The role of universities on the consolidation of knowledge-based sectors: A spatial econometric analysis of KIBS formation rates in Spanish regions

1 This is a pre-print of an article published in Socio-Economic Planning Sciences. The final authenticated version is available online at: https://doi.org/10.1016/j.seps.2020.100900 Horváth, K.; Berbegal-Mirabent, J. (2020). The role of universities on the consolidation of knowledge-based sectors: A spatial econometric analysis of KIBS formation rates in Spanish regions. Socio-Economic Planning Sciences. DOI: 10.1016/j.seps.2020.100900. The role of universities on the consolidation of knowledge-based sectors: A spatial econometric analysis of KIBS formation rates in Spanish regions Krisztina Horváth* Regional Innovation and Entrepreneurship Research Center (RIERC), Faculty of Business and Economics, University of Pécs Pécs, Rákóczi út 80. 7622-Hungary E-mail: [email protected] Jasmina Berbegal-Mirabent Department of Economy and Business Organization, Universitat Internacional de Catalunya (UIC) C. Immaculada, 22. 08017. Barcelona, Spain E-mail: jberbeg[email protected] Abstract This study evaluates how features related to the regional configuration of universities—i.e., the number of universities in a region and the proportion of public universities—influence the regional formation rate of knowledge-intensive business service (KIBS) firms. Using a sample of 47 Spanish regions (provinces) during 2009-2013, the results of the spatial econometric panel analysis give support to the argument that regions with a greater concentration of universities and with higher proportion of public universities attract more new KIBS firms. However, the findings also indicate a substitution effect between these university-based variables and the region’s industry specialization. Thus, new KIBS businesses tend to locate in regions where they expect either stronger knowledge inputs from universities or higher demand from potential industrial customers. We also reveal the presence of interregional spillover effects. The paper offers insights on how territories may attract more knowledge-based businesses by encouraging the development of the local higher education system. Keywords: knowledge-intensive business service (KIBS), university, public university, territorial servitization, spatial econometrics, Spain * Corresponding author 2 The role of universities on the consolidation of knowledge-based sectors: A spatial econometric analysis of KIBS formation rates in Spanish regions 1. Introduction Governments have increasingly devoted considerable resources to finance science, either through tax policy or direct investments. The recent trend of facilitating socio-economic progress via the creation and/or development of knowledge-based economies has resulted in reforms and investments in research and development (R&D) infrastructures, including universities and research centers (Izsak et al., 2013; Cruz-Castro and Sanz-Menéndez, 2016). Additionally, the growing awareness of the importance of universities as key pillars for the consolidation of knowledge-based economies has led European governing bodies to adopt specific policies within the EU 2020 strategic plan aimed at stimulating knowledge creation and diffusion, as well as scientific productivity among universities (European Commission, 2015). In parallel, over the last decade economies have witnessed a drastic change in their industry configuration, and knowledge-intensive business service (KIBS) firms—such as management consultancy or research and development (R&D) services—have become one of the key engines for the consolidation of knowledge-based economies (European Commission, 2012). For instance, KIBS businesses can contribute to regional development (e.g., economic and employment) through the creation and transfer of knowledge to regional partners such as manufacturing firms, a concept recently referred to as territorial servitization (Lafuente et al., 2017; 2019). However, the weight of KIBS firms across territories is heterogeneous and conditioned by several factors, such as agglomeration economies, the quality of the entrepreneurial environment or the density of potential clients (Meliciani and Savona, 2015; Lafuente et al., 2017; Horváth and Rabetino, 2019). The role of universities on regional development invites to research their potential contribution to the creation of KIBS firms at territorial level (Bonaccorsi et al., 2014; Calcagnini et al., 2016; Valero and Van Reenen, 2019). Universities might be strongly connected to the knowledge creation and diffusion processes that KIBS perform as both sources (e.g., by providing students as future employees, business idea incubators or accelerators) and recipients of knowledge (e.g., as clients or via informal connections) (Pinto et al., 2015; Agasisti et al., 2019). Nevertheless, universities’ capacity to facilitate KIBS formation rates in a focal territory may not be homogeneous in terms of knowledge diffusion metrics. Additionally, knowledge spillovers that occur between territories may influence the level of knowledge available within an area (Acs et al., 2009). Therefore, our study seeks to obtain a better grasp of how features related to the regional configuration of universities influence the regional formation rate of KIBS firms. More specifically, we test the effect of universities’ concentration and ownership status—two measures potentially perceived important in the location decision of new KIBS businesses—on KIBS formation rates in 47 3 Spanish provinces during 2009-2013. Besides, our study analyzes whether (and how) the region’s industry specialization conditions the impact of these university-linked factors. Following the debate on the importance of efficiency of public universities from a policy perspective (e.g., Agasisti and Pérez-Esparrells, 2010; Lafuente and Berbegal-Mirabent, 2019), this research aims at investigating the extent to which universities are shaping the creation of new KIBS businesses within and between territories. To account for knowledge spillover processes between neighboring territories, we employ spatial econometric models—i.e., spatial Durbin panel models— that quantify the intraregional and interregional (spillover) effects of the analyzed variables. The analysis of the role of higher education institutions on the formation of knowledge-intensive businesses has been largely sidelined in prior work (Fischer et al., 2018). Yet, the use of spatial