How Internal and External Sources of Knowledge Contribute to Firms' Innovation Performance Anja Cotic Svetina Igor Prodan This paper investigates the extent to which different knowledge sources contribute to firms' innovation performance. The empirical analysis estimates the relationships in the structural model of the influence of knowledge sources on innovative performance using data collected through personal interviews at 303 firms. The results reveal that internal sources have the most important influence on firms' innovative per- formance and confirm that, in their innovation process, firms mostly rely on knowledge developed through in-house r&d efforts, continu- ous improvement, and internal education and training programs. The data show that in-house learning is not sufficient for generating in- novation and that firms need to supplement internal knowledge with knowledge acquired outside the firm. They mainly need to secure links with firms and institutions in the global environment if they want to secure the inflow of new ideas and approaches that will eventually lead to innovations. Key Words: knowledge, innovation, structural equation modeling jEL Classification: 030, 031 Introduction An interactive view of innovation has been developed within the frame- work of a learning economy, in which innovation is seen as a techni- cal and social process based on the complex interaction between firms and their environment (Asheim and Isaksen 1997). Most authors agree that the use of internal and external knowledge sources contributes pos- itively to firms' innovation performance, but the relationship has been empirically tested only to a limited extent (Capello 1999; Caloghirou, Kastelli, and Tsakanikas 2004; Capello and Faggian 2005). This paper in- vestigates the extent to which various knowledge sources contribute to firms' innovation performance. More specifically, it identifies on the one DrAnja Cotic Svetina is an Assistant at the Faculty of Economics, University of Ljubljana, Slovenia. Dr Igor Prodan is an Assistant Professor at the Faculty of Economics, University of Ljubljana, Slovenia. Managing Global Transitions 6 (3): 277-299 hand the level of importance of internal knowledge sources embodied mainly in in-house r&d efforts. On the other hand, it looks at exter- nal sources of knowledge and identifies how the use of local, national, and international knowledge sources determines firms' innovation per- formance. This paper extends the work of other scholars as to what are the sources of innovation, by considering how knowledge sources at dif- ferent spatial levels influence the innovation performance of firms. While most authors analyzed the role of external knowledge sources in general (Caloghirou, Kastelli, and Tsakanikas 2004; Willoughby and Galvin 2005; Tsai and Wang 2007; Love and Mansury 2007), we divide them according to the geographical proximity to the observed firm. As such, this is one of the few empirical papers that assesses simultaneously to what extent internal, local, national and international knowledge sources contribute to firms' innovation. The empirical analysis is based on a survey that was carried out in seven European countries: the Czech Republic, Germany, Italy, Poland, Romania, Slovenia, and the United Kingdom. The relationships in the structural model of the influence of sources of knowledge on innovation performance are estimated using data collected through personal inter- views at 303 firms. The results reveal that internal sources have the most important influence on firms' innovative performance and confirm that, in their innovation process, firms mostly rely on knowledge developed through in-house r&d efforts, continuous improvements and internal education, and training programs. The data show that in-house learn- ing alone is not sufficient for generating innovation and the firms need to supplement internal knowledge with knowledge acquired outside the firm. They mainly need to secure links with firms and institutions in the extra-local environment in order to secure the inflow of new ideas and approaches that will eventually lead to innovations. The paper is structured in five sections. The next section presents the theoretical framework on which the empirical analysis is based. The main focus is on the literature describing the importance of internal and external knowledge sources and how they contribute to firms' innova- tive performance. Then four hypotheses are developed, which are later empirically tested. The third section describes the methodology used, in- cluding the sampling and data collection process, data analysis, and op- erationalization and measure validation. The fourth section is dedicated to presenting the empirical findings together with a graphic presenta- tion of the structural model. The results are summarized and the main findings discussed in the last section. Acquiring Knowledge for Innovation: Theory and Hypotheses Development Until the 1980s, understanding of the innovation process was strongly influenced by the linear model of innovation, which suggested that de- velopment of innovations follows a straight research-to-market trajec- tory. In this model a central role was given to r&d activity, and firms' innovative performance was mainly seen as a consequence of r&d in- vestment. This research-based and technocratic view of the innovation process could not explain the success of several sme firms that had lim- ited resources for in-house r&d but were able to base their competitive- ness on constant innovation. This phenomenon of innovative smes has become especially apparent in several sme clusters that have emerged all over Europe and the rest of the world. Since then, several scholars and practitioners have tried to reveal the dynamics behind small and medium-sized firms' innovativeness. More than a decade ago it became obvious that innovations rarely occur as creative acts of individual ge- niuses, but more often as a result of interactive processes. Individuals can not learn new things in a cognitive vacuum and learning always takes place in relation to some kind of social context (Johnson 1992; Lundvall 1992). From the perspective of innovation, new knowledge is not only developed in r&d departments but also in connection with ordinary production activities of firms and other actors through the interactive learning process (Eriksson 2005). Firms cooperate with their suppliers, customers, knowledge institutions (universities, laboratories, etc.), and even with their competitors when developing new products and services or improving production processes. The interactive model of innovation