TY - JOUR
AU - Rigby, David, L
AB - Abstract Explanations for why firms in some clusters outperform others rest on the assumed benefits of local and nonlocal interaction. In this article, we extend research on knowledge sharing by modeling local and global interactions between firms distributed across industrial clusters. Our simulation model develops an evolutionary framework where firms explore and exploit knowledge sets accumulated over time by recombining technologies held by local and nonlocal partners. In investigating the opportunity cost structure of local and nonlocal interaction, our simulation raises two important questions that call for additional theoretical and empirical analysis. First, can too much local interaction induce technological lock-in and restrict innovation in clusters? Second, does nonlocal interaction entail opportunity costs to clusters that can outweigh its benefits? 1. Introduction The long-run fortunes of cities and regions rest on the development of new technologies and on the spatial distribution of the resulting rents. Individual firms still dominate these processes, competing over the acquisition of knowledge and the translation of that knowledge into profitable technologies (Kogut and Zander, 1992; Teece and Pisano, 1994). The generation of new technologies is broadly viewed as a process of recombining existing ideas (Kauffman, 1993; Fleming and Sorenson, 2001), and thus the size and heterogeneity of knowledge stocks are critical drivers of the pace and the direction of technological change (Rigby and Essletzbichler, 2006). Because the breadth of knowledge that individual firms can internalize is limited, invention increasingly relies on collaboration between economic agents (Owen-Smith and Powell, 2004). It is clear from research on agglomeration and clusters that firms benefit in many different ways by their interaction with one another (Marshall, 1920; Duranton and Puga, 2004). Local interaction, or buzz, precipitated by dense clusters of co-located economic agents, is generally seen as an engine of innovation (Saxenian, 1994; Storper and Venables, 2004; Storper et al., 2015), though it may also be a harbinger of technological lock-in and decay (Grabher, 1993; Hassink, 2010). Nonlocal interaction, the knowledge pipelines that connect economic agents between clusters, is regarded as a solution to technological stagnation, replenishing place-bound stocks of ideas and refueling economic growth (Bathelt et al., 2004). Though broadly supportive of these claims, we suggest that they are deployed rather too generally such that the conditions under which different forms of interaction confer benefits on cluster members remain unclear. In this article, we develop a simulation model to explore the mechanisms through which global pipelines and local buzz shape the innovativeness of local clusters. We find that neither buzz nor pipelines are unconditionally good. With respect to buzz, the results from our simulation model show that clusters containing firms that share some knowledge outperform clusters with independent firms, but that technological growth declines when firm boundaries largely dissolve. In other words, too much buzz can ferment the negative technological lock-in of clusters. With respect to pipelines, our model indicates that pipelines expand the knowledge heterogeneity embodied within clusters and raise aggregate performance. However, we also find that there are costs to clusters whose agents interact with partners elsewhere and that these costs can exceed the benefits of nonlocal interaction. Together, these results raise important questions as to the conditions under which buzz and pipelines enhance innovation and when they may fail to generate the dynamism frequently suggested by the literature. We employ a simulation model to explore the costs and benefits of local and nonlocal interaction because it allows precise control over the environment within which interaction occurs. In this sense, simulation aids theory construction (Davis et al., 2007). Given the stochastic nature of recombinant knowledge production that we employ in our model, the heterogeneity of economic agents and the clusters within which they operate, analytical solution to the evolutionary framework we deploy is impractical (Nelson and Winter, 1982). Our model is not offered as an alternative to empirical work, rather it is developed as a complement, permitting us to control relationships that cannot easily be isolated in empirical settings while also suggesting novel areas for further empirical research (Chandrasekaran et al., 2016). In the discussion below, we provide a brief sketch of a literature that views technological change as driven by the interaction of agents. We review geographical extensions of these claims that highlight the spatially differentiated character of regional knowledge pools and of interaction within and across those pools. Thereafter, we change gears to discuss the structure of a simulation model developed to address the questions raised above. We first discuss the simulation informally and then shift to more formal presentation. The results from the simulation are discussed at length and we conclude with our main findings, prompting considerations of questions for further research within the buzz and pipelines tradition. 2. Literature review 2.1. A recombinant model of technological change Invention, or knowledge creation, is increasingly imagined as a process of recombination, of new technologies constructed by combining existing ideas (Arthur, 2009). Limits on the possibilities of recombination are linked to the architecture of knowledge by Simon (1962). Integrating recombinant invention in a model of endogenous growth, Weitzman (1998) shows that the number of idea combinations rapidly overwhelms the processing capacity of individual economic agents. Wuchty et al. (2007) builds on this argument to illustrate the rising importance of collaboration in knowledge generation. The growing heterogeneity of knowledge stocks and the rise of their complexity make the long-run management of technology precarious. Individual firms have limited capacity to effectively manage those stocks and so specialize, organizing to exploit and extend the knowledge embodied within their workers and routines while guarding it from others (Kogut and Zander, 1992; Grant, 1996; Pavitt, 1999). Within this environment, knowledge production increasingly involves the combination of ideas that move across the boundaries of individual firms. For some (Powell et al., 1996; Dyer and Singh, 1998; Kogut, 2000), competitive advantage resides in distributed network resources and routines that emerge from cooperative practices. Whether such linkages are formal (strategic alliances, supply-chain partnerships or R&D joint ventures) or informal (a geographical or technological community), there is broad agreement on the generation of benefits, though much less on what forms of interaction maximize returns (Ahuja, 2000; Chesbrough, 2003; Freitas et al., 2011). While the growing complexity of knowledge has increased the importance of interaction, complex knowledge is difficult to exchange without the aid of face-to-face communication (Gertler, 2003; Storper and Venables, 2004). As evidenced by patent citation records, the interactions capable of transmitting complex knowledge tend to occur between agents that reside in the same city (Jaffe et al., 1993; Sonn and Storper, 2008; Balland and Rigby, 2017; though see Breschi and Lissoni, 2009). However, face-to-face communication is not necessarily trapped in specific locations, and complex knowledge can defy geography if the social conditions are right (Boschma, 2005; Maskell et al., 2006). The nature and relative importance of local and nonlocal interaction for the process of innovation are frequently debated under the headings of buzz and pipelines. 