econometrics model together with the analysis of direct/indirect effects constitutes a novelty in the field, becoming one of the first studies that evaluate the relationship between universities and KIBS formation rates employing this technique. Accordingly, this work contributes to the existing literature both from a theoretical (i.e., how universities stimulate KIBS business creation at regional level) and a methodological (i.e. using spatial econometrics) angle. The results of the spatial econometric models highlight the positive effect of universities on regional KIBS formation processes. More specifically, we find both an increased role of concentration of universities and public universities in attracting KIBS businesses to the territory. However, we also observe that these effects only prevail in areas with a relatively low industry specialization, which results in lower KIBS formation rates in territories with both high number of universities (high proportion of public universities) and high industry specialization. In line with Horváth and Rabetino (2019), this result shows that market demand and related regional characteristics can have an important modifying effect on the location decisions of KIBS businesses. The remainder of the paper is structured as follows. Section 2 provides the theory that reveals the role of universities on regional KIBS formation processes. For hypotheses development, two university-linked drivers and the potentially moderating effect of industry specialization are considered. As a framework of our empirical analyses, data, variables and estimation strategy are provided in Section 3. Results are presented in Section 4, while Section 5 gives space to concluding remarks. 2. Background theory and hypotheses development 2.1. Growing in universities’ backyard: A fertile ground for new KIBS businesses Increasingly, regional development relies on territories’ capacity to innovate and transfer knowledge to local businesses (Boschma, 2005; Crespo et al., 2014), and KIBS firms play a decisive role in this process (e.g., Lafuente et al., 2010; European Commission, 2012; Corrocher and Cusmano, 2014; Gebauer and Binz, 2019). Based on Miles et al. (1995, p. 18) who first coined the term, KIBS businesses are business-to-business (B2B) service providers that deal with “…economic activities 4 which are intended to result in the creation, accumulation or dissemination of knowledge”. Beyond knowledge diffusion as their main value-added, KIBS businesses may contribute to the territory’s performance via own, internal innovations and as mediators of external innovations too (Muller and Zenker, 2001; Shearmur, 2012). A significant part of KIBS businesses’ knowledge lies in their employees, and therefore, the quality of their human resources is essential (Tether and Hipp, 2002). Basically, two stages of business-employee interactions determine this quality: employee selection— e.g., initial quality of knowledge inputs such as qualification, skills, and experience—and afterintegration employee development. As we see in Pinto et al. (2015), over time, KIBS businesses gain the human capital and relevant expertise required to their operations, and they develop greater independence from external knowledge sources such as science and technology parks. But, when it comes to businesses facing different liabilities—e.g., liability of newness and smallness, and resource constraints (Aldrich and Auster, 1986; Hudson et al., 2001)—tapping external knowledge sources for high-quality human capital and knowledge may be of utmost relevance (Koch and Stahlecker, 2006; Johnston and Huggins, 2018). In this sense, universities may serve as important fertilizers to KIBS businesses and provide them with strategic advantages, especially in their early stages of development. Besides their traditional core functions (i.e., education and research), the growing importance of knowledge and innovation as key pillars of development and growth has led universities to become decisive actors of territorial economic development (e.g., Etzkowitz, 2003; Shore and McLauchlan, 2012; Degl’Innocenti et al., 2019). A relevant task within this so-called “third mission” is knowledge transfer from universities to businesses. According to Laredo (2007), these interactions take place mainly via future employees’ knowledge and skills; but also research contracts and intermediary roles such as internships or consultancy activities. Pinto et al. (2015, p. 1880) added to this list informal relationships, defined as “all types of engagement not supported by a formal contract or agreement”, with an even superior relevance. For instance, informal relationships can be established in conferences, workshops, or via unofficial discussion of research results and business ideas. Businesses—including the KIBS sector—may absorb and convert this knowledge to wealth-creating processes such as new product development, advanced service provision or increased competitiveness (Agasisti et al., 2019). Therefore, universities have turned into one of the most important engines for regional economic development as well as for business consolidation (e.g., Harris, 2001; Hegde, 2005). For instance, Ponds et al. (2010) detected in the Netherlands a positive effect of universities’ research and development (R&D) on regional innovation, while Moutinho et al. (2015) provided evidence that higher university R&D employment in Europe has the potential to reduce unemployment and increase gross-value added figures. From this viewpoint, Spain—the focus of our study—constitutes a special case since universities proved themselves as relevant external R&D partners for innovative businesses (Segarra-Blasco and Arauzo-Carod, 2008; Fernández-López et al., 2019). 