explains the process of innovation as a network of knowledge-flows both within the organization, and in the relationship between the organiza- tion and the environment (Santos 2000). internal and external sources of knowledge This section aims to show how complex the process of knowledge acqui- sition is, and to present the idea that firms need to acquire new knowl- edge from numerous internal and external sources in order to constantly generate innovations and maintain their competitive edge. According to the general trend towards more composite knowledge, where new products and processes typically combine many technolo- gies from several scientific disciplines, it is important to understand that firms today can hardly learn and innovate in isolation (Pavitt 1998; John- son, Loren, and Lundvall 2002). While in large firms information and knowledge are still mainly transferred through functional interaction among r&d, production, marketing, and organization departments and functional teams (Capello 1999), small and medium-sized firms increas- ingly need to rely on external knowledge sources. Accordingly, knowl- edge sources can be firstly divided into internal and external sources, whereas external sources can be further divided into local, national, and international sources, depending on where the source of knowledge is located (Belussi, McDonald, and Borras 2002). Internally, firms acquire knowledge through in-house research and development activities and by learning from continuous improvements in processes. Employee skills represent another important source of new knowledge, and firms often organize internal education and training programs in order to further build and improve the internal knowledge base. If firms do not have ap- propriate knowledge inside the firm, they can acquire it externally by cooperating with customers and suppliers, as well as other firms, or by forming partnerships with public, semi-public, and private institutions. In terms of geographic location, these external actors can be located in close geographic proximity (locally), somewhere in the country (nation- ally), or elsewhere (internationally). Among external sources of knowledge, inter-firm collaboration has probably received the most widespread research attention. It is widely recognized that the innovative process often involves interaction between the manufacturer and users of products. Usually such interaction be- tween producers and end users involves not only an exchange of tech- nical knowledge but also important information about market require- ments and trends. Another important source of knowledge comes from the other side of the supply chain. Suppliers of equipment and mate- rial (Geenhuizen 1997) can bring important insight into the organiza- tion of production, logistics and other functions. But inter-firm coop- eration extends far beyond the relationships that develop between sup- ply chain partners. Studies of successful firms reveal that some sort of collaborative arrangements develop between business partners as well as between competitors. For example, a study of the Cambridge region re- vealed that 76% of firms possess close links with other firms (Keeble et al. 1998). When analyzing the nature of inter-firm cooperation they identi- fied everything from joint ventures, subcontracting, and research collab- orations to the sharing of equipment and information about customers. Accordingly, we perceive both vertical as well as horizontal inter-firm re- lationships as sources of important external sources of knowledge and interactive learning (Camagni 1993; Yeung 2005; Steiner and Hartmann 2006). Knowledge exchange not only appears between firms but can often be found between firms and institutions. Universities, research institutes, science parks, incubators, and other knowledge institutions are actively involved in a set of relationships occurring in the business environment (Gunasekara 2006) and are particularly seen as lead players in the in- novative activity of firms providing scientific research inputs for inno- vating firms (Keeble and Wilkinson 2000). According to Gambarotto and Solari (2004), in addition to channeling information and knowl- edge, support organizations can also help translate academic codified knowledge into practical and accessible know-how. In line with the mod- ern understanding of innovation, the research process is oriented toward problem-solving and as such requires two-way research interaction be- tween knowledge organizations and industry actors combined with sev- eral other institutions. Inter-firm collaboration, as well as partnerships with institutions, were long believed to be mainly limited to the local level and were studied within the context of clusters. However, with globalization and advances in information and communication technology, the geographic scope of this interaction is widening and often spreads across national borders. If firms want to succeed in the innovation race, they need to have ac- cess to the most advanced technical and organizational knowledge in their fields, which means they have to search for appropriate knowledge with no regard to its location. The use of geographically close sources has several benefits that stem from constant face-to-face interactions, knowledge spillovers, and the transmission of tacit knowledge (Cam- agni 1991; Keeble 2000; Capello and Faggian 2005). However, this does not imply that the mere use of local knowledge sources is sufficient in terms of knowledge creation and innovation. Research shows that lim- iting knowledge acquisition to the local level can lead to a lock-in effect (Grabher 1993; Keeble and Wilkinson 1999; 2000). In order to maintain a constant inflow of new knowledge, firms need to nurture links inside, as well as outside, the cluster. toward the research hypotheses The following paragraphs present the main theoretical arguments for the role ofknowledge sources in firms' innovation performance, and develop four hypotheses. In-house r&d efforts have been consistently proven to contribute to firms' innovative potential. A systematic review of studies investigating the use of knowledge in smes has shown that internal managerial and entrepreneurial teams, as well as other employees, play a crucial role in knowledge creation and, consequently, innovation (Thorpe et al. 2005). Additionally, a firm's in-house expertise for r&d has a considerable pos- itive effect on the absorptive capacity of firms. Cohen and Levinthal (1990) define absorptive capacity as 'the ability of a firm to recognize the value of new, external information, assimilate it, and apply it to com- mercial