2.2. Buzz and pipelines While a broad base of research illustrates the benefits of co-location on firm performance (Glaeser et al., 1992; Audretsch and Feldman, 1996), it is also increasingly well understood that forms of institutional and technological lock-in can weaken the competitive fortunes of local clusters and threaten their long-run resilience (Grabher, 1993; Hassink, 2010; Simmie and Martin, 2010; Rodriguez-Pose, 2013). Responding to a lack of theoretical clarity on the structure of knowledge and its movement, and to mixed evidence regarding the dominance of local over nonlocal knowledge flows, Bathelt et al. (2004) explore forms of knowledge-based interaction between firms that are distributed across spatial clusters. Within the cluster, they develop the notion of local buzz to reinforce the idea that rich communication ecologies may develop within local settings that enhance understanding, interaction and knowledge production (Bathelt and Gluckler, 2011). However, they also make clear that local buzz is insufficient to maintain competitive advantage in isolated clusters of economic agents. The most vibrant clusters are those containing firms that build explicit linkages, or pipelines, to partners and differentiated knowledge resources that are embedded within extra-local clusters (Maskell et al., 2006). The ‘buzz–pipeline model’ has prompted significant conceptual debate around the characterization of buzz (Asheim et al., 2007) and proximity (Boschma, 2005; Torre and Rallet, 2005), on forms of interaction (Moodysson, 2008) and on the nature of innovation across different knowledge bases (Asheim and Coenen, 2005). Buzz–pipelines claims may also be linked to a new round of empirical work exploring the geography of interaction and, in particular, the importance of local and nonlocal linkages in knowledge creation. Tödtling et al. (2006) find broad support for the arguments of Bathelt et al. (2004), noting that Austrian industries with different knowledge bases use distinct forms of technological inputs that are gathered across geographical scales. The importance of knowledge bases to understand the geography of biotech activity in part of Sweden is echoed by Moodysson (2008). In more recent work on the Vienna software cluster, Trippl et al. (2009) burrow more deeply into the nature of knowledge-based interaction and its geography. They find the buzz and pipelines model too restrictive, noting that informal network linkages are critical to the firms they sampled and that such links are found at all spatial scales, not just locally. They also report the importance of formal partnerships at both local and national levels. Fitjar and Rodriguez-Pose (2011,, 2013,, 2017) report somewhat different results on the basis of firm surveys in Norwegian city-regions. Focusing on formal interactions between economic agents, they show little evidence of local/regional interaction supporting firm innovation. It should be clear that there is considerable heterogeneity in the character of knowledge-based linkages between firms, the geography of such linkages and their impact on innovation. At the same time, there is strong support for the idea that local and nonlocal forms of interaction benefit the firms involved. For us, however, the most interesting question is not whether partners benefit from interaction, but whether interactions between firms generate returns to clusters that exceed the cost of such interactions. This question lies at the core of the buzz–pipelines framework, yet it is a question that has received relatively little attention, save the important work of Giuliani and Bell (2005) and Morrison et al. (2013). Heiman and Nickerson (2002) and Tether (2002) argue that technological search and interaction are associated with costs as well as benefits. Developing new technological possibilities in-house or through interactions with other firms are resource using activities that carry significant opportunity costs. Even buzz, imagined in restricted form as local, informal relationships, cannot be free, for the cost of scanning within the cluster implies fewer interactions of other kinds. In his analysis of embeddedness on economic performance, Uzzi (1996, 1997) argues that one cost of inefficient search is reduced access to novel technological possibilities. Giuliani (2007) too, argues that networking is time-consuming and costly, drawing, in part, on the empirical findings of Schrader (1991). Gilsing et al. (2011) and Tripsas et al. (1995) discuss the cost of interaction in the form of barriers to technology transfer and the need for governance in R&D networks to discourage opportunistic behavior. Greater local interaction may also not always benefit the cluster. At first blush, it would seem that frequent local interaction enables the recombination of diverse ideas held by individual firms. Following the same logic, closed firm borders should restrict local recombination. However, if the frequency of interaction within a cluster and the extent of local knowledge heterogeneity are not independent of one another, high levels of local interaction may cause firms to develop overlapping knowledge stocks, reduce knowledge heterogeneity and slow the production of knowledge at the local level (see also Cowan and Jonard, 2003 and Morrison et al., 2013). In general, the cluster literature assumes that greater levels of local interaction between economic agents are beneficial. However, might local interaction also be capable of reducing the buzz of diverse ideas and generating technological lock-in? This concern extends the process identified by Grabher (1993) and motivates our first question: Question 1: Ceteris paribus, can local openness and firm interaction be too intense, to the point that it reduces the local heterogeneity of knowledge and thus dampens recombinant knowledge production within a cluster? Although currently the buzz and pipelines model is relatively silent on the cluster-level opportunity costs associated with local and nonlocal interaction, from the arguments above it seems untenable to assume that these costs are zero. With respect to nonlocal interaction, these costs are created by the diversion of attention away from the local cluster. Because the opportunity cost of nonlocal interaction is reduced local interaction, the magnitude of its cost should depend on the normal level of local interaction within the cluster. These arguments lead us to question 2. Question 2: Do pipelines confer similar costs and benefits on agents located in clusters with different intensities of local interaction? For which strengths of local interaction do the benefits of pipelines exceed the costs, and for which strengths of local interaction do the costs exceed the benefits? In the simulation model outlined below, we do not explicitly distinguish between formal or informal interactions in our consideration of local and nonlocal linkages. That said, we have chosen to frame nonlocal interaction as an ephemeral cluster (a tradeshow) that is relatively short-lived. It is likely that formal linkages between nonlocal partners may generate larger benefits than those obtained through informal linkages. At the same time, extended nonlocal linkages may generate additional costs to the home clusters of participants. Currently, the conceptual and empirical literature on buzz and pipelines does not formally explore the net effect of nonlocal interaction on clusters. 3. Informal discussion of the simulation model 3.1. Introduction We use a simulation model to examine the questions raised above regarding local and nonlocal interaction in a formal setting. In this section of the article, we walk through the set-up of the simulation model, presenting a more formal explication of that framework in Section 4. Firms interact locally and nonlocally by sharing knowledge stocks. In a first version of the simulation (Version 1A), we vary the amount of interaction that occurs within each cluster and explore over a series of time-steps how interaction affects the total volume of knowledge produced at the cluster level. In a second version (Version 2A), one firm from each cluster is selected as a participant in a global pipeline, modeled as a tradeshow, where it interacts with nonlocal firms and accesses nonlocal knowledge that it later brings to its home cluster. Whether or not that nonlocal knowledge is distributed in the home cluster depends on the level of local interaction. We thus explore the joint influence of local and nonlocal interaction on cluster knowledge production. We do not explore variations in institutional or organizational proximity, other than assuming that the borders of firms are more or less porous from one cluster to the next, and so allow varying amounts of local knowledge sharing. More broadly, institutional and organizational factors may be regarded as fixed and do not influence the processes examined. We also take spatial proximity, the spatial clustering of firms into clusters, as a fixed condition. Thus, the results of the simulation are driven solely by varying the propensity for firms to interact with one another, either locally or nonlocally. 3.2. Firms Firms compete in a single market by producing a range of goods that are linked directly to the different kinds of knowledge that they possess. In order to focus the simulation on the supply-side dynamics of the economy, we assume that demand is equally robust for all types of outputs. Therefore, the prices of outputs are determined by supply conditions alone. Outputs that are short in supply command high prices and monopolistic rents, while outputs that are supplied in abundance command low prices. Firms utilize the individual components of their knowledge stocks by competing in one of two ways. In a first form of competition, they exploit existing knowledge by utilizing it to produce a specific range of goods. Firms choose to exploit knowledge when they expect the resulting output will generate above average returns.1 In a second form of competition, when firms have technologies that are used to produce goods with expected prices below the average, firms remove these technologies from the market and use them for exploration. The process of exploration involves recombining all pairs of technologies known to the firm that are not used to produce existing goods in the market. New technologies generated through exploration are used by firms to produce new goods that are later introduced to the economy. We build our analysis around an evolutionary simulation model of recombinant technological change to help make clear how technology-based interaction impacts the heterogeneity of cluster knowledge stocks and their growth. Each period firms reinvest revenues in search for additional profit. Firms with knowledge types that yield high rents increase their productive capacity in the same knowledge subsets. We assume that once a firm understands a particular technology, it can expand use of that technology through reinvestment: the firm does not need to invent that technology again. However, over time the rents associated with existing technologies fall as the dated knowledge diffuses to other firms and clusters. When a technology becomes widely exploited, the associated good tends to generate below average returns and this induces further exploration. 