5 A traditionally central factor deemed to determine universities’ knowledge diffusion capacity is geographic or spatial proximity (e.g., Siegel et al., 2003; Abramovsky et al., 2007). Although it ceased to be a pre-requisite for university-led knowledge spillover to take place (e.g., Feldman, 1994; Raspe and Van Oort, 2011; Varga and Sebestyén, 2017), scientific evidence shows that it still remained a decisive characteristic of university-business relationships (e.g., Crescenzi et al., 2017; Drejer and Østergaard, 2017). For instance, Bresnahan et al. (2001) revealed that important elements of universities’ “knowledge toolkit” such as provision of skilled human capital, managerial and entrepreneurial skills, are key factors in the formation of regional entrepreneurial and innovation clusters (e.g., Silicon Valley). Besides, Fernandes and Ferreira (2013) found a positive relationship between geographic proximity, KIBS businesses’ cooperation with universities and their number of innovations. After a general insight into universities’ knowledge generation capacity for new KIBS businesses, in Section 2, hypotheses development follows. In the first subsections (Sections 2.2.1 and 2.2.2), we explore universities’ ability to foster regional KIBS formation and articulate the hypotheses for two potential drivers of this process. Then, in Section 2.2.3 we investigate the potential impact of regions’ industry specialization that possibly moderates universities’ effect on the location choices of new KIBS businesses. 2.2. Hypotheses development 2.2.1. Concentration of universities to steer regional KIBS business formation Businesses’ choice to locate close to universities might be conditioned by both their differing need for university knowledge and the quality of this knowledge. First of all, not all businesses require university knowledge to the same extent. For instance, some influencing factors that determine the need for university knowledge are industry characteristics (Anselin et al., 2000), cost of internal knowledge production, acquisition from a larger geographic distance (Harhoff, 2000; Fischer et al., 2018) as well as growth ambition of businesses (Segarra-Blasco and Arauzo-Carod, 2008). Based on the industry characteristics, cost sensitivity and growth—or at least survival—ambitions of new KIBS firms (see Section 2.1 before), they are likely to have a greater need for university knowledge, and thus, a solid reason to locate close to universities. This intuition is supported by Baptista et al. (2011) who showed for the specific case of Portugal that establishing a new university in Portuguese municipalities enhanced the level of business entry in knowledge-based sectors. Second, a link can be developed between the need for geographic proximity and the accessibility—i.e., the codified vs. tacit nature (Foray and Lundvall, 1998)—of knowledge. Audretsch et al. (2005) found that both the knowledge output (research or human capital) and the knowledge orientation (natural science and social science) of universities may condition the location of businesses. Particularly, they observed that in Germany, new knowledge-based businesses from highly innovative industries tend to locate further from universities with better quality research in the field of 6 natural sciences and closer to universities with better quality social sciences research. The effect was the opposite in terms of human capital: firms chose to locate closer to universities with more students in the field of natural sciences, while more potential future employees in social sciences did not evoke this effect. In Spain, Acosta et al. (2011) partially confirmed these results. The authors found a positive significant relationship only between the number of graduates in science and technology and the location of new high-technology businesses. Besides, their results revealed a potential need for the match of specialization. In Italy (Calcagnini et al., 2016) and in Brazil (Fischer et al., 2018) findings partially overlapped with those observed in the previous countries. Finally, the third stream of studies highlights the role of the quality of knowledge on the location decision of businesses. The presence of universities in a territory itself may reflect a good-quality local infrastructure that can positively influence the location decision of knowledge-based service firms (Lafuente et al., 2010). Woodward et al. (2006) found a significant relationship between university R&D expenditures and the number of newly created high-technology plants by county, and Kirchhoff et al. (2007) reported similar results in the US. Based on the findings by Agasisti et al. (2019), co-location with more efficient and better performing universities might be a logical choice— as far as businesses are able to perceive it—too as they may have a higher capacity to foster intraregional economic development. Because businesses tend to lack comprehensive information about the qualities of a given university and especially the qualities of all the universities in an area, perceptions constitute an essential link in establishing university-business relationships. For instance, in their case study Johnston and Huggins (2018) found that in the UK small firms tend to engage in collaborative relationships with universities based on their perceived credibility. However, their judgment is mostly shaped by the credibility of an academic individual (e.g., their achievements, recommendations, and knowledge field) and not of an organization as a whole. Besides, although the authors list several individual traits whose complex assessment would reduce the risk of coupling with the wrong university, most study firms confine themselves to the evaluation of only some of these characteristics. Nevertheless, as opposed to collaboration decisions made by existing businesses in a territory, when it comes to preceding business location decisions a broader context analysis seems to take place to gauge the territory’s knowledge potential. As Alcácer and Chung (2007) suggested—among others—businesses choose to locate in areas that may provide them the most knowledge gains. Given the limited capacity and/or willingness to measure these gains (e.g., efficiency or performance of universities), a good proxy in business decisions could be the spatial concentration of universities. Concentration likely indicates high-quality university knowledge available in the area, due to the presence of competitors that provoke development efforts in the co-locating universities (Porter, 1994). Empirical evidence supports this argument. For instance, Audretsch et al. (2005) revealed that the higher the number of universities in a territory, the closer a new knowledge-based business is located 7 to the closest university. For innovative start-ups in general, Calcagnini et al. (2016) confirmed these results. Valero and Van Reenen (2019) added that a high number of universities is associated with greater regional economic development. Agasisti (2009) and Agasisti et al. (2019) came to the same result, and—supporting the previously assumed beneficial role of co-location with competitors— explained this phenomenon by the higher level of competition among universities that may lead to increased university performance, and trigger economic development. Therefore, taking as a basis the results of the aforementioned studies, we propose the following hypothesis: Hypothesis 1. A higher concentration of universities contributes to a greater KIBS business formation rate in the same region. 