ends.' This means that continuous improvements in its internal knowledge base are also important for increasing a firm's capability to assimilate and transform external knowledge and information into new products, services, and processes. As Lundvall and Nielsen (1999) argue, a strong internal knowledge base is the key to successful innovation. In line with their arguments, we posit that the greater the use of internal knowledge sources, the more innovations the firm will be able to create, as well as exploit knowledge from external sources and transform it into innovation. Therefore, we have formulated the following hypothesis: H1 The extent of usage of internal knowledge positively influences firms' innovative performance. What collective and interactive learning literature argues in general is that a firm's learning capacity does not depend solely on individual skills and the organization of the firm (internal to the firm), but it is also context dependent on the institutional set-up of its business environ- ment (Lorenzen 1998; Tomassini and Sarcina 2005). In recent decades companies have been facing an increase in uncertainty and risk (Geen- huizen 1997). Firms in many industries are facing a turbulent environ- ment with changes taking place in market, technology, and industrial organization. Responding to various uncertainties, companies have in- creasingly externalized their sources of knowledge. In order to increase or deploy their own knowledge effectively, firms often need to supple- ment their knowledge with that of other firms and organizations, which often happens in some form of collaborative arrangements. The growing importance of inter-organizational collaboration can be explained by the nature of contemporary knowledge (Dunning 2000): the development of new knowledge can be highly expensive; the outcome of much in- vestment in augmenting knowledge (by r&d) is highly uncertain; many kinds of knowledge become obsolete quite quickly; and complex prob- lems require multi-disciplinary team solutions. In addition, competitive pressures are forcing firms to introduce new products and services to the market at an increasing pace, and for many (especially small and medium-sized) firms it is impossible to rely only on internal resources for necessary knowledge production. Consequently, firms and other or- ganizations are increasingly engaging in inter-organizational coopera- tion projects (Eriksson 2005). There are several institutional environments in which firms acquire knowledge and learn. Lundvall (1992) emphasizes the national level as an institutional framework for learning and innovation, because of its homogeneity with respect to culture, technical and educational institu- tions, and historically-built relations between actors and firms. Other researchers focus on the regional and local levels as the most impor- tant environments for knowledge acquisition and innovation. Recently an increasing number of scholars have proven that firms often search for knowledge internationally (Malmberg and Power 2005). What these the- ories have in common is the fact that external knowledge sources provide an important complement to in-house learning and innovation efforts, and thus contribute to improved innovative performance (Caloghirou, Kastelli, and Tsakanikas 2004). The importance of inter-organizational relationships has been mainly developed and studied in the context of localized clusters, where a num- ber of firms, knowledge and research institutions, and other actors are located in close geographic proximity. The literature on localized and collective learning argues that the local level is the most appropriate en- vironment for knowledge exchange and interactive learning due to cul- tural, social, and organizational proximity (Lundvall 1992; Belussi and Pilotti 2000; Steiner 2006), which has led to formulation of the following hypothesis: H2 The more a firm uses local sources of knowledge, the more it develops knowledge sharing that positively contributes to innovative perfor- mance. The main problem of the localized learning literature is that it has sometimes been read in a way that places local knowledge acquisition as a superior form that might sometimes take the place of knowledge acqui- sition and learning at the national and international levels; some authors even believe that it can replace internal r & d efforts (Capello and Faggian 2005). This stream of literature describes clusters and other local net- works as being somehow self-sufficient in knowledge terms. This has led to a rather heated debate in recent years (Malmberg and Maskell 2006), proving that interactions with distant partners are at least as important as those with local actors. Several authors have empirically proven that learning might be best understood as a combination of close and distant interactions (Malmberg and Maskell, 2006; Wolfe and Gertler 2006; Brit- ton 2003; Cumbers, MacKinnon, and Chapman 2003; Henry and Pinch 2001; Tödling and Kaufmann 1999). Recently many authors have stressed the importance of linkages with external firms, institutions, or even networks, which provide access to external knowledge and technology, and prevent the lock-in effect. The most recent contribution to this discussion comes from Malmberg and Maskell (2006), who submit that neither the argument for localized in- teractive learning nor the existence of localized capabilities in any way presupposes that most knowledge exchange and learning interaction should be local. They believe that extra-local knowledge flows can be expected to connect to the local knowledge flows so that the two become mutually reinforcing. This happens when the external sources 'pump' information and news about markets and technologies into the local en- vironment and consequently intensify the local interaction and benefit the local actors. The main idea of this literature is that intense localization within a cer- tain local environment does not mean isolation from the extra-local en- vironment. Firms form networks and partnerships with firms and other organizations at the local, national and international levels in order to enhance their knowledge base and innovation potential. Business and social contacts can be more frequent, intensive, and easier to maintain if they are facilitated by proximity (all types); however, firms must nur- ture their relationships with firms and organizations outside the local area and even try to engage in global networks (Malmberg and Maskell 2006; Bathelt, Malmberg, and Maskell 2002). This will provide them with access to information on rapidly changing technologies and market op- portunities