3.3. Invention through buzz, or local interaction The first version of our simulation model (Version 1A) investigates how the intensity of local interaction influences knowledge production, firm growth and technological heterogeneity. We begin with a simulation universe with five clusters, each containing five firms. Within each of those clusters, firms are endowed with a common propensity to interact locally. Across clusters, that propensity varies. In cluster 1, we exogenously maximize the amount of local interaction between firms by setting the degree of local interaction (local interaction strength, or LIS) equal to one. In this hypothetical cluster, firm borders are open so all firms in the cluster have access to the knowledge stocks of their local partners that are used for exploration, along with their own knowledge stocks designated for exploration. We endow the second cluster with slightly weaker interaction (LIS = 0.75), where 75% of firm technologies that are used for exploration spill over between firms. Firms in the third cluster (LIS = 0.5) recombine half of their exploring technologies locally and the other half internally, and so on, until the fifth cluster, which we endow with LIS = 0. In this last cluster, firm borders are effectively closed and no knowledge flows between local firms. Figure 1 shows how LIS factors into the recombination of technologies between firms in the same cluster. Ties drawn between technologies indicate hypothetical patterns of recombination that link subsets of the knowledge stocks within and between firms. In the first cluster, where LIS = 1, an equal number of recombinant ties are made between firms as within firms. In the second cluster, where local interaction (LIS) is weaker, there are many more within-firm recombinant ties than between-firm ties.2 Figure 1 View largeDownload slide Recombination in clusters with varying LIS. Invention of technologies (nodes) through recombination (edges) within and between firms (large circles). Local interaction is stronger in the cluster to the left, so more edges span the two firms. Letters indicate the qualitative content of each idea. All A’s are identical. Recombination of identical technologies produces no new technological variety (see Section 4.2). Figure 1 View largeDownload slide Recombination in clusters with varying LIS. Invention of technologies (nodes) through recombination (edges) within and between firms (large circles). Local interaction is stronger in the cluster to the left, so more edges span the two firms. Letters indicate the qualitative content of each idea. All A’s are identical. Recombination of identical technologies produces no new technological variety (see Section 4.2). As we start the simulation model in time period 1, each firm is endowed with the same technologies. The simulation model is run multiple times for many time periods. We examine how cluster performance varies with the strength of local interaction. Performance is measured by calculating the quantity of technologies present in each cluster, the range or diversity of those technologies and their technological rents or value in terms of ubiquity. 3.4. Invention through pipelines, or nonlocal interaction We develop a second simulation model (Version 2A) in which we introduce pipelines to explore how global interaction, or interaction between firms from different clusters, influences the relationship between LIS and cluster performance. In the pipelines version of the simulation, a ‘tradeshow’ is held during each model run. We model pipelines as tradeshows because it is the simplest way to capture the effects of global knowledge transfer on the knowledge heterogeneity of clusters, and because it eliminates the need to examine how firms choose partners (Maskell et al., 2006). We acknowledge that our representation of pipelines as tradeshows is limited. In particular, it does not allow us to explore the various effects of strategies used by firms when they decide whom to interact with bilaterally. While the tradeshow is in session, one firm from each cluster halts interaction within its own cluster and turns its attention to nonlocal tradeshow partners. Firms participating in the tradeshow share and learn from the other participants as though they are members of a temporary economic cluster.3 When the tradeshow ends, participants resume local interaction and nonlocal partnerships dissolve. A simplified tradeshow, with only two clusters of two firms each, is depicted in Figure 2. Similar to the simulation model of Morrison et al. (2013), we assume that participation in the tradeshow occurs at the cost of local interaction such that pipeline participation is not free either to the firms directly involved or to their local partners. This is one of the most critical arguments of our simulation framework. While firms may participate in local interaction as well as nonlocal interaction, with limited resources they cannot participate in both forms of interaction with the same level of intensity. For simplicity, we have chosen to capture this resource constraint by making local and nonlocal interaction mutually exclusive. Again, this simplified rendering of tradeshows has the tradeoff in that it does not allow us to explore the strategic behavior of firms when they choose how much to interact and with whom to interact. Figure 2 View largeDownload slide Recombination with pipelines. Figure 2 View largeDownload slide Recombination with pipelines. We run the two versions of the simulation model (Versions 1A and 2A) and explore variations in the values of key parameters for both sets of models. Those parameter variations, capturing changes in the total time horizon (simulation length), the timing of the tradeshow, the number of clusters, the number of firms per cluster and the strength of local interaction (LIS) at the tradeshow, allow the robustness of our core arguments to be examined. The parameter sets are given in Table 1 and the results in Table 2. Comparison of the local interaction models (Versions 1A–1J) with the local and nonlocal interaction models (Versions 2A–2J) reveals the opportunity costs and the benefits of both forms of interaction. We are less interested in the firm-level benefits of direct interaction, an issue which is well-examined, both theoretically and empirically, in the literature noted above. Rather, our goal is to understand whether greater openness leads to more efficient knowledge production in local clusters and whether nonlocal knowledge flows always lead to greater knowledge production at the level of the cluster. Table 1 Simulation parameters Parameter Version 1 Version 2 Number of clusters 3, 5, 8 3, 5, 8 Number of firms per cluster 3, 5, 8 3, 5, 8 Number of firms in tradeshow — 3, 5, 8 Time horizon 30, 40, 50 30, 40, 50 Tradeshow time periods — 20–25, 25–30, 30–35 LIS in cluster [1, .5, 0] [1, .5, 0] [1, .75, .5, .25, 0] [1, .75, .5, .25, 0] [1, .86, .71, .57, .43, .29, .14, 0] [1, .86, .71, .57, .43, .29, .14, 0] Tradeshow cluster interaction strength — 0.5 ρ (risk aversion) 95% 95% Number of model runs 5000 5000 Parameter Version 1 Version 2 Number of clusters 3, 5, 8 3, 5, 8 Number of firms per cluster 3, 5, 8 3, 5, 8 Number of firms in tradeshow — 3, 5, 8 Time horizon 30, 40, 50 30, 40, 50 Tradeshow time periods — 20–25, 25–30, 30–35 LIS in cluster [1, .5, 0] [1, .5, 0] [1, .75, .5, .25, 0] [1, .75, .5, .25, 0] [1, .86, .71, .57, .43, .29, .14, 0] [1, .86, .71, .57, .43, .29, .14, 0] Tradeshow cluster interaction strength — 0.5 ρ (risk aversion) 95% 95% Number of model runs 5000 5000 Note: Parameters for the base models (Versions 1A and 2A) in bold. View Large Table 1 Simulation parameters Parameter Version 1 Version 2 Number of clusters 3, 5, 8 3, 5, 8 Number of firms per cluster 3, 5, 8 3, 5, 8 Number of firms in tradeshow — 3, 5, 8 Time horizon 30, 40, 50 30, 40, 50 Tradeshow time periods — 20–25, 25–30, 30–35 LIS in cluster [1, .5, 0] [1, .5, 0] [1, .75, .5, .25, 0] [1, .75, .5, .25, 0] [1, .86, .71, .57, .43, .29, .14, 0] [1, .86, .71, .57, .43, .29, .14, 