2.2.2. Public vs. private universities – Is there a difference in regional knowledge generation? As suggested in the previous section (Section 2.2.1), universities possess different capacities to disseminate knowledge, and thus, they may have a diverse impact on regional KIBS business formation rates. A common distinction originating in the organization theory and made between universities is their public / private status (e.g., Denis et al., 2015). Similar to most European countries, Spanish public universities are state-owned and receive most of their funding from public taxes (Lafuente and Berbegal-Mirabent, 2019). Their financing model implies that although they ask for a tuition fee from their students, the partial coverage of educational costs by the state can make knowledge accessible to a relatively wide range of inhabitants (e.g., even in less developed areas) (Escardíbul et al., 2017). Besides, the use of public money comes with a responsibility to invest this amount efficiently and contribute to the region’s development. The efficient utilization and justification of public funding can be examined via the comparison between the efficiency and performance of public vs. private universities. Most evidence comes from the USA. For instance, Hegde (2005) found that public universities tend to adapt to private sector demand less efficiently, and their patents and research target local innovation needs to a less extent as compared to private universities. Also, in their analysis of more than 7,000 US colleges and universities, Guironnet and Peypoch (2018) found that in terms of education efficiency, public universities perform better while private universities excel at research efficiency. Public universities’ excellence in educational efficiency might be traced back to a more rigorous initial quality filter of students, while higher research efficiency in private universities—consistently with the findings by Hegde (2005)—manifests in a higher rate and quality of publications. Nevertheless, as for the private-public context of universities, results are sensitive to countryspecific characteristics, and therefore, it is important to analyze the knowledge spillover capability of public and private universities in the national environment, which is the actual focus of this study. For Spain, Caldera and Debande (2010) found that public universities have a higher R&D income, higher number of R&D contracts but private universities issue more licenses. Analyzing a more extended set of knowledge outputs, Casani et al. (2014) found that on average public universities outperform 8 private ones in several aspects associated with the quantity and quality of produced knowledge: for instance, in the number of staff with PhDs, research productivity, percentage of papers published in first quartile (Q1) in Web of Science, and in engagement in innovation. In terms of efficiency, de la Torre et al. (2017) found that the greater flexibility of private universities seem to help them make a better adjustment between inputs and outputs than public ones. However, this result is no longer valid after the economic crisis. Data from the period 2013-2014 revealed that public universities have caught up with private ones although having their inputs substantially reduced. However, against the higher state expectations towards public universities, evidence shows the limited capacity of Spanish public universities to excel at each of their three hypothetical missions (Berbegal-Mirabent et al., 2019). As Sánchez-Barrioluengo (2014) found, while research and knowledge transfer activities develop in synergy, teaching is less compatible with these activities. But, in Spain, on average, public universities seem to outperform private ones in terms of the quality of graduates and research—i.e., knowledge outputs with the most impact on territorial development—and thus, they may have a more pronounced positive effect on intraregional development and on KIBS formation rates at territorial level. Keep in mind that in Spain the territorial landscape is dominated by peripheral regions. In these geographies public universities can be of great relevance to less developed businesses, as they can provide them cost advantages due to their proximity and more affordable service offerings. Innovative businesses can also benefit from their presence as a compensation of missing complementary services (e.g., professional KIBS for technological KIBS businesses) (Pinto et al., 2015). Consequently, in the context of our study, we hypothesize that: Hypothesis 2. A higher proportion of public universities contributes to a greater KIBS business formation rate in the same region. 2.2.3. The moderating role of the region’s industry specialization In addition to potential knowledge gains from universities, market demand likely constitutes an important location factor for KIBS businesses. Besides, beyond a pure trade relationship, KIBS’ existing knowledge stock can develop with accumulated experience on customized and time-changing problem solving for their business clients (Tether and Hipp, 2002; Scarbrough et al., 2004). More specifically, a business segment that has proved to be one of the most important clients of KIBS businesses is the industry sector, and more specifically, the manufacturing industry (den Hertog, 2000; Vendrell-Herrero et al., 2017; Bustinza et al., 2019). For instance, Bienkowska (2015) documents that a remarkable share of jobs in the manufacturing industry in Europe (40%) is related to high value added services. In addition, over the last few decades, manufacturers’ demand for knowledgeintensive services has increased due to various emerging and valuable forms of cooperation such as outsourcing (e.g., Hätönen and Eriksson, 2009) and combined product-service offerings, that is, business servitization (e.g., Vendrell-Herrero et al., 2017; Bustinza et al., 2019). 