طراحی شبکه های سازمانی جهت نوآوری: بررسی تجربی تنظیمات شبکه، شکل گیری و اداره

Strategic SME networks have received significant policy attention, yet a review of the current literature reveals limited attention to the factors that contribute to network innovation. This study examines the influence of the number of member firms (network size), the extent to which a network is based on firm incentives (bottom-up formation), and the extent of development of the governance structure (size of administrative function) on a network's innovative performance. Latent growth modeling with longitudinal data from 53 networks reveals that larger networks and bottom-up formed networks achieve greater innovative performance, and that the administrative function partially mediates these effects.

Moving away from individualized business with local and regional customers and competitors, small- and medium-sized enterprises (SMEs) are increasingly required to be innovative to confront globalization and ever-increasing competition. Such demands have historically been met with the establishment and maintenance of strong interpersonal relationships among entrepreneurs, dyadic business relationships with a few selected partners, and the development of cooperative capabilities (Collins and Hitt, 2006 and Tyler, 2001). However, in addition to such relationships and activities, SMEs increasingly involve informal networks of interorganizational relationships among several SMEs that work together as independent firms but with shared objectives. As a more formal cooperative form, strategic SME networks have been defined as “intentionally formed groups of small- and medium-sized profit-oriented companies in which the firms: (1) are geographically proximate, (2) operate within the same industry, potentially sharing inputs and outputs, and (3) undertake direct interactions with each other for specific business outcomes” (Human and Provan, 1997, p. 372). This view on such networks is consistent with what has been reported in the existing literature (see, e.g., Fukugawa, 2006, Huggins, 2001, Lipparini and Sobrero, 1994, Sherer, 2003, Vanhaverbeke, 2001, Wincent, 2005a and Wincent, 2005b), where specific business outcomes are related to various forms of process, product, and technological innovation. These networks can be rather large in size (there are networks that comprise up to 100 member firms), wherein the member firms join together as independent firms under a common trademark and aim to develop innovation by joint forces. For example, the Swedish network YWOOD includes approximately 50 manufacturing firms that work together to develop new products related to the wood industry and experiment with complementary product concepts. The idea is that all customers will find what they are looking for within the network and that the network will offer competitive products to consumers on the international market. The idea is also to have a competitive and innovative product base to offer to large global retailers. Within this network, the participating firms have developed both manufacturing equipment to process wood in new ways and various new products that the member firms have successfully exported. Obviously, there are no guarantees for success and there are many networks that have failed to produce innovation for their participants (Rosenfeld, 1996 and Sherer, 2003). Despite the current significant policy attention and widespread implementation of SMEs as a tool to strengthen innovation and competitiveness (Hanna and Walsh, 2002 and Rosenfeld, 1996), the academic study of how to design strategic SME networks is apparently still in its infancy. Even though a literature stream on strategic SME networks has started to develop and to research a variety of questions, scholars face a situation in which practitioners must design strategic SME networks without much specific support from the research community to guide their decisions (see, e.g., Ahlström-Söderling, 2003 and Chaston, 1995). Thus, in an economy where innovation is increasingly important, we organized the research for this paper around the following two questions: what factors are important to consider when developing strategic SME networks that support innovation? What broad areas of influential factors have been mentioned in the research so far and how important would a selection of such factors be for innovative performance in strategic SME networks? We believe that addressing these questions makes important contributions to the literature. To present the positioning and contribution of this study, we describe the field of research (i.e., as is published in academic journals) on strategic SME networks in Table 1. Various studies have addressed questions related to strategic SME networks without providing direct guidance for network design, but most efforts in Table 1 report evidence from small-scale research (including one or two networks) of an exploratory character (83 percent did not test ex ante-developed hypotheses), which leaves anecdotal evidence for the cases they are designed to examine. The studies are also highly fragmented and research a host of problems that does not directly focus on the key outcome of strategic SME networks, namely, innovation.Although the studies make important contributions and descriptions related to phenomena such as knowledge transfer from science to SMEs in networks (Major and Cordey-Hayes, 2000), the creation of strategic SME networks (Ahlström-Söderling, 2003), the role of broker competences in facilitating the progression of such networks (Chaston, 1995), and even ways to reduce opportunism in strategic SME networks (Hammami et al., 2003), there are no efforts that attempt specifically to systematically test potential key factors that influence innovative performance in strategic SME networks, the relative impact of possible factors, or the existence of possible relationships among such factors. Supported by the notion of the importance of integrative research and acknowledging