0] Tradeshow cluster interaction strength — 0.5 ρ (risk aversion) 95% 95% Number of model runs 5000 5000 Parameter Version 1 Version 2 Number of clusters 3, 5, 8 3, 5, 8 Number of firms per cluster 3, 5, 8 3, 5, 8 Number of firms in tradeshow — 3, 5, 8 Time horizon 30, 40, 50 30, 40, 50 Tradeshow time periods — 20–25, 25–30, 30–35 LIS in cluster [1, .5, 0] [1, .5, 0] [1, .75, .5, .25, 0] [1, .75, .5, .25, 0] [1, .86, .71, .57, .43, .29, .14, 0] [1, .86, .71, .57, .43, .29, .14, 0] Tradeshow cluster interaction strength — 0.5 ρ (risk aversion) 95% 95% Number of model runs 5000 5000 Note: Parameters for the base models (Versions 1A and 2A) in bold. View Large Table 2 Results of number of technologies by LIS and model number and name Model number and name LIS 1 0.75 0.5 0.25 0 Local interaction models 1A Base no tradeshows 256 263 245 204 75 1B Shorter time horizon 209 217 210 182 85 1C Longer time horizon 303 309 295 224 65 Local and global interaction models 2A Base with tradeshows 252 257 248 202 80 2B Shorter time horizon 209 211 205 183 90 2C Longer time horizon 298 302 286 223 72 2D Earlier tradeshow 253 257 246 205 79 2E Later tradeshow 252 254 246 201 81 2F Stronger TS interaction 252 258 240 208 81 2G Weaker TS interaction 250 251 245 204 80 2H No opportunity cost 262 265 254 207 80 Three firms, three clusters LIS 1 0.5 0 1I Base without tradeshows 93 94 72 2I Base with tradeshows 93 94 73 Eight firms, eight clusters LIS 1 0.86 0.71 0.57 0.43 0.29 0.14 0 1J Base without tradeshows 527 535 483 416 359 280 216 75 2J Base with tradeshows 511 488 447 425 364 282 228 85 Model number and name LIS 1 0.75 0.5 0.25 0 Local interaction models 1A Base no tradeshows 256 263 245 204 75 1B Shorter time horizon 209 217 210 182 85 1C Longer time horizon 303 309 295 224 65 Local and global interaction models 2A Base with tradeshows 252 257 248 202 80 2B Shorter time horizon 209 211 205 183 90 2C Longer time horizon 298 302 286 223 72 2D Earlier tradeshow 253 257 246 205 79 2E Later tradeshow 252 254 246 201 81 2F Stronger TS interaction 252 258 240 208 81 2G Weaker TS interaction 250 251 245 204 80 2H No opportunity cost 262 265 254 207 80 Three firms, three clusters LIS 1 0.5 0 1I Base without tradeshows 93 94 72 2I Base with tradeshows 93 94 73 Eight firms, eight clusters LIS 1 0.86 0.71 0.57 0.43 0.29 0.14 0 1J Base without tradeshows 527 535 483 416 359 280 216 75 2J Base with tradeshows 511 488 447 425 364 282 228 85 Note: Results from models using the base parameter sets in bold. View Large Table 2 Results of number of technologies by LIS and model number and name Model number and name LIS 1 0.75 0.5 0.25 0 Local interaction models 1A Base no tradeshows 256 263 245 204 75 1B Shorter time horizon 209 217 210 182 85 1C Longer time horizon 303 309 295 224 65 Local and global interaction models 2A Base with tradeshows 252 257 248 202 80 2B Shorter time horizon 209 211 205 183 90 2C Longer time horizon 298 302 286 223 72 2D Earlier tradeshow 253 257 246 205 79 2E Later tradeshow 252 254 246 201 81 2F Stronger TS interaction 252 258 240 208 81 2G Weaker TS interaction 250 251 245 204 80 2H No opportunity cost 262 265 254 207 80 Three firms, three clusters LIS 1 0.5 0 1I Base without tradeshows 93 94 72 2I Base with tradeshows 93 94 73 Eight firms, eight clusters LIS 1 0.86 0.71 0.57 0.43 0.29 0.14 0 1J Base without tradeshows 527 535 483 416 359 280 216 75 2J Base with tradeshows 511 488 447 425 364 282 228 85 Model number and name LIS 1 0.75 0.5 0.25 0 Local interaction models 1A Base no tradeshows 256 263 245 204 75 1B Shorter time horizon 209 217 210 182 85 1C Longer time horizon 303 309 295 224 65 Local and global interaction models 2A Base with tradeshows 252 257 248 202 80 2B Shorter time horizon 209 211 205 183 90 2C Longer time horizon 298 302 286 223 72 2D Earlier tradeshow 253 257 246 205 79 2E Later tradeshow 252 254 246 201 81 2F Stronger TS interaction 252 258 240 208 81 2G Weaker TS interaction 250 251 245 204 80 2H No opportunity cost 262 265 254 207 80 Three firms, three clusters LIS 1 0.5 0 1I Base without tradeshows 93 94 72 2I Base with tradeshows 93 94 73 Eight firms, eight clusters LIS 1 0.86 0.71 0.57 0.43 0.29 0.14 0 1J Base without tradeshows 527 535 483 416 359 280 216 75 2J Base with tradeshows 511 488 447 425 364 282 228 85 Note: Results from models using the base parameter sets in bold. View Large 4. Formal discussion of the simulation model 4.1. Knowledge and production of outputs We now introduce the specifics of the simulations. Each firm is described by a vector in which the elements record the amount of knowledge of each type the firm knows. Knowledge types translate directly into output. Specifically, to produce output of type a, we assume a one-to-one production function so that firms compete entirely on the basis of their knowledge portfolio. The production function is given by Outputa,f=Knowledgea,f Firms sell their outputs in the global market where they compete with the outputs supplied by all other firms. The price that firms receive for a given output is determined by the relative supply and demand for that output. To simplify matters, we assume that demand is uniformly distributed across all outputs.4 Therefore, the price of an output is determined by the quantity supplied by all firms. Because the price of a is a decreasing function of the quantity supplied, we write its price as the inverse of its quantity: Pricea,t=1Quantitya,t Firms may control more than one unit of each type of technology, so the total revenue generated by a given technology for a firm is given by the market price of the outputs scaled by the quantity of the technology controlled by the firm. Revenuea,f,t=Pricea,t*Quantitya,f,t Each period, firms reinvest their revenues, increasing or decreasing the quantity of each knowledge type that they hold. We assume that firms must pay a cost to expand their knowledge-based productive capacity. The cost of increasing capacity is assumed constant for all goods and is given by the average price of all goods from the previous time period. Costt=∑anQuantitya,t-1*Pricea,t-1∑anQuantitya,t-1 To make things a little more concrete, imagine that firms use knowledge along with a generic input, say labor, to produce outputs. In time period t, a firm purchases a number of units of this generic input. The volume of such inputs is given by the firm’s revenue from period t − 1 and the cost of the input, a general wage in the case of labor, which is equal to the average price of all goods in period t − 1. Once hired, the generic inputs of labor utilize the different technologies in the firm, absorbing knowledge and thus becoming units of technology-and-output specific labor. Thus, through hiring and firing labor, firms increase and decrease the quantity of outputs they produce. Workers are paid the same wage regardless of the technology employed: all technology rents accrue to the firm.5 To keep firm-level agency straightforward, we assume that firms reinvest their revenues generated by one type of technology in that same technology. Therefore, the quantity of a that firm f will control in the next time period is given by the revenue it generated from a divided by the numeraire cost of generic inputs: Quantitya,f,t+1 = Revenuea,f,tCostt The above equation allows firms to adjust their knowledge structures based on market prices. Those stocks increase without friction for goods whose price is above the average and they decrease without friction for goods whose price is below the average, which would lead to a rapid convergence of prices for all types of outputs. We therefore introduce friction by making Quantitya,f,t+1 a function of how much a the firm already knows Quantitya,f,t+1 = 12Quantitya,f,t +Quantitya,f,t+1  It is important to recognize that by assuming all firms use the same generic inputs, we consciously downplay the importance of cost-based forms of competition. As a consequence, firms gain competitive advantage by developing novel technologies through recombination. Knowledge stocks are held by individual firms, though portions of those stocks are shared among members of local and nonlocal clusters. Therefore, the network structures of interaction are the critical determinant of cluster performance. Clusters with firms that are able to access the most heterogeneous knowledge stocks are those that are most likely to succeed. New technologies, by virtue of the model’s dynamics, tend to be relatively scarce, generate above average returns and growth in the short-run and increase technological variety and the possibility of higher levels of recombinant interaction in the long-run. 4.2. Exploitation and exploration Firms continue to use, or exploit, technologies so long as the resulting outputs command market prices above the average. The exploitation of scarce technologies creates Schumpeterian dynamics: firms exploit technologies that are relatively scarce and profitable, but in doing so increase their ubiquity thereby eroding technological rents. Declining rents incentivize firms to explore and recombine technologies that have become relatively