9 By interacting with customers, KIBS firms face the challenge of both dealing with unprecedented cases that require additional knowledge and managing frequent interactions in an efficient way (Garicano and Wu, 2012). Therefore, and similar to accessing university-based advantages, building (tacit) knowledge through interactions with clients likely requires geographic proximity, in particular, at the beginning of their lifecycle and establishing their client base (Koch and Stahlecker, 2006). For KIBS businesses, proximity to clients has been identified as a crucial location factor by several scientific studies (e.g., Keeble and Nachum, 2002; Meliciani and Savona, 2015). In this regard, there is a growing stream in the academic literature that is gaining momentum investigating the characteristics and territorial effects of the interactions between KIBS and manufacturing sector, recently referred to as “territorial servitization” (Lafuente et al., 2017). Among others, this research line suggests that territories with a consolidated manufacturing base will likely have greater KIBS formation rates (e.g., Lafuente et al., 2017; Gomes et al., 2019). Notwithstanding, this also implies that a pure mass of manufacturers does not automatically induce the spread of new KIBS businesses in a territory. Instead, the scientific literature points to the relevance of a favorable entrepreneurial environment that can serve as a seedbed for elevated KIBS formation. As Acs et al. (2014) describe, the relevance of a healthy, developed entrepreneurial environment—or “entrepreneurial ecosystem” (p. 490)—lies in its high-quality business actors and institutional background that may serve as a catalyst to new business success through its selfreinforcing processes. In terms of interaction with manufacturers, the findings by Horváth and Rabetino (2019) demonstrate two regional development paths for KIBS businesses. First, they reveal that a healthy entrepreneurial ecosystem (e.g., via the provision of developed digital infrastructure and high-quality human capital) per se is able to stimulate KIBS formation in a territory. Second, they find that increased manufacturing specialization only spurs KIBS formation if the region accommodates a healthy entrepreneurial ecosystem. Some ecosystem elements that have proved to be relevant for elevated KIBS formation are the size, dynamism (e.g., generation of spin-off companies) and cooperation willingness (e.g., need for external R&D efforts) of their local client base (Koch and Stahlecker, 2006; Bellandi and Santini, 2019). Similarly, Wyrwich (2019) showed that the recognition of new, large-scale business opportunities in a territory can also contribute to an increased start-up rate of its KIBS businesses. Based on the KIBS generation competence of the industry sector—that mostly incorporates manufacturing businesses—, we hypothesize that it strengthens the positive effect of universities— both in terms of concentration of universities and the proportion of public universities—on KIBS formation in the region. Accordingly, we conjecture that: Hypothesis 3. The region’s industry specialization moderates the positive relationship between concentration of universities and KIBS business formation rate in the same region. Hypothesis 4. The region’s industry specialization moderates the positive relationship between the proportion of public universities and KIBS business formation rate in the same region. 16 universities is higher as compared to private universities. Therefore, we confirm our second hypothesis (H2). These findings correspond with our complementary analysis in which regional level effects are clearly separated from the effects coming from the surrounding territories (direct effects). Nevertheless, when we look at the results in Model 2, it turns out that universities’ positive effect on regional KIBS formation—both in terms of the number of universities and the proportion of public universities in the territory—is not homogeneous across space but conditioned by the level of the region’s industry specialization. Based on our theoretical underpinning, it is reasonable to think that territories with both a higher industry specialization and more universities (higher proportion of public universities) would contribute to enhanced KIBS business creation rates. However, as indicated in Model 2, this is not the case: a substitution effect exists between the industrial specialization and the given university-based characteristics of a region. To better gauge and interpret these effects, we estimated and plotted the marginal effects of university-based variables in regions with low and high industry specialization using pooled OLS models (Figures 2 and 3). Figure 2 illustrates the moderating role of industry specialization on the effect of concentration of universities—measured by the number of universities in a region—on regional KIBS formation rate. On the one hand, and against our expectations, regions with various universities only experience a more pronounced KIBS formation rate if they have a low industry specialization. On the other hand, however, regions more specialized in industrial activities, even if their university landscape is scarcer, have the capacity to attract new KIBS businesses. Consequently, we cannot give support to our third hypothesis (H3). A logical explanation to this phenomenon is that these locations are chosen by new KIBS businesses for different reasons. First, regions with a strong university base, even if they have a relatively low industry share, may be able to channel valuable resources such as competitive knowledge and innovative thinking to new KIBS businesses. This can compensate for a less extensive local industrial client base as it can make knowledge-based services attractive at extraregional level or even in an international context. Some regions’ capacity to facilitate interaction with clients at a spatially less bounded level is evidenced by Keeble and Nachum (2002) who found that global activity of KIBS businesses may be more intense in territories characterized by collective learning and networking opportunities. Second, regions with high industry specialization—as described in Section 2.2.3—may provide twofold advantages to new KIBS businesses: a critical mass of potential customers and experiential learning via interactions with customers. Nevertheless, based on the results of Model 1, the nonsignificant, unclear effect of industry specialization raises awareness to the heterogeneity and different capacity of these areas to contribute to regional KIBS formation. 