the use of prior work, a study that includes indications or suggestions of possible influences on innovation from earlier studies and that integrates efforts and key ideas should provide a foundation for progress in a research field (Cornelius et al., 2006). To this background, we developed and tested a model of influences on innovative performance in strategic SME networks on a large longitudinal sample of such networks in Sweden. We expanded prior work by focusing on how a selection of hypothesized factors jointly influences performance. We were also able to theorize and examine the influence of those factors across time. Expanding on previous research, we selected potentially important variables related to network design in three areas: network configuration (few or many member firms), network formation (bottom-up or top-down initiatives), and network governance (small or large administrative function) to include in a model that centrally placed network governance (i.e., a mediation model). All these variables seem contextually important for several reasons, but they have not been explicitly related to innovative performance. We also examined the potential role of partial mediation; such that the effect of network configuration and formation strengthens innovative performance in strategic SME networks both directly and indirectly through its respective influence on network governance. To our knowledge, this study offers the first systematic and longitudinal large-sample test of the effect of a selection of network design variables on innovative performance in strategic SME networks. This article is structured as follows. First, we give a short description of some distinct characteristics of strategic SME networks. Next, we integrate prior research on strategic SME networks with our development of hypotheses that examine the role of different network designs for network innovation, which is followed by a section on methodological issues related to how we carried out the empirical testing. Finally, we draw from this experience to discuss implications for theory and practice.

Table 2 presents descriptive statistics for all variables used in the analyses in the form of average values and standard deviations at each measurement occasion. We measured network size and the bottom-up formation process at time 1, and we measured the size of the administrative function and network innovative performance at times 1, 2, and 3. A simple visual inspection of the means shows that both network innovative performance and the size of the administrative function increased over time.Before estimating the full model, we separately examined the latent growth models for innovative performance and size of the administrative function (see Table 3). Following general guidelines (i.e., model fit indicators) we found that both models (innovative performance and size of the administrative function) represented the data well and were usable in further model testing. Table 3 reports on loadings, which have been set to 0, 1, and 2 for the three yearly measurement points examined because of reasons mentioned in the beginning of Section 5. Table 3 also reports on the average and variance of initial levels and rates of change in both size of administrative function and innovative performance. All values are significant, which illustrates that we can estimate initial levels of both the administrative function and innovative performance, that the rate of change is significant, and that we therefore see general tendencies in the administrative function and innovative performance growing or declining over time.Specifically, Table 3 reveals that the average size of the administrative function was initially 3.91 (mean initial level) and the average rate of change was .44 (mean rate of change). This implies a significant developmental increase in the administrative function (size of the administrative function = 3.91 + 0.44t, at t = 1, 2, and 3). The average level of innovative performance was initially 1.54 (mean initial level), with a predicted average rate of change of 0.18 (mean rate of change). Thus, we also find a significant development increase in innovative performance (innovative performance = 1.54 + 0.18t, at t = 1, 2, and 3). Model fit values are reported at the end of the table in form of chi-square (χ2) and goodness-of-fit, where both are within acceptable levels, indicating the validity of modeling innovative performance and size of the administrative function as a linear growth model for the three examined years. To further test the validity of a linear representation, we also tested potential quadratic forms of both the size of administrative function and innovative performance. We did so to be able to exclude the possibility that the growth in both administrative function and innovative performance are better described as nonlinear effects. Because of the inclusion of quadratic terms we could confirm that the linear estimation of the administrative function and innovative performance prevailed over the respective quadratic forms. Therefore, we continued estimating linear effects in the growth patterns of the administrative function and innovative performance. To determine the mediating role of the network administrative function, we performed a nested model test (see Table 4). Our model, and its pertinent hypotheses, suggests that network size and bottom-up formation influence innovative performance both directly and indirectly—indirectly by influencing the size of the administrative function. As such, we suggest that the size of the administrative function partially mediates the effects of network size and bottom-up formation on innovative performance. In comparison to our proposed partially mediated model, we also estimated a fully mediated model and a model that excluded mediating effects. The fully mediated model excluded paths from network size and bottom-up formation process on innovative performance, and thereby assumes that network size and bottom-up formation