ubiquitous. Firms explore with the technologies that they do not exploit through recombination. For example, a firm exploring with technologies a and b invents the new technology ab. As firms continue to recombine technologies, they invent more complex hybrid technologies. The recombination of technologies a and ab produces the technology aab. Because aab is a new technology, it is unlikely to be known by other firms and thus commands high rents. The inventing firm may be able to exploit and grow from the production of aab until other firms learn this technology, the technology’s ubiquity increases, and falling rents spur the search for new technology. Over time, recombination gives rise to a tree-like structure of knowledge. As firms continue to explore and recombine subsets of knowledge, new recombinations are formed and the tree grows increasingly large and complex. Because recombination results in a rapidly growing set of possible technologies, we introduce a parameter ρ that can be interpreted as an exogenous, uniform degree of risk aversion, associated with the possibility of exploration search failure. The primary purpose of ρ is to curtail the amount of successful exploration that occurs in the simulation. Because the value ρ is constant across all clusters and all firms, it does not bias the key findings in our results. Lastly, we discuss how firms recombine technologies. When a firm recombines, it produces all possible exploring technology pairings. For example, a firm f in cluster c exploring with one a and one b explores with the exploring knowledge vector Exploring Knowledgef,c,t=1af,c,t;1bf,c,t Recombining a with b generates the possible pairs vector.6 Possible Pairsf,c,t=2abf,c,t;1aaf,c,t;1bbf,c,t Because each exploring technology pairing should produce exactly one new unit of technology, we normalize the possible pairs vector by the sum of the exploring knowledge vector to arrive at the realized pairs vector: Realized Pairsf,c,t=1abf,c,t;0.5aaf,c,t;0.5bbf,c,t The realized pairs vector contains two technologies (aa and bb), which have added no new variety to their base technologies, a and b. We therefore condense to Realized Pairsf,c,t=1abf,c,t;0.5af,c, t;0.5bf,c,t The exploration process concludes once firms arrive at the final realized pairs vector. This vector represents the new technologies that the firm has added through exploration. Together with the older technologies the firm has retained through exploitation, the firm will use these technologies to produce outputs and generate rents in the next time period. 4.3. Local and nonlocal interaction in the process of exploration We exogenously fix the degree of local interaction in each cluster. If a firm explores with the exploring vector Exploring Knowledgef,c,t=1af,c,t;1bf,c,t it divides its exploring technologies between two vectors, an internal exploring knowledge vector and a local exploring knowledge vector based on its cluster’s LIS. LIS ranges from 0 to 1. Internal Exploring Knowledgef,c,t=1af,c,t;1bf,c,t*(1-LISc) Local Exploring Knowledgef,c,t=1af,c,t;1bf,c,t*(LISc) The firm’s local exploring knowledge vector contains the firm’s contribution to the buzz in a cluster. This is a set of technologies that all firms in the cluster may explore with. Thus, when firm f explores locally, it recombines technologies in its own local exploring knowledge vector with the locally shared technologies contained in the cluster exploring knowledge vector, given by Cluster Exploring Knowledgec,t=∑fnaf,c,t;∑fnbf,c,t*(LISc) Local exploration otherwise occurs identically to internal firm exploration. When a firm participates in a tradeshow, LIS at the tradeshow is set to the middle value, 0.5. This captures the idea that the firm attends a tradeshow with the purpose of exchanging knowledge, yet it remains guarded in terms of openness. We vary this parameter in our robustness checks. Tradeshow-based interaction otherwise occurs identically to interaction in at-home clusters. 4.4. Collection of results We develop two versions of the simulation: a simulation with only local interaction (Version 1A), and a simulation with both local interaction and nonlocal interaction (Version 2A). To generate our results, we record the number of technologies in each cluster at the conclusion of each run of the simulation. Each simulation run lasts 40 periods of time, though we explore variations in the time horizon in robustness checks. We run each version many times (5000 repetitions) in order to smooth the stochasticity introduced by the rho (⁠ ρ ⁠) term. At the conclusion of the 5000 repetitions, we take the mean number of technologies generated in each cluster and we report these values in Table 2 as the primary indicator of cluster performance. To examine the robustness of our findings, we experiment with different values of key parameters. We augment the number of clusters in our universe, the number of firms in each cluster, the length of the time horizon, the time period of the tradeshow and the LIS at the tradeshow. We change the parameters one at a time, with the exception of changes in the number of clusters and the number of firms. We adjust these two parameters in tandem: the three-cluster simulation contains three firms per cluster, and the eight-cluster simulation contains eight firms per cluster. We do this to keep the tradeshow cluster size consistent with the size of regional clusters when the tradeshow is not in session. We depict the parameter sets of the different simulations in Table 1, with the parameter values of our base versions (Versions 1A and 2A) given in bold. 5. Results We present our core results in Table 2. The derivatives of simulation Version 1 of the simulation model focus only on local interaction; they do not involve a tradeshow. The derivatives of Version 2 explore the combination of local interaction together with a tradeshow. The letter following the simulation version numbers indicates the parameter sets used. Versions can be compared across numbers (1 and 2) to explore the contribution of nonlocal interaction to local interaction and cluster performance, and across letters (A through J) to explore how changes in parameters impact the results. 5.1. Results: local and nonlocal interaction No-tradeshow versions of the simulation model, versions 1A–1J, show that the innovativeness of clusters has an inverse-U shaped relationship with the strength of local interaction. The average number of technologies in clusters peaks when LIS equals 0.75. Local interaction beyond this level of strength is associated with reduced innovation: in the base version (Version 1A), moving from an LIS of 0.75 to 1 is associated with a reduction from 263 to 257 technologies in the cluster, a 2.7% decline. The simulation results answer our first research question: too much local interaction, in the absence of nonlocal interaction, negatively impacts cluster performance. Results from the simulations also make clear that participation in pipelines does not always increase the innovativeness of clusters from which tradeshow participants originate. For clusters with specific levels of LIS, comparing the simulation models with and without pipelines shows that the number of technologies in a cluster is often, though not always, higher without tradeshow participation. This result is driven largely by the cost of pipeline participation that reflects lower levels of local interaction within clusters that contain fewer firms while the tradeshow is in progress and answers part of our second research question: nonlocal interaction does not always raise the innovative performance of clusters. Versions 2A–2J of the simulation results in Table 2 show that the negative impacts of excessive local interaction are mitigated in clusters with nonlocal pipelines. In the base model (Version 2A) moving from an LIS of 0.75 to 1 decreases the number of technologies in the cluster by 1.9%. Thus, while clusters with an LIS of 1 are less innovative than clusters with an LIS of 0.75 even when they have a firm participating in the tradeshow, the negative impact of very strong local interaction is less than in the scenario with no tradeshow. The mitigating effect of tradeshows is robust across each model version and is strongest in the model containing eight clusters of eight firms, (Version 2J), where the introduction of nonlocal interaction causes technological growth to increase monotonically with LIS. Overall, these findings also speak to our second research question: buzz and pipelines are complementary in that the negative effects of too much local interaction can be reduced through building nonlocal partnerships. 