17 Table 2. Spatial Durbin model: Regression results (Dependent variable: KIBS formation rate, number of regions: 47, time period: 2009–2013, N=235) Model 1: Baseline model Model 2: Full model Coefficient (Std. error) Direct effect Indirect Effect Total effect Coefficient (Std. error) Direct effect Indirect effect Total effect NumberUniv (ln) 0.8865** (0.3858) 1.0912** (0.4360) 3.1916 (2.2572) 4.2829* (2.5098) 1.4450*** (0.4951) 1.5799*** (0.5117) 3.4166 (2.2975) 4.9966** (2.5080) PropPublic 1.2507** (0.5230) 1.5815*** (0.5902) 5.1155 (3.2033) 6.6970* (3.5495) 3.0407*** (0.9237) 3.3780*** (0.9207) 9.0966* (4.7185) 12.4746** (5.0833) IndSpec 0.0104 (0.0568) 0.0203 (0.0557) 0.1326 (0.1093) 0.1529 (0.1145) 0.2328*** (0.0701) 0.2419*** (0.0799) 0.3102 (0.3919) 0.5521 (0.4507) NumberUniv (ln)  IndSpec –0.0843** (0.0389) –0.0712* (0.0387) 0.1916 (0.1983) 0.1204 (0.2411) PropPublic  IndSpec –0.2280*** (0.0650) –0.2232*** (0.0729) –0.1525 (0.3367) –0.3757 (0.3879) GDP_pc (ln) 0.0424 (0.0690) 0.0677 (0.0737) 0.4176* (0.2455) 0.4853* (0.2811) 0.0898 (0.0731) 0.1315 (0.0826) 0.6323** (0.2735) 0.7738** (0.3265) PopDens (ln) –0.1818 (0.7858) –0.3226 (0.7588) –2.5775 (2.1464) –2.9001 (2.1645) –0.0051 (0.7209) –0.2469 (0.6993) –2.9780 (2.0085) –3.2049 (2.1400) W_KIBS_fr (Spatial rho) 0.3909*** (0.0924) 0.3716*** (0.0918) 18 Table 2. Continued. Note: Time dummies are included in the model specification (2013 is the omitted time category). Robust standard errors adjusted by heteroskedasticity are presented in brackets. “W_” indicates the spatially lagged variables, calculated with row-standardized QUEEN weight matrix. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively. Model 1: Baseline model Model 2: Full model Coefficient (Std. error) Direct effect Indirect Effect Total effect Coefficient (Std. error) Direct effect Indirect effect Total effect W_NumberUniv (ln) 1.6239 (1.3754) 1.5083 (1.5106) W_PropPublic 2.7089 (1.9254) 4.8262* (2.9907) W_IndSpec 0.0762 (0.0812) 0.0649 (0.2224) W_NumberUniv (ln)  IndSpec 0.1884* (0.1055) W_PropPublic  IndSpec 0.0341 (0.1972) W_GDP_pc (ln) 0.2492* (0.1470) 0.3928*** (0.1285) W_PopDens (ln) –1.5667 (1.4223) –1.7360 (1.3155) Time dummies Yes Yes Diagnostic statistics R2 (within) 0.0778 0.0952 Log (pseudo) likelihood value 270.0140 278.5652 Wald test (chi2) 72.64*** 58.02*** Spatial diagnostics LM error (QUEEN) 31.71*** 30.73*** Robust LM error (QUEEN) 20.61*** 13.70*** LM lag (QUEEN) 22.17*** 17.59*** Robust LM lag (QUEEN) 11.07*** 0.56 Wald test: SDM / SAR 11.18* 17.28** Wald test: SDM / SEM 11.68* 16.94** 19 Figure 2. The moderating role of industry specialization on the effect of number of universities on regional KIBS formation rate Figure 3 depicts the outcome of the moderating role of industry specialization on the effect of proportion of public universities. In this case, we can also observe that regions with a more dominant public university segment can have a more pronounced KIBS formation rate, even if the territory lacks the potentially inviting industrial businesses. This indicates that public universities may be able to compensate for some regional economic disadvantages—in our case, the lack of a consolidated industry base—in territories. Therefore, our fourth hypothesis (H4) is also rejected. However, one can see that the effect of public universities exceeds regional boundaries. While public universities have the potential to facilitate KIBS formation in a territory, neighboring regions turn out to have a more pronounced, though less significant (p<10%), positive effect. This result indicates that although regions may benefit from the presence of public universities, the density of new KIBS businesses may be higher in neighboring territories that have a more favorable entrepreneurial environment (e.g., Madrid and Valencia), and are able to offer more economic benefits to new KIBS businesses. Nevertheless, the regional combination of higher industry specialization and a university landscape characterized by more private universities can also encourage a more intense formation of KIBS businesses in the area. As we previously mentioned, for instance, Guipúzcoa and Navarra regions in the Basque Country constitute a positive example, where the density of new KIBS businesses is one of the highest in Spain. -.5 0.5 11.5 Estimated KIBS formation rate (logged value) -5 -4 -3 -2 -1 0 1 2 3 4 5 Number of universities (logged value) Low industry specialization High industry specialization 20 Figure 3. The moderating role of industry specialization on the effect of proportion of public universities on regional KIBS formation rate Finally, our findings also reveal a positive development shift (in terms of GDP per capita), as knowledge spillovers take place from economically developed regions to less developed regions. This phenomenon might be nurtured by regional measures that trigger economic development in less developed regions (e.g., more favorable tax environment), while businesses can still locate close enough to profit from the advantages of more developed neighboring regions. Based on Lafuente et al. (2010), non-economic location factors can also overwrite the profit-oriented thinking of businesses by offering a desirable personal choice for the business founder, in terms of, for instance, quality of life or nice childhood memories from the place where the founder was raised. Our result is thus in line with Gallego and Maroto (2015) and Horváth and Rabetino (2019) who found that KIBS formation is more intense in less developed regions than in already developed territories. 5. Discussion and implications of the results Supported by previous studies (e.g., Audretsch et al., 2005; Agasisti et al., 2019), our results confirm that regions with diverse university system configurations have different capacity to contribute to regional KIBS formation rates. More specifically, we found that a higher concentration of universities can lead to a superior KIBS formation rate in the same region. Also, and supporting the need for public interventions, we uncover that a greater weight of public universities in the region can positively impact KIBS formation rates. Nevertheless, and more importantly, we conclude that these university-led constructive forces are conditioned by the region’s industry specialization. In a regional environment with lower industry share, characterized by either—though with differing knowledge .4 .6 .8 1 Estimated KIBS formation rate (logged value) 0.1 .2 .3 .4 .5 .6 .7 .8 .9 1 Proportion of public universities Low industry specialization High industry specialization 21 support potential and retaining power—a developed university system or