and provide a constant influx of new knowledge needed in the innovation process. The role of the national knowledge sources was extensively discussed in the literature dealing with national systems of in- novation (Lundvall, 1988; 1992; Lundvall et al. 2002), while the role of in- ternational sources has mainly been studied in the context of r&d part- nerships (Knudsen 2006). Based on this literature, both national and in- ternational sources of knowledge are expected to positively contribute to firm's innovation performance. As Simmie (2006) suggests, innovation must be understood in terms of trading nodes in an international system that encompasses local, national and international knowledge spillovers and multilayered economic linkages that extend over several different spatial scales. The above discussion underpins the last two hypotheses: H3 The more a firm uses national sources of knowledge, the more it de- velops knowledge sharing that positively contributes to innovative performance. H4 The more a firm uses international sources of knowledge, the more it develops knowledge sharing that positively contributes to innovative performance. Methodology The methodology is discussed in terms of the sampling and data collec- tion process, data analyses, operationalization, and measure validation. sampling and data collection process Data for testing the structural equation model for explaining the in- fluence of knowledge sources on innovation activity were collected within the research project weid (West-East id: Industrial Districts' Re- Location Processes; Identifying Policies in the Perspective of the Euro- pean Union Enlargement) conducted under the 5th eu Framework Pro- gram. Eleven European research partners were included in the project: Fondazione Istituto Guglielmo Tagliacarne (Italy), Eurochambres Aisbl (Belgium), Istituto per lo Sviluppo della Formazione dei Lavoratori (Italy), Libera Universita Internazionale degli Studi Sociali 'Guido Carli' (Italy), Manchester Metropolitan University (United Kingdom), Om- nimotio s. r. o. (Czech Republic), Landesinstitut Sozialforschungsstelle Dortmund (Germany), University of Ljubljana (Slovenia), University of Reading (United Kingdom), University of Roskilde (Denmark), and University of Aurel Vlaicu in Arad (Romania). Based on the literature review, interviews with managers, and work with focus groups within the weid research group, a questionnaire for in-depth interviews was developed. The questionnaire was initially prepared in English and then first translated into the local languages (Czech, German, Italian, Polish, Romanian, and Slovenian), and after that back-translated into English (Brislin 1970; Brislin 1980; Hambleton 1993). The translation followed the 'etic approach' - an approach where there is little or no attempt to decenter or adapt the measure to another cultural context (Craig and Douglas 2005). In-depth interviews with top executives from manufac- turing firms from the Czech Republic, Germany, Italy, Poland, Roma- nia, Slovenia, and the United Kingdom were conducted on the basis of the structured questionnaire developed. For the analyses, 303 usable re- sponses were obtained. The composition of the sample was comparable to the population. data analyses Reliability was assessed using Cronbach's (1951) alpha. Construct and discriminant validity, as well as convergent validity, were assessed us- ing exploratory and confirmatory factor analysis (Floyd and Widaman 1995). Exploratory factor analysis and reliability analysis was conducted in spps. The eqs Multivariate Software version 6.1 (Bentler and Wu 2006) was utilized for confirmatory factor analysis and testing of the proposed structural model. Since a small amount of non-normality was found in the data, the Elliptical Reweighted Least Square (erls) estima- tion method was used (Sharma, Durvasula, and Dillon 1989). As recom- mended by Shook, et al. (2004), the fit of the model was assessed with multiple indices: nfi (the normed-fit-index), nnfi (the non-normed-fit index), cfi (the comparative fit index), gfi (the goodness-of-fit index), srmr (the standardized root mean square residual), and rmsea (the root mean square error of approximation). Values of nfi, nnfi, cfi, and gfi greater than 0.90 indicate a good model fit (Hair et al. 1998; Byrne 2006). Hu and Bentler (1999) suggest that values of srmr smaller than 0.08 indicate an acceptable fit. Values of rmsea less than 0.05 indicate good fit, and values as high as 0.08 represent reasonable errors of approx- imation in the population (Browne and Cudeck 1992). The chi-square is reported, but is not given major consideration because it is highly sen- sitive to sample size and the number of items in the model (Bentler and Bonett 1980). operationalization and measure validation In this study, independent and dependent variables were measured through scales previously tested and developed by the weid research group. Internal Sources of Knowledge Internal sources of knowledge were measured with six items. Respon- dents were asked to indicate (on a 5-point Likert-type scale ranging from 'not important at all' to 'very important') how important the following internal sources of knowledge are for their company: knowledge gained through in-house research and development (1NT01), knowledge gained from continuous improvement of production processes (1NT02), knowl- edge developed through their company's internal education and train- ing programs (1NT03), organizational skills learned from continuous im- provement of their production processes (1NT04), organizational skills of the professional managers within their local company (1NT05), and or- ganizational skills gained from their company's internal education and training programs. The factor analysis indicated that all factor loadings were above 0.4 and significant. Cronbach's alpha of 0.80 indicates strong internal consistency of six items operationalized to measure this con- struct. Local, National, and International Sources of Knowledge Local, national, and international sources of knowledge were each mea- sured with 10 items. Respondents were asked to indicate (on a 5-point Likert-type scale ranging from 'not important at all' to 'very impor- tant') how important the following local, national, and international sources of knowledge are for their company: knowledge derived from interactions with clients and/or suppliers (local clients and/or customers - L0C01; national clients and/or customers - NAT01; and international clients and/or customers - 1NAT01), knowledge derived from cooper- ation with other companies (L0C02, NAT02, and 1NAT02), knowledge gained from