5.2. Interpretation: micro-mechanisms of the costs of local and nonlocal interaction An important and surprising result that emerges from our simulation model is that very strong local interaction can be detrimental to the technological growth of a cluster. To investigate why this situation arises, we develop a simplified variant of the model to illustrate the impacts of very strong local interaction on the technological heterogeneity of a cluster. This model contains two clusters, each with two firms. The first cluster serves as a control and has local interaction set to a moderate value (0.5) for the full duration of the model. The second cluster begins with an LIS of 0.5 and is ‘treated’ by increasing its LIS to 1.0 when t = 25. Each time period, we calculate the cognitive distance between firm i and firm j in each cluster using the standard Euclidean distance formula to explore how technological heterogeneity evolves before and after we administer the treatment in cluster two. We run this simplified model 15,000 times and average the results across model runs. The results of our simplified simulation model are presented in Figure 3. While the average technological distance between the firms of each cluster, or their technological heterogeneity, is identical until the treatment is applied to the second cluster at t = 25, the treatment (denoted by a star) causes a relative decrease in the cognitive distance between its firms. Thus, we conclude that when the strength of local interaction is so high that it eliminates the ability of firms to shield some technologies from their neighbors, the knowledge stocks of local firms begin to converge and isomorphism sets in (DiMaggio and Powell, 1983). Figure 3 View largeDownload slide Controlled experiment of local interaction and heterogeneity. Technological distance between firms as an indication of technological heterogeneity is measured using Euclidean distance (see Section 5.2). Figure 3 View largeDownload slide Controlled experiment of local interaction and heterogeneity. Technological distance between firms as an indication of technological heterogeneity is measured using Euclidean distance (see Section 5.2). The second key finding from our simulation is that pipelines do not unequivocally increase the innovativeness of their anchor clusters. To better understand the micro-level processes behind this outcome, we develop a version of the simulation with fewer moving parts to look at the size and the technological diversity of individual firms before, during and after the tradeshow. The simplified simulation model has three clusters, each with an LIS of 0.5. Each cluster has three firms, and one firm in each cluster participates in the tradeshow. We calculate the technological heterogeneity of individual firms, and we calculate firm size, again measured as the number of technologies they utilize, to explore how technological diversity translates into innovative growth. In Tables 3 and 4, we report the evolution of firms’ technological heterogeneity and size for two periods of time: between the beginning of the tradeshow and the end of the tradeshow (t = 24 to t = 30), and between the end of the tradeshow and the conclusion of the model (t = 31 to t = 40). Results for technological heterogeneity are given in Table 3 while results for technological size are given in Table 4. Table 3 Tradeshow effects on firm-level technological heterogeneity Firm participated in TS? Yes No Time period measured Beginning of tradeshow to end of tradeshow (t = 24: 30) 0.171 0.101 End of tradeshow to end of simulation (t = 31: 40) 0.321 0.327 Firm participated in TS? Yes No Time period measured Beginning of tradeshow to end of tradeshow (t = 24: 30) 0.171 0.101 End of tradeshow to end of simulation (t = 31: 40) 0.321 0.327 Notes: Results from a simplified three-cluster model where each cluster contains three firms. Technological heterogeneity of individual firms calculated using Euclidean distance formula. View Large Table 3 Tradeshow effects on firm-level technological heterogeneity Firm participated in TS? Yes No Time period measured Beginning of tradeshow to end of tradeshow (t = 24: 30) 0.171 0.101 End of tradeshow to end of simulation (t = 31: 40) 0.321 0.327 Firm participated in TS? Yes No Time period measured Beginning of tradeshow to end of tradeshow (t = 24: 30) 0.171 0.101 End of tradeshow to end of simulation (t = 31: 40) 0.321 0.327 Notes: Results from a simplified three-cluster model where each cluster contains three firms. Technological heterogeneity of individual firms calculated using Euclidean distance formula. View Large Table 4 Tradeshow effects on firm-level technological size Firm participated in TS? Yes No Time period measured Beginning of tradeshow to end of tradeshow (t = 24: 30) 1.17 0.831 End of tradeshow to end of simulation (t = 31: 40) 1.77 1.53 Firm participated in TS? Yes No Time period measured Beginning of tradeshow to end of tradeshow (t = 24: 30) 1.17 0.831 End of tradeshow to end of simulation (t = 31: 40) 1.77 1.53 Notes: Results from a simplified three-cluster model where each cluster contains three firms. View Large Table 4 Tradeshow effects on firm-level technological size Firm participated in TS? Yes No Time period measured Beginning of tradeshow to end of tradeshow (t = 24: 30) 1.17 0.831 End of tradeshow to end of simulation (t = 31: 40) 1.77 1.53 Firm participated in TS? Yes No Time period measured Beginning of tradeshow to end of tradeshow (t = 24: 30) 1.17 0.831 End of tradeshow to end of simulation (t = 31: 40) 1.77 1.53 Notes: Results from a simplified three-cluster model where each cluster contains three firms. View Large Tables 3 and 4 advance three propositions of how pipelines, simulated as a tradeshow affect the dynamics of firms and clusters. First, firms that do not participate in pipelines lose out when a neighboring firm in the same cluster attends a tradeshow. The top rows of both Tables 3 and 4 show that firms that participate in tradeshows grow and gain heterogeneity more rapidly than the firms that do not participate. This result is driven by two mechanisms. First, there are more firms in the tradeshow cluster than the at-home clusters during the tradeshow. Larger clusters imply greater technological diversity, recombinant possibilities and technological growth. Second, the firms attending the tradeshow have not previously interacted, so tradeshows allow for novel recombination between new partners. The second proposition advanced by this simplified model is that pipeline benefits spillover to nonparticipating firms. In the bottom row of Tables 3 and 4, we see that after the tradeshow concludes, the tradeshow-participant firms add slightly less technological heterogeneity than the nonparticipants do (0.321 compared with 0.327). This outcome results from three mechanisms: the outflow of knowledge developed in the at-home cluster during the tradeshow to the firm returning from the tradeshow, a re-balancing of cluster sizes and the spillover of tradeshow-sourced knowledge to the at-home cluster. Because we found in model Version 2A of Table 2 that tradeshow participation is most beneficial in clusters with strong local interaction, we conclude this third spillover mechanism is largely responsible for the strong performance of nonparticipant firms after the tradeshow concludes. The third proposition advanced by our simplified model is that pipelines create lasting first-mover benefits. Comparing the bottom row of Table 3 to the bottom row of Table 4, we see that while the growth in technological heterogeneity essentially equalizes across the firms after the tradeshow concludes, the size growth advantage persists for firms that participated in the tradeshow. In the bottom row of Table 4, firms that participated in the tradeshow grow by an average of 1.77 technologies between the conclusion of the tradeshow and the end of the simulation while nonparticipants in tradeshows grow on average by 1.53 technologies. The more rapid growth of tradeshow-participant firms after the tradeshow concludes is driven by these firms’ early entry into new and valuable technologies. Early entry allows firms that participated in the tradeshow to exploit the monopolistic rents associated with their new technologies before they diffuse to their neighboring firms. Thus, while the firms in a cluster may benefit from pipelines even if they do not participate in pipelines themselves, the extent to which they benefit is influenced by the speed at which they can access and make use of the pipeline knowledge held by their neighbors. 