stronger public university presence, universities can take the lead in KIBS business creation. In highly industrialized regions, however, where (public) universities’ weight is lower, connecting with industry businesses may constitute the main incentive to KIBS business formation processes. Our results offer various insights for the scientific community, both in terms of the impact of universities embedded in regions with different level of industry activity and the application of an analytical approach that can better quantify the actual regional and extraregional effects. First, it turns out that the concentration and proportion of public universities may play a crucial role in regional knowledge generation, and more specifically, in KIBS business formation. These findings— recognizing the role of the extent of the local customer base—highlight that regions with more universities may serve as entrepreneurial and knowledge hubs and can compensate for a lower share of potential industry clients in the area. Similarly, regions which are potentially less attractive for KIBS businesses in terms of the degree of industrialization in the area, may overcome the shortage of local industry demand with increased public university presence. Second, and especially in terms of justifying the benefits of public universities on regional development, the use of a spatial model separating intraregional and extraregional effects may be important for future empirical analyses. A number of relevant policy implications can be drawn from our results. It should be noted that both the characteristics of new KIBS businesses and the environment should be taken into account when designing future policies for regional development. Based on the findings of this study, and considering the substitution role of regions’ university system and industry specialization, regional policy makers should target most of their resources to create places of excellence for new KIBS businesses with different location considerations. First, to attract new KIBS businesses to less industrialized areas, a differentiation should be made due to quality differences between regions with more universities (e.g., Madrid, Barcelona) and regions in which public universities dominate the university landscape (e.g., Sevilla, Málaga, Alicante). In the first case, regions abundant in universities are likely to follow a natural development path within the country, as new KIBS businesses can enjoy the nurture of an extensive and high-quality knowledge and entrepreneurial community. Therefore, these leading regions should facilitate that their new KIBS businesses can also adapt to international standards and gain a competitive edge in international markets. In the second type of regions, where public universities are more dominant in the university system, policy makers can provide support to public universities by several means. For instance, public universities should design more competitive and transformative programs (e.g., degree programs, incubators for new KIBS businesses) that adjust to the knowledge demand of existing and potentially unexploited economic strengths of the area. However, as Pinto et al. (2015) suggest, because of their public funding, public universities may offer services cheaper than KIBS businesses, and this way crowd them out of the market. Therefore, to avoid this unwanted substitution effect, it is also important that universities exclusively focus on contributing to develop a consolidated stock of 22 new KIBS businesses in the territory. Besides, governmental policy could encourage interregional knowledge transfers from high-quality universities located in more developed regions to improve the services of public universities in less developed areas. Second, in regions with a solid industry base, but with less universities / relatively more private universities, KIBS-industry interactions should be encouraged by public incentives to motivate industry businesses to innovate and acquire knowledge from KIBS businesses. In this role, it may be important to dissolve the potential bottlenecks in the system such as low networking propensity of local industry businesses and high entry barriers for new KIBS businesses (Koch and Stahlecker, 2006). Wyrwich (2019) suggests that developing the local industry sector should be sufficient to foster the market-led emergence of KIBS businesses. Nevertheless, universities can take an active role in this process by matching the demand of the industry sector to the supply of new KIBS firms (e.g., related and competitive services), and thus, the government should compensate universities for their increased third mission contributions. A successful regional transformation story is the case of the Basque Country that managed to move from an old industrial region to the road of economic renewal (Morgan, 2016). 6. Concluding remarks Universities are invoked as relevant knowledge conduits with the capacity to contribute to the consolidation of knowledge-based economies. In this sense, we sought the answer to the economically crucial question of how universities can contribute to the formation of new knowledge-intensive business service firms at a regional level. In this regard, the original contribution of our study stems from being the first in shedding light on the potentially moderating role of a region’s industry specialization on this process. To analyze regions’ KIBS generation capacity linked to universities and their industry share, we used a sample of Spanish provinces for a five-year-period. We employed two variables related to the regional concentration and composition—i.e., in terms of their public or private status—of universities, and the region’s industry specialization that we measured by the proportion of industry businesses in the region. To accurately quantify the intraregional impact of these variables and to separate effects that spill over to neighboring regions, we employed spatial econometric regression models. Our study is not exempt from limitations. First, we use regional level data that limit the depth of analysis in terms of the study phenomenon. Second, against the focus of our study, it should be noted that a high level of KIBS business formation does not necessarily signal healthy development processes in a region. Due to data availability reasons, we do not analyze the weight of universitybased characteristics in the location decisions of KIBS businesses with different value added for the region. Therefore, a differentiation between innovative vs. non-innovative, and exporting vs. nonexporting KIBS businesses could be made. This separation might be especially important, as territories do not benefit from KIBS businesses with different innovation/export potential to the same extent. 