interactions with public institutions such as universities, public research centers, local government, and so on (L0C03, NAT03, and INAT03), knowledge gained from interactions with semi-public in- stitutions such as chambers of commerce, industry associations, trade unions, and so on (loc04, NAT04, and inat04), knowledge provided by consultants and private research centers (loc05, NAT05, and INAT05), organizational skills gained from interactions with clients and/or sup- pliers (loc06, NAT06, and INAT06), organizational skills gained from cooperation with other companies (loc07, NAT07, and INAT07), organi- zational skills learned from interactions with public institutions such as universities, public research centers, local government, and so on (loc08, NAT08, and INAT08), organizational skills learned from inter- actions with semi-public institutions such as chambers of commerce, industry associations, trade unions, and so on (loc09, NAT09, and INAT09), and organizational skills learned from consultants and pri- vate research centers (loc10, nat10, and inat10). The factor analysis indicated that all factor loadings were above 0.4 and significant for all three constructs. To test for convergent validity of the constructs and to compare the one-factor structure with the three-factor structure (where factors are correlated), the confirmatory factor analysis was conducted. The results showed that one-factor structure is not appropriate because of the overall poor model fit (chi-square = 1420.029, 368 df, probability 0.000; nfi = 0.80; nnfi = 0.81; cfi = 0.84; gfi = 0.69; srmr = 0.12; and rmsea = 0.10). The confirmatory factor analysis showed that the three-factor structure fits the data reasonably well, with the following fit indices: chi-square = 681.457, 365 df, probability 0.000; nfi = 0.90; nnfi = 0.94; cfi = 0.95; gfi = 0.82; srmr = 0.08; and rmsea = 0.05. Cronbach's alphas of 0.85 (local sources of knowledge), of 0.86 (national sources of knowledge), and of 0.86 (international sources of knowledge) indicate strong internal consistency of items operationalized to measure these constructs. Innovation Performance Innovation performance was measured with five items. Respondents were asked to indicate whether their company had registered patents abroad in the last three years (ip01), and to indicate whether their com- pany had introduced or adopted any major changes to their products (ip02), processes (ip03), organization of production (ip04), and organi- zation of sales and distribution (ip05). The factor analysis indicated that all factor loadings were above 0.4 and significant. Cronbach's alpha of 0.75 indicates strong internal consistency of five items operationalized to measure this construct. Control Variables Control variables were also included and operationalized as follows: (1) firm's size was operationalized as the number of employees, and (2) the region was operationalized as a dichotomous variable, where '0' repre- sented western European countries (Italy, Germany, and United King- dom) and '1' represented eastern European countries (Czech Republic, Poland, Romania, and Slovenia). Findings The structural relationships in the model of the influence of sources of knowledge on the innovation performance were estimated using the El- liptical reweighted least square (erls) method in eqs 6.1 (Bentler and Wu 2006). eqs reported that parameter estimates appeared in order, and that no special problems were encountered during the optimization. The resulting model goodness-of-fit indices indicated a moderately good model fit (chi-square = 1600.305, 812 df, probability 0.000; nfi = 0.86; nnfi = 0.92; cfi = 0.93; gfi = 0.76; srmr = 0.08; and rmsea = 0.06). The variance explained for the innovation performance was 20%. The model, which includes hypothesized relationships and results of the model test, is depicted in figure 1. An examination of our hypotheses is presented in the following section. hypotheses testing Hypothesis H1 proposed that the extent of the usage of internal sources of knowledge is positively related to the innovation performance. The results presented in figure 1 show that the internal sources of knowledge have a significant, positive, and high path coefficient of 0.31. The result thus provides strong support for hypothesis H1. Hypothesis H2 proposed a positive relationship between local sources of knowledge and firms' innovation performance. Hypothesis H2 was not supported by the findings (significant standardized path coefficient of -0.26), because the result was the opposite of what was predicted, in- dicating that local sources of knowledge are negatively related to innova- tion performance. Hypothesis H3 assessed the relationship between national sources of knowledge and firms' innovation performance. Hypothesis H3 was not supported by the findings (non-significant standardized coefficient of - 0.01). Hypothesis H4 predicted that the extent of international sources of knowledge would be positively related to firms' innovation perfor- mance. The results indicate a significant relationship between interna- tional sources of knowledge and firms' innovation performance (positive significant standardized coefficient of 0.25). The results thus support hy- pothesis H4. other findings Other findings will be discussed in terms of the impact of control vari- ables and the relationships between variables. The impact of firm size and region as a dichotomous control vari- able was assessed (western European countries versus eastern European CJNATOT}. i 1 1 «-0.47». •-0.67*' (j^ATOT^)' CJNATOT>- -16,* - CJnatiO")- FIGURE 1 The model of the influence of sources of knowledge on the innovation performance (bolded parameters are fixed; * sig. < 0.05) countries). Although the model fit indices and the structural coefficients of the relationship between independent variables and innovation per- formance did not reveal substantial variations with the introduction of control variables, both control variables were found significantly re- lated to the innovation performance. The results indicate that firms from eastern European countries are significantly less innovative than firms from western European countries (negative significant standardized co- efficient of -0.19), and that larger firms are significantly more inno- vative than smaller firms (positive significant standardized coefficient of 0.20). The results also show that internal, local, national, and international sources of knowledge are significantly correlated among each other. While the correlations between internal and international, local and in- ternational, and national and international sources of knowledge