6. Conclusions At the core of the buzz and pipelines model is the claim that nonlocal interaction between economic agents boosts the gains from local interaction in clusters. This is an important claim that speaks to literatures on research topics as diverse as agglomeration and urban growth, regional competitive advantage and regional resilience. To date, considerable empirical evidence has been marshalled in both local and nonlocal contexts to show how formal and informal linkages influence the innovativeness of individual firms. However, this work has cast insufficient light on the question of when and how these firm-level interactions impact the performance of the clusters in which they are located. This is not an easy question to tackle empirically because of the complexity of feedbacks effects that operate at the cluster level. Supportive of the buzz–pipelines argument, we seek to further theoretical and empirical analysis in this research field with a simulation model that provides additional specificity regarding the conditions under which local and nonlocal forms of interaction generate benefits that diffuse throughout clusters of local economic agents. A simulation approach is useful insofar as it demands careful specification of the exact nature of the relationships between model components and explicit identification of the assumptions on which those relationships are based. At the same time, simulation approaches sacrifice complexity for conceptual tractability. Our simulation is predicated on a rather simple model of firm-level agency, which restricts us from analyzing how bilateral interactions create additional costs and benefits associated with buzz and pipelines. In general, the results from our simulation model of recombinant invention suggest that local interaction between firms within a cluster raises the average performance of those firms in terms of the number and variety of technologies available to cluster members. Our simulation also reveals that nonlocal interaction, modeled as a short-run tradeshow connecting firms from different clusters, tends to raise the average performance of the home clusters from which tradeshow participants originate. These results accord with the core arguments of the buzz–pipelines model. However, our simulation generates two other important outcomes that we believe add important qualifiers to the results just outlined. First, we find that greater local interaction is not always beneficial for clusters, for high levels of local interaction may cause co-located firms to develop overlapping technological capabilities which result in reduced knowledge recombination, decreased innovation in the cluster and negative technological lock-in. Second, the participation of firms in pipelines generates opportunity costs within their home clusters that, at least in the scenarios we explored, can outweigh the benefits of such participation. Together, these findings show that the benefits of local and nonlocal interaction cannot be taken for granted. While papers within the buzz and pipelines tradition identify some costs to nonlocal interaction on home clusters (see Henn, 2012), clearly more work on this topic is needed. Continued exploration with our simulation model has generated a series of related issues that are ripe for further theoretical and empirical analysis. The first of these has to do with the nature of pipelines and the extent to which the costs generated by ephemeral forms of interaction such as tradeshows may differ from the costs generated by more formal and enduring partnerships. While the benefits of formal versus informal interaction have received some research attention (Maskell et al., 2006; Allen et al., 2007), analysis of potential variation in the costs of such relationships on home clusters and the strategies that firms use to mitigate these costs is warranted. A second closely related issue has to do with the number or the share of economic agents within clusters that engage in different forms of interaction. Are there critical numbers of local interacting agents that are required to hold a cluster together when other agents increasingly look to distant partnerships? Specifically, how do the aggregate costs and benefits of nonlocal interaction vary with the number and size of firms engaged in pipelines, and in relation to the size of clusters themselves? This question is linked to the evolution of clusters and to the variable role of local and nonlocal interaction at different stages of cluster development (Taube et al., 2018). A third issue is connected to the burgeoning literature on geographies of knowledge sourcing (see Cantwell and Mudambi, 2011) and the factors that influence the variability of returns to pipeline participation. Our simulation model makes clear that interaction with partner firms from clusters with nonoverlapping knowledge bases will enhance returns, but this argument begs questions about cognitive proximity and the capacity of agents across clusters to source and combine unfamiliar knowledge types. The literature on related and unrelated variety speaks directly to these issues regarding absorptive capacity and seems to offer fruitful possibilities for extending research on buzz and pipelines. Acknowledgements The authors acknowledge helpful comments from Pierre-Alex Balland, Koen Frenken and workshop participants at the 2015 Geography of Innovation conference (Toulouse, France), the 2016 Association of American Geographers Annual Meeting (San Francisco), the 2016 Regional Economic Development and Innovation Workshop (UCLA) and seminar participants at Utrecht University. The usual disclaimer applies. Footnotes 1 Expected returns in time t are given by the real (observed) prices in the previous time period, t − 1. 2 Our fixed structure of inter-agent interactions resembles the simulation model of Cowan and Jonard (2003). 3 We assign all firms participating at the tradeshow medium strength local interaction (LIS = 0.5) within the temporary cluster, regardless of the local interaction strength of their home clusters. We explore alternative tradeshow LIS values in robustness checks. 4 We follow the tradition in evolutionary economic modeling (Nelson and Winter, 1982) in assuming invariant demand in order to focus the model on the supply-side dynamics of the knowledge-based competition. 5 Note that we could make these same arguments employing a generic bundle of raw material inputs that are rendered technology-and-output specific through their transformation in the firm in more or less the same way. 6 We ignore the order of technologies, so the b-a becomes a second a-b. References Ahuja G. ( 2000 ) Collaboration networks, structural holes, and innovation: a longitudinal study . Administrative Science Quarterly , 45 : 425 – 455 . Google Scholar Crossref Search ADS Allen J. , James A. , Gamlen P. ( 2007 ) Formal versus informal knowledge networks in R&D: a case study using social network analysis . R&D Management , 37 : 179 – 196 . Google Scholar Crossref Search ADS Arthur B. ( 2009 ) The Nature of Technology: What Is It and How It Evolves . New York : Free Press . Asheim B. , Coenen L. ( 2005 ) Knowledge bases and regional innovation systems: comparing Nordic clusters . Research Policy , 34 : 1173 – 1190 . Google Scholar Crossref Search ADS Asheim B. , Coenen L. , Vang J. ( 2007 ) Face-to-face, buzz, and knowledge bases: sociospatial implications for learning, innovation, and innovation policy . Environment and Planning C: Government and Policy , 25 : 655 – 670 . Google Scholar Crossref Search ADS Audretsch D. , Feldman M. ( 1996 ) R&D spillovers and the geography of innovation and production . The American Economic Review , 86 : 630 – 640 . Balland P.-A. , Rigby D. ( 2017 ) The geography of complex knowledge . Economic Geography , 93 : 1 – 23 . Google Scholar Crossref Search ADS Bathelt H. , Gluckler J . ( 2011 ) The Relational Economy: Geographies of Knowing and Learning . Oxford, UK : Oxford University Press . Bathelt H. , Malmberg A. , Maskell P. ( 2004 ) Clusters and knowledge: local buzz, global pipelines, and the process of knowledge creation . Progress in Human Geography , 28 : 31 – 56 . Google Scholar Crossref Search ADS Boschma R. ( 2005 ) Proximity and innovation: a critical assessment . Regional Studies , 39 : 61 – 74 . Google Scholar Crossref Search ADS Breschi S. , Lissoni F. ( 2009 ) Mobility of skilled workers and co-invention networks: an anatomy of localized knowledge flows . Journal of Economic Geography , 9 : 439 – 468 . Google Scholar Crossref Search ADS Cantwell J. , Mudambi R. ( 2011 ) Physical attraction and the geography of knowledge sourcing in multinational enterprises . Global Strategy Journal , 1 : 206 – 232 . Google Scholar Crossref Search ADS Chandrasekaran A. , Linderman K. , Sting F. J. , Benner M. J. ( 2016 ) Managing R&D project shifts in high-tech organizations: a multi-method study . Production & Operations Management , 25 : 390 – 416 . Google Scholar Crossref Search ADS Chesbrough H. ( 2003 ) The Era of Open Innovation . Boston : Harvard Business School Press . Cowan R. , Jonard N. ( 2003 ) The dynamics of collective invention . Journal of Economic Behavior and Organization , 52 : 512 – 532 . Google Scholar Crossref Search ADS Davis J. , Eisenhardt K. , Bingham C. ( 2007 ) Developing theory through simulation methods . Academy of Management Review , 32 : 480 – 499 . Google Scholar Crossref Search ADS DiMaggio P. , Powell W. ( 1983 ) The iron cage revisited: institutional isomorphism and collective rationality in organizational fields . American Sociological Review , 48 : 147 – 160 . Google Scholar Crossref Search ADS Duranton G. , Puga D. ( 2004 ) Micro-foundations of urban agglomeration economies. In Henderson J. V. , Thisse J. F. (eds) Handbook of Regional and Urban Economics . pp. 2063–2117. North Holland : Elsevier . Dyer J. , Singh H. ( 1998 ) The relational view: cooperative strategy and sources of interorganizational competitive advantage . Academy of Management Review , 23 : 660 – 679 . Google Scholar