influence innovative performance only by their respective influence on the size of the administrative function.The model representing no mediation excluded the influence of the intercept and slope of the administrative function on network innovative performance, thereby suggesting that the effects from network size and bottom-up formation are direct and that there is no effect through the size of administrative function. Table 4 reports on several of the more commonly used goodness-of-fit statistics for the three separate tables (including chi-square statistics, RMSEA, GFI, CFI, and NFI), as well as on chi-square differences for the three competing models. The goodness-of-fit statistics reveal that our hypothesized model (the partially mediated model) represented the data well, with a nonsignificant chi-square and with relative- and absolute-fit indicators well within recommended limits (i.e., RMSEA < .08; GFI, CFI, and NFI all greater than .90). Both the fully mediated model and the nonmediated model showed a significant chi-square and a goodness-of-fit index below recommended levels. To further evaluate the models, we also compared the chi-square of the different models to assess which model best represented the data. When comparing the partially mediated model with the fully mediated model (model 1 versus model 2 in Table 4) we found that including direct effects from network size and bottom-up formation on innovative performance significantly improves the fit of the model with the underlying data, which suggests that the partially mediating model better represents the data than does the fully mediated model. Similarly, when comparing the partially mediated model with the nonmediated model (model 1 versus model 3 in Table 4), we found that the inclusion of the mediation effect shows a significantly better fit than does the option of not including mediation in the model. As such, we have tested and confirmed that the partially mediated model prevails over the two competing models; thus, we proceed by estimating and analyzing the specific relationships in the partially mediated model. Table 5 reports the results for the structural equation estimation of the hypothesized model (as specified in Fig. 1). Judging by the significances and the size and direction of the coefficients of the intercepts of the dependent variables, we found support for our three hypotheses on predictors of network innovative performance. Hypothesis 1 predicted that a larger number of network firms would be related to greater innovative performance in strategic SME networks. As illustrated in Table 5, network size has a positive and significant coefficient in relation to the intercept of the network innovative performance (β = .47, p < .001), which is consistent with Hypothesis 1.Hypothesis 2 argued for greater innovative performance in strategic SME networks formed by a bottom-up formation process compared to those formed by a top-down process. The coefficient of bottom-up formation is positively and significantly related to the intercept of network innovative performance (β = .32, p < .001). Thus, Hypothesis 2 is supported. Hypothesis 3 posited that a larger administrative function would be related to greater innovative performance in strategic SME networks. Board size has a positive and significant coefficient in relation to the intercept of the network innovative performance (β = .21, p < .05), which supports Hypothesis 3. Similarly, we found support for our two hypotheses on predictors of the size of the strategic SME network administrative function. Hypothesis 4 predicted that a larger number of network firms would be related to a larger administrative function of strategic SME networks. The coefficient of network size is positively and significantly related to the intercept of board size (β = .38, p < .001), which is consistent with Hypothesis 4. Hypothesis 5 argued that a bottom-up network formation is associated with a larger strategic SME network administrative function. A positive and significant coefficient of bottom-up formation in relation to the intercept of the size of the strategic SME network administrative function provides support for Hypothesis 5 (β = .23, p < .05). As illustrated in Table 5, the joint effects of network size, bottom-up formation, and the size of the administrative function explained 58 percent of the variance in latent baseline network innovative performance, while network size and bottom-up formation accounted for 24 percent of the variance in latent baseline strategic SME network administrative function. Besides our formulated hypotheses we also posited the possibility that some of the effects of the independent variables would not only be related to the intercept of innovative performance and size of the administrative function, but also be related to the rate of growth in these dependent variables over time. As such, without any formal hypotheses we suggested that the design variables examined could possibly be related to the growth in the size of the administrative function and the level of innovative performance over time. In terms of latent growth (reported as slope in Table 5), coefficients for network size did not significantly predict variance in either network innovative performance or size of the strategic SME network's administrative function. However, we found positive and significant coefficients for the bottom-up formation process, indicating that it predicts variance in latent growth in both network innovative performance (β = .16, p < .05) and in the size of the strategic SME network administrative function (β = .18, p < .05). Furthermore, we found latent growth in the size of the strategic SME network administrative function to be significantly and positively related to latent growth in network innovative performance (β = .18, p < .05). Overall, the estimated model reveals 10 percent explained variance in latent growth innovative performance and 3 percent explained variance in latent growth administrative function size.