were moderate (correlation coefficient of 0.25, 0.35, and 0.44 respectively), the correlations among internal, local, and national were somewhat higher (correlation coefficient between 0.54 and 0.57). Nevertheless, multi-collinearity was not detected among any of the variables in the multivariate model. Discussion and Conclusion The results presented in the previous section reveal that internal knowl- edge sources are only some of the sources of innovation. Our research confirmed that in-house learning is crucial for firms' innovation perfor- mance; however, interactive learning outside the firm also significantly contributes to innovativeness. According to these results, it is mainly co- operation with international business partners that contributes to inno- vation. The significant, positive, and high path coefficient confirms the im- portance of in-house r&d activities, continuous process improvements, and internal education programs, which together boost firms' innova- tiveness. This means that innovation performance to a great extent de- pends on a firm's own efforts. This is not surprising, given the fact that innovations strongly influence a firm's competitive position in the market. Consequently, firms try to keep the innovation inside the firm, mainly relying on internal knowledge sources. Know-how historically was - and in large measure remains - a kind of knowledge developed within the confines of a firm (already discussed in Hudson 1999), and our results have proven that the boundaries of the firm are still signif- icant for knowledge related to innovations that are central to the core competencies and strategic goals of the company. Nevertheless, the increasing complexity of the knowledge base upon which the production process depends is increasing the social division of labor in knowledge production, yet is also resulting in growing long- term cooperation between firms (Hudson 1999). According to local- ized learning literature, we expected local knowledge sources to posi- tively contribute to firms' innovation performance; however, the results proved the opposite. The findings might at first seem surprising because they indicate that the use of local knowledge sources impedes innova- tion. However, much of the literature warns that sole dependence on local knowledge sources can lead to the lock-in effect, whereby firms are 'locked' into the existing technological trajectory of the local envi- ronment and are unable to continuously develop new products and ser- vices and implement innovations in processes and organization (Visser and Boschma 2004; Malmberg and Maskell 2006). Camagni (1991) has already stressed that firms need linkages with the external business en- vironment. Especially in times of rapid technological change, external (non-local) links might provide local firms with the complementary assets that are needed to adapt to the changing economic and techno- logical environment. In areas of production characterized by fast in- novation and technological change, 'local firm involvement in wider national and global networks is absolutely essential for long-term re- gional growth,' and 'the milieu has to open up to external energy in order to avoid 'entropic death' and a decline in its own innovative ca- pacity' (Camagni 1991, 139). Our results are not in line with the older literature on localized learning (Capello 1999), which often positioned learning at the local level as somehow superior to that at other spa- tial levels. Nevertheless, our study confirms what most of the recent literature is arguing by saying that innovation performance is a result of combining several internal as well as external knowledge sources, the latter coming from different geographical levels. Local knowledge sources are important for firms to a certain extent, as geographic proxim- ity and concentration of firms can provide enormous opportunities for the transmission of sticky, non-articulated forms of knowledge between firms (Tödling, Lehner, and Trippl 2004). However, localized learning does not necessarily lead to innovation. Our results indicate that ac- cess to codified external knowledge should be secured through inter- action with firms and institutions outside the local environment, and we show that new value is created by combining these various types of knowledge. In today's globalized economy, where supply chains are distributed all around the globe and specialized knowledge and research institutions are scattered in numerous locations, there is no reason to believe that a firm will find the precise knowledge needed in its innovation process within the local environment. Accordingly, firms search for the necessary knowledge elsewhere and often look for appropriate innovation partners irrespective of the geographic space. While our research did not reveal a significant influence of national knowledge sources, it has proven that international sources have a strong, positive, and significant influence on firms' innovation performance. Keeble and Wilkinson (2000) have already supported these ideas with the empirical findings of the Euro- pean network. Numerous firms possess close functional links with firms and knowledge centers in their countries and abroad, and view such wider networks as very important for successful research and innova- tion. Extra-local networking appears to be an important process whereby high-tech firms sustain their innovative activity and competitive advan- tage. Our research confirms that firms need to incorporate the internal learning process with knowledge acquired outside the firm. They need to secure extra-local links in order to secure the inflow of new knowl- edge needed in the innovation process and prevent the lock-in effect. As Oinas and Malecki (2002) suggest, the innovation system can be un- derstood as being internationally distributed and not only as an activity primarily confined within a given local environment. In line with their approach, Simmie (2006,133) suggests that 'innovation must be under- stood in terms of trading nodes in an international system that encom- passes both local and international knowledge spillovers and multilay- ered economic linkages extending over several different spatial scales.' To sum up, one can conclude that internal learning and interactive learn- ing with firms and institutions in a wider business environment mutu- ally reinforce each other and bring optimal results in terms of innova- tion performance. In this respect, our results are in line with existing studies (Caloghirou, Kastelli, and Tsakanikas 2004; Love and Mansury, 2007) that verify the importance of external sources and