Crossref Search ADS Fitjar R. , Rodriguez-Pose A. ( 2011 ) When local interaction does not suffice: sources of firm innovation in urban Norway . Environment and Planning A , 43 : 1248 – 1267 . Google Scholar Crossref Search ADS Fitjar R. , Rodriguez-Pose A. ( 2013 ) Firm collaboration and modes of innovation in Norway . Research Policy , 42 : 128–128 . Google Scholar Crossref Search ADS Fitjar R. , Rodriguez-Pose A. ( 2017 ) Nothing is in the air . Growth and Change , 48 : 22 – 39 . Google Scholar Crossref Search ADS Fleming L. , Sorenson O. ( 2001 ) Technology as a complex adaptive system: evidence from patent data . Research Policy , 30 : 1019 – 1039 . Google Scholar Crossref Search ADS Freitas I. , Clausen T. , Fontana R. , Verspagen B. ( 2011 ) Formal and informal external linkages and firms’ innovative strategies: a cross-country comparison . Journal of Evolutionary Economics , 21 : 91 – 119 . Google Scholar Crossref Search ADS Gertler M. ( 2003 ) Tacit knowledge and the economic geography of context, or the undefinable tacitness of being (There) . Journal of Economic Geography , 3 : 75 – 99 . Google Scholar Crossref Search ADS Gilsing V. , Bekkers R. , Freitas I. , Van der Steen M. ( 2011 ) Differences in technology transfer between science-based and development-based industries: transfer mechanisms and barriers . Technovation , 31 : 638 – 647 . Google Scholar Crossref Search ADS Giuliani E. ( 2007 ) The selective nature of knowledge networks in clusters: evidence from the Chilean wine industry . Journal of Economic Geography , 7 : 139 – 168 . Google Scholar Crossref Search ADS Giuliani E. , Bell M. ( 2005 ) The micro-determinants of meso-level learning and innovation: evidence from a Chilean wine cluster . Research Policy , 34 : 47 – 68 . Google Scholar Crossref Search ADS Glaeser E. , Kallal H. , Scheinkman J. , Shleifer A. ( 1992 ) Growth in cities . Journal of Political Economy , 100 : 1126 – 1152 . Google Scholar Crossref Search ADS Grabher G. ( 1993 ) (ed.) The weakness of strong ties: the lock-in of regional development in the Ruhr area. In The Embedded Firm: On the Socioeconomics of Industrial Networks , pp. 255 – 277 . London : Routledge . Grant R. ( 1996 ) Toward a knowledge-based theory of the firm . Strategic Management Journal , 17 : 109 – 122 . Google Scholar Crossref Search ADS Hassink R. ( 2010 ) Locked in decline? On the role of regional lock-ins in old industrial areas. In Boschma R. , Martin R. (eds) Handbook of Evolutionary Economic Geography , pp. 450 – 468 . Cheltenham, UK : Edward Elgar . Heiman B. , Nickerson J. ( 2002 ) Towards reconciling transaction cost economics and the knowledge-based view of the firm: the context of interfirm collaborations . International Journal of the Economics of Business , 9 : 97 – 116 . Google Scholar Crossref Search ADS Henn S. ( 2012 ) Transnational entrepreneurs, global pipelines and shifting production patterns: the example of the Palanpuris in the diamond sector . Geoforum , 43 : 497 – 506 . Google Scholar Crossref Search ADS Jaffe A. , Trajtenberg M. , Henderson R. ( 1993 ) Geographic localization of knowledge spillovers as evidenced by patent citations . Quarterly Journal of Economics , 108 : 577 – 598 . Google Scholar Crossref Search ADS Kauffman S. ( 1993 ) The Origins of Order: Self-Organization and Selection in Evolution . Oxford, UK : Oxford University Press . Kogut B. ( 2000 ) The network as knowledge: generative rules and the emergence of structure . Strategic Management Journal , 21 : 405 – 425 . Google Scholar Crossref Search ADS Kogut B. , Zander O. ( 1992 ) Knowledge of the firm, combinative capabilities, and the replication of technology . Organization Science , 3 : 383 – 397 . Google Scholar Crossref Search ADS Marshall A. ( 1920 ) Principles of Economics . London : Macmillan . Maskell P. , Bathlet H. , Malmberg A. ( 2006 ) Building global knowledge pipelines: the role of temporary clusters . European Planning Studies , 14 : 997 – 1013 . Google Scholar Crossref Search ADS Moodysson J. ( 2008 ) Principles and practices of knowledge creation: on the organization of “Buzz” and “Pipelines” in life science communities . Economic Geography , 84 : 449 – 469 . Google Scholar Crossref Search ADS Morrison A. , Rabellotti R. , Zirulia L. ( 2013 ) When do global pipelines enhance the diffusion of knowledge in clusters? Economic Geography , 89 : 77 – 96 . Google Scholar Crossref Search ADS Nelson R. , Winter S. ( 1982 ) An Evolutionary Theory of Economic Change . Cambridge, MA : Harvard University Press . Owen-Smith J. , Powell W. ( 2004 ) Knowledge networks as channels and conduits: the effect of spillovers in the Boston biotechnology community . Organization Science , 15 : 5 – 21 . Google Scholar Crossref Search ADS Pavitt K. ( 1999 ) Technology, Management and Systems of Innovation. Cheltenham, UK : Edward Elgar . Powell W. , Koput L. , Smith-Doerr L. ( 1996 ) Interorganizational collaboration and the locus of innovation: networks of learning in biology . Administrative Science Quarterly , 41 : 116 – 145 . Google Scholar Crossref Search ADS Rigby D. , Essletzbichler J. ( 2006 ) Technological variety, technological change, and a geography of production techniques . Journal of Economic Geography , 6 : 45 – 70 . Google Scholar Crossref Search ADS Rodriguez-Pose A. ( 2013 ) Do institutions matter for regional development? Regional Studies , 47 : 1034 – 1047 . Google Scholar Crossref Search ADS Saxenian A. ( 1994 ) Regional Advantage: Culture and Competition in Silicon Valley and Route 128 . Cambridge : Harvard University Press . Schrader S. ( 1991 ) Informal technology transfer between firms: cooperation through information trading . Research Policy , 20 : 153 – 170 . Google Scholar Crossref Search ADS Simmie J. , Martin R. ( 2010 ) The economic resilience of regions: towards and evolutionary approach . Cambridge Journal of Regions, Economy and Society , 3 : 27 – 43 . Google Scholar Crossref Search ADS Simon H. ( 1962 ) The architecture of complexity . Proceedings of the American Philosophical Society , 106 : 467 – 482 . Sonn J. , Storper M. ( 2008 ) The increasing importance of geographical proximity in knowledge production: an analysis of US patent citations, 1975-1997 . Environment and Planning A , 40 : 1020 – 1039 . Google Scholar Crossref Search ADS Storper M. , Venables A. ( 2004 ) Buzz: face-to-face contact and the urban economy . Journal of Economic Geography , 4 : 351 – 370 . Google Scholar Crossref Search ADS Storper M. , Kemeny T. , Makarem N. , Osman T. ( 2015 ) The Rise and Fall of Urban Economies: Lessons from San Francisco and Los Angeles . Palo Alto, CA : Stanford University Press . Taube F. , Karna A. , Sonderegger P. ( 2018 ) Economic geography and emerging market clusters: a co-evolutionary study of local and non-local networks in Bangalore . International Business Review (in press), doi: 10.1016/j.ibusrev.2018.03.011. Teece D. , Pisano G. ( 1994 ) The dynamic capabilities of firms: an introduction . Industrial and Corporate Change , 3 : 537 – 556 . Google Scholar Crossref Search ADS Tether B. ( 2002 ) Who co-operates for innovation, and why: an empirical analysis . Research Policy , 31 : 947 – 967 . Google Scholar Crossref Search ADS Tödtling F. , Lehner P. , Tripp M. ( 2006 ) Innovation in knowledge intensive industries: the nature and geography of knowledge links . European Planning Studies , 14 : 1035 – 1058 . Google Scholar Crossref Search ADS Torre A. , Rallet A. ( 2005 ) Proximity and localization . Regional Studies , 39 : 47 – 59 . Google Scholar Crossref Search ADS Trippl M. , Tödtling F. , Lengauer L. ( 2009 ) Knowledge sourcing beyond buzz and pipelines: evidence from the Vienna software sector . Economic Geography , 85 : 443 – 462 . Google Scholar Crossref Search ADS Tripsas M. , Schrader S. , Sobrero M. ( 1995 ) Discouraging opportunistic behavior in collaborative R&D: a new role for government . Research Policy , 24 : 367 – 389 . Google Scholar Crossref Search ADS Uzzi B. ( 1996 ) The sources and consequences of embeddedness for the economic performance of organizations: the network effect . American Sociological Review , 61 : 674 – 698 . Google Scholar Crossref Search ADS Uzzi B. ( 1997 ) Social structure and competition in interfirm networks: the paradox of embeddedness . Administrative Science Quarterly , 42 : 35 – 67 . Google Scholar Crossref Search ADS Weitzman M. ( 1998 ) Recombinant growth . Quarterly Journal of Economics , 113 : 331 – 360 . Google Scholar Crossref Search ADS Wuchty S. , Jones B. , Uzzi B. ( 2007 ) The increasing dominance of teams in production of knowledge . Science , 316 : 1036 – 1039 . Google Scholar Crossref Search ADS PubMed © The Author(s) (2018). Published by Oxford University Press. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Buzz and pipelines: the costs and benefits of local and nonlocal interaction
JO - Journal of Economic Geography
DO - 10.1093/jeg/lby039
DA - 2019-05-01
UR - https://www.deepdyve.com/lp/oxford-university-press/buzz-and-pipelines-the-costs-and-benefits-of-local-and-nonlocal-pNNzVi3gob
SP - 753
VL - 19
IS - 3
DP - DeepDyve
ER -