imply that in- novations come from a number of sources and develop in a number of ways (Willoughby and Galvin 2005). However, those studies mainly fo- cus on the type of sources (for example suppliers and customers, scien- tific system, public institutions, etc.) but do little to explain how location of those knowledge sources influences the innovation performance of firms. In this respect our study brings additional insight into the com- plex process of innovation and proves that not all external knowledge sources are equally important for innovation. According to our results, firms need to establish and nurture collaboration with different partners in the wider international environment in order to boost their innova- tiveness. Although this study has many strengths, it also has some limitations that need to be acknowledged. Firstly, with regard to local knowledge sources, the problem of knowledge internalization deserves mention; that is, when firms overestimate the role of in-house activities and down- grade the role of the local environment in which they operate. The knowledge exchange between local firms and institutions mainly hap- pens in a socialized way in the form of knowledge spillovers. As soon as a firm acquires this local knowledge, it incorporates it into the existing knowledge base, making it internal to the firm (Henry and Pinch 2000; Cole 2007). Accordingly, firms might underestimate the importance of being located in the local environment, because they take for granted the benefits of the specialized local labor market, the proximity of simi- lar firms, and close linkages with local universities and other knowledge organizations. Secondly, the model of the influence of sources of knowl- edge on innovation performance is not comprehensive (it includes a limited number of elements in order to make the empirical examination feasible) because it ignores some other factors that influence innova- tion performance. Thirdly, although the causal directions hypothesized in the model were suggested by the theory, the cross-sectional nature of this study cannot prove the causation but can only support a set of hypothesized paths (Kline 2005). Therefore, the possibility of reverse causality cannot be eliminated. References Asheim, B., and A. Isaksen. 1997. Location, agglomeration and innovation: Towards regional innovation systems in Norway. European Planning Studies 5 (3): 299-330. Bathelt, H., A. Malmberg, and P. Maskell. 2002. Clusters and knowledge: Local buzz, global pipelines and the process of knowledge creation. druid Working Paper 02-12. Belussi, F., and L. Pilotti. 2000. Knowledge creation and collective learn- ing in the Italian local production systems. Http://www.decon.unipd .it/assets/pdf/dp/0021.pdf. Belussi, F., F. McDonald, and S. Borras. 2002. Industrial districts: State of the art review. Research report, Project West-East id: Industrial dis- tricts re-location processes; Identifying policies in the perspective of the European Union enlargement. Bentler, P. M., and D. G. Bonett. 1980. Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin 88 (3): 588-606. Bentler, P. M., and E. J. C. Wu. 2006. eqs 6.1 for Windows. Encino: Multi- variate Software. Brislin, R. W. 1970. Back-translation for cross-cultural research. Journal of Cross-Cultural Psychology 1 (3): 185-216. -. 1980. Translation and content analysis of oral and written mate- rial. In Handbook of cross-cultural psychology, ed. H. C. Triandis and J. W. Berry. Boston, ma: Allyn and Bacon. Britton, J. N. H. 2003. Network structure of an industrial cluster: Electron- ics in Toronto. Environment and Planning 35 (6): 983-1006. Browne, M. W., and R. Cudeck. 1992. Alternative ways of assessing model fit. Sociological Methods Research 21 (2): 230-58. Byrne, B. M. 2006. Structural equation modeling with eqs: Basic concepts, applications, and programming. 2nd ed. Mahwah, nj: Erlbaum. Caloghirou, Y., I. Kastelli, and A. Tsakanikas. 2004. Internal capabilities and external knowledge sources: Complements or substitutes for in- novative performance? Technovation 24 (1): 29-39. Camagni, R. 1991. Local milieu, uncertainty and innovation networks: To- wards a new dynamic theory of economic space. In Innovation net- works: Spatial perspective, ed. R. Camagni, 121-43. London: Belhaven. -. 1993. Inter-firm industrial networks: The costs and benefits of co- operative behaviour. Journal of Industry Studies 1 (1): 1-15. Capello, R. 1999. Spatial transfer ofknowledge in high technology milieux: Learning versus collective learning processes. Regional Studies 33 (4): 353-65. Capello, R., and A. Faggian. 2005. Collective learning and relational capital in local innovation processes. Regional Studies 39 (1): 75-87. Cohen, W. M., and D. A. Levinthal. 1990. Absorbtive capacity: A new per- spective on learning and innovation. Administrative Science Quarterly 35 (1): 128-52. Cole, A. 2007. The knowledge-based theory of the cluster. A critical re-examination. Papers in Evolutionary Economic Geography 0708, Utrecht University. Craig, C. S., and S. Douglas. 2005. International marketing research. 3rd ed. Chichester: Wiley. Cronbach, L. J. 1951. Coefficient alpha and the internal structure of tests. Psychometrika 16 (3): 297-334. Cumbers, A., D. MacKinnon, and K. Chapman. 2003. Innovation, collab- oration and learning in regional clusters: A study of smes in the Ab- erdeen oil complex. Environment and Planning A 35 (9): 1689-706. Dunning, J. 2000. Regions, globalisation and the knowledge-based economy. New York: Oxford University Press. Eriksson, M. L. 2005. Organising regional innovation support: Sweden's industrial development centres as regional development coalitions. phD diss., Linköping University. Floyd, F. J., and K. F. Widaman. 1995. Factor analysis in the development and refinement of clinical assessment instruments. Psychological As- sessment 7 (3): 286-99. Gambarotto, F., and S. Solari. 2004. The role of reconnection of competen- cies and institutions in the collective learning process Http://www.eco .unipmn.it/eventi/eadi/papers/gambarottosolari.pdf. Geenhuizen, M. 1997. A microscopic view on spatial innovation: The case of the Netherlands. In The dynamics of innovative regions: The gremi approach, ed. E. Ratti, A. Bramanti, and R. Gordon, 265-77. Aldershot: Ashgate. Grabher, G., ed. 1993. The embedded firm: On socioeconomics of industrial networks. London: Routledge. Gunasekara, C. 2006. The generative and developmental roles of univer- sities in regional innovation systems. Science and Public Policy 33 (2): 137-50. Hair, J. F., R. E. Anderson, R. L. Tatham, and W. C. Black. 1998. Multivari- ate data analysis. 5th ed. Upper Saddle River, Nj: Prentice-Hall. Hambleton, R. K. 1993. Translating achievement tests for use in cross- national studies. European Journal of Psychological Assessment 9 (1): 57-68. Henry, N., and S. Pinch. 2001. Neo-marshallian nodes, institutional thick- ness, and Britain's 'motor sport valley': Thick or thin. Environment and Planning 33 (7): 1169-83. Hu, L., and P. M. Bentler. 1999. Cutoff criteria for fit indexes in covari- ance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling 6 (1): 1-55. Hudson, R. 1999. The learning economy, the learning firm and the learn- ing region: A sympathetic critique of the limits to learning. European Urban and Regional Studies 6 (1): 59-72. Johnson, B. 1992. Institutional learning. In National systems of innovation: Towards a theory of innovation and interactive learning, ed. B. A. Lund- vall, 23-44. London: Pinter. Johnson, B., E. Lorenz, and B.A. Lundvall. 2002. Why all this fuss about codified and tacit knowledge? Industrial and Corporate Change 11 (2): 245-62. Keeble, D. 2000. Collective learning processes in European high-technol- ogy milieux. In High technology clusters, networking and collective learning in Europe, ed. D. Keeble and F. Wilkinson, 199-229. Adelshot: Ashgate. Keeble, D., and F. Wilkinson. 1999. Collective learning and knowledge de- velopment in the evolution of regional clusters of high-technology smes in Europe. Regional Studies 33 (4): 295-303. -. 2000. High-technology clusters, networking and collective learning in Europe. Adelshot: Ashgate. Keeble, D., C. Lawson, H. Lawton Smith, F. Wilkinson, and B. Moore. 1998. Collective learning processes and inter-firm networking in innovative high-technology regions. Working Paper 86, escr Centre for Business Research, University of Cambridge. Kline, R. B. 2005. Principles and practice of structural equation modeling. 2nd ed. New York: Guilford. Knudsen, L. G. 2006. Determinants of 'openness' in r&d collaboration: the roles of absorptive capacity and appropriability. Paper presented at the druid-dime Academy Winter 2006 phD Conference, Aal- borg. Lorenzen, M. 1998. Localised learning: Why are inter-firm learning pat- terns institutionalised within particular localities? druid Working Pa- per 05-22. Love J., and M. A. Mansury. 2007. External linkages, r&d and innovation performance in us business services. Industry and Innovation 14 (5): 477-96. Lundvall, B. A. 1988. Innovation as an interactive process: From user- producer interaction to national system of innovation. In Technical change and economic theory, ed. G. Dosi, C. Freeman, G. Silverberg, and L. Soete, 269-349. London: Pinter. -. 1992. National systems of innovation: Towards a theory ofinnova- tion and interactive learning. London: Pinter. Lundvall, B. A., and P. Nielsen. 1999. Competition and transformation in the learning economy: The Danish case. Revue d'Economie Industrielle 88 (2): 67-90. Lundvall, B. A., B. Johnson, E. S. Andersen, and B. Dalum. 2002. National systems of production, innovation and competence building. Research Policy 31 (2): 213-31. Malmberg, A., and D. Power. 2005. (How) do (firms in) clusters create knowledge? Industry and Innovation 12 (4): 409-31. Malmberg, A., and P. Maskell. 2006. Localised learning revisited. Growth and Change 37 (1): 1-18. Oinas, P., and Malecki, E. J. 2002. The evolution of technologies in time and space: From national and regional to spatial innovation systems. International Regional Science Journal 25 (1): 101-31. Pavitt, K. 1998. Technologies, products and organisation in the innovat- ing firm: What Adam Smith tells us and Joseph Schumpeter doesn't. Industrial and Corporate Change 7 (3): 433-52. Santos, D. 2000. Innovation and territory: Which strategies to promote re- gional innovation systems in Portugal? European Urban and Regional Studies 7 (2): 147-57. Sharma, S., S. Durvasula, and W. R. Dillon. 1989. Some results on the be- havior of alternate covariance structure estimation procedures in the presence of non-normal data. Journal of Marketing Research 26 (2): 214-21. Shook, C. L., D. J. Ketchen, G. T. M. Hult, and K. M. Kacmar. 2004. An as- sessment of the use of structural equation modeling in strategic man- agement research. Strategic Management Journal 25 (4): 397-404. Simmie, J. 2006. Do clusters or innovation systems drive competitive- ness? In Clusters and regional development: Critical reflections and ex- plorations, ed. B. Asheim, P. Cooke, and R. Martin, 164-88. New York: Routledge. Steiner, M. 2006. Do clusters 'think?': An institutional perspective on knowledge creation and diffusion in clusters. In Clusters and regional development: Critical reflections and explorations, ed. B. Asheim, P. Cooke, and R. Martin, 199-218. New York: Routledge. Steiner, M., and C. Hartmann. 2006. Organisational learning in clusters: A case study on material and immaterial dimensions of cooperation. Regional Studies 40 (5): 493-506. Thorpe, R., R. Holt, A. Macpherson, and L. Pittaway. 2005. Using knowl- edge with small and medium-sized firms: A systematic review of the evidence. International Journal of Management Reviews 7 (4): 257-81. Tödling, F., and A. Kaufmann. 1999. Innovation systems in regions of Europe: A comparative perspective. European Planning Studies 7 (6): 699-717. Tödling, F., P. Lehner, and M. Trippl. 2004. Knowledge intensive indus- tries, networks and collective learning. Paper presented at the 44th Eu- ropean Congress of the European Regional Science Association, Uni- versity of Porto, Portugal. Tomassini, M, and R. Sarcina. 2005. Knowledge and learning. Research re- port, Project West-East id: Industrial districts' re-location processes; Identifying policies in the perspective of the European Union enlarge- ment. Tsai, K.-H., and J.-C. Wang. 2007. A longitudinal examination of perfor- mance of two ways on innovation in Taiwan: Internal r&d investment and external technology acquisition. International Journal of Technol- ogy Management 39 (3-4): 235-47. Visser, E.J., and R. Boschma. 2004. Learning in districts: Novelty and lock- in in a regional context. European Planning Studies 12 (6): 793-808. Willoughby K., and P. Galvin 2005. Inter-organizational collaboration, knowledge intensity, and the sources of innovation in the bioscience- technology industries. Knowledge, Technology, and Policy 18 (3): 56-73. Wolfe, D. A., and M. S. Gertler. 2006. Clusters from inside out: Local dy- namics and global linkages. In Cluster in urban and regional develop- ment, ed. A. Cumbers and D. MacKinnon, 101-24. New York: Rout- ledge. Yeung, H. W. C. 2005. The firm as social networks: An organisational per- spective. Growth and Change 36 (3): 307-28.