23 Third, the innovative and competitive nature of incumbent KIBS businesses as well as the networking characteristics and growth aspirations of the local industry sector may influence regional KIBS formation rates too (Koch and Stahlecker, 2006; Szerb et al., 2019). These are potentially relevant aspects that should be addressed in future studies. Finally, instead of provinces, the analysis of cities or labor market areas could provide us a more reliable picture on the scope of interregional spillover effects. Funding: This work was funded by the Hungarian government through the project titled “Sustainable, intelligent and inclusive regional and city models” [grant number: EFOP-3.6.2-16-2017-00017]. The project was supported by the European Union and co-financed by the European Social Fund. References Abramovsky, L., Harrison, R., Simpson, H. (2007). University research and the location of business R&D. The Economic Journal, 117(519), 114-141. Acosta, M., Coronado, D., Flores, E. (2011). University spillovers and new business location in hightechnology sectors: Spanish evidence. Small Business Economics, 36(3), 365-376. Acs, Z., Braunerhjelm, P., Audretsch, D. B., Carlsson, B. (2009). The knowledge spillover theory of entrepreneurship. Small Business Economics, 32(1), 15-30. Acs, Z. J., Autio, E., Szerb, L. (2014). National Systems of Entrepreneurship: Measurement issues and policy implications. Research Policy, 43(3), 476-494. Agasisti, T. (2009). Market forces and competition in university systems: theoretical reflections and empirical evidence from Italy. International Review of Applied Economics, 23(4), 463-483. Agasisti, T., Pérez-Esparrells, C. (2010). Comparing efficiency in a cross-country perspective: the case of Italian and Spanish state universities. Higher Education, 59(1), 85-103. Agasisti, T., Barra, C., Zotti, R. (2019). Research, knowledge transfer, and innovation: The effect of Italian universities’ efficiency on local economic development 2006− 2012. Journal of Regional Science, 59(5), 819-849. Alcácer, J., Chung, W. (2007). Location strategies and knowledge spillovers. Management Science, 53(5), 760-776. Aldrich, H., Auster, E. R. (1986). Even dwarfs started small: Liabilities of age and size and their strategic implications. Research in Organizational Behavior, 8, 165-198. Anselin, L. (1988). Spatial econometrics: methods and models. Kluwer Academic, Boston. Anselin, L., Varga, A., Acs, Z. (2000). Geographical spillovers and university research: A spatial econometric perspective. Growth and Change, 31(4), 501-515. Anselin, L., Rey, S. J. (2014). Modern Spatial Econometrics in Practice: A Guide to GeoDa, GeoDaSpace and PySAL. GeoDa Press LLC, Chicago, IL. 24 Arnold, J. M., Javorcik, B., Lipscomb, M., Mattoo, A. (2016). Services reform and manufacturing performance: Evidence from India. The Economic Journal, 126(590), 1-39. Audretsch, D. B., Lehmann, E. E. (2004). Mansfield's missing link: the impact of knowledge spillovers on firm growth. Journal of Technology Transfer, 30(1), 207-210. Audretsch, D. B., Lehmann, E. E., Warning, S. (2005). University spillovers and new firm location. Research Policy, 34(7), 1113-1122. Baptista, R., Lima, F., Mendonça, J. (2011). Establishment of higher education institutions and new firm entry. Research Policy, 40(5), 751-760. Bellandi, M., Santini, E. (2019). Territorial servitization and new local productive configurations: the case of the textile industrial district of Prato. Regional Studies, 53(3), 356-365. Berbegal-Mirabent, J., Gil-Doménech, D., de la Torre, R. (2019). Dealing with heterogeneity: an analysis of spanish universities. TEC Empresarial, 13(3), 58-77. Bienkowska, E. (2015). Reindustrialization of Europe: Industry 4.0 – Innovation, growth and jobs, Forum Europe conference [Speech]. Retrieved from European Commission, https://ec.europa.eu/commission/commissioners/20142019/bienkowska/announcements/reindustrialisation-europe-industry-40-innovation-growth-andjobs-forum-europe-conference_en (18/08/2017) Bonaccorsi, A., Colombo, M. G., Guerini, M., Rossi-Lamastra, C. (2014). The impact of local and external university knowledge on the creation of knowledge-intensive firms: Evidence from the Italian case. Small Business Economics, 43(2), 261-287. Boschma, R. (2005). Proximity and innovation: a critical assessment. Regional Studies, 39(1), 61-74. Bresnahan, T., Gambardella, A., Saxenian, A. (2001). ‘Old economy’ inputs for ‘new economy’ outcomes: Cluster formation in the new Silicon Valleys. Industrial and Corporate Change, 10(4), 835-860. Bustinza, O. F., Lafuente, E., Rabetino, R., Vaillant, Y., Vendrell-Herrero, F. (2019). Make-or-buy configurational approaches in product-service ecosystems and performance. Journal of Business Research, 104, 393-401. Calcagnini, G., Favaretto, I., Giombini, G., Perugini, F., Rombaldoni, R. (2016). The role of universities in the location of innovative start-ups. Journal of Technology Transfer, 41(4), 670-693. Caldera, A., Debande, O. (2010). Performance of Spanish universities in technology transfer: An empirical analysis. Research Policy, 39(9), 1160-1173. Casani, F., De Filippo, D., García-Zorita, C., Sanz-Casado, E. (2014). Public versus private universities: Assessment of research performance; case study of the Spanish university system. Research Evaluation, 23(1), 48-61. Corrocher, N., Cusmano, L. (2014). The ‘KIBS engine’ of regional innovation systems: Empirical evidence from European regions. Regional Studies, 48(7), 1212-1226. 25 Crescenzi, R., Filippetti, A., Iammarino, S. (2017). Academic inventors: collaboration and proximity with industry. The Journal of Technology Transfer, 42(4), 730-762. Crespo, J., Suire, R., Vicente, J. (2014). Lock-in or lock-out? How structural properties of knowledge networks affect regional resilience. Journal of Economic Geography, 14(1), 199-219. Cruz-Castro, L., Sanz-Menéndez, L. (2016). The effects of the economic crisis on public research: Spanish budgetary policies and research organizations. Technological Forecasting and Social Change, 113, 157-167. de la Torre, E. M., Gómez-Sancho, J. M., Pérez-Esparrells, C. (2017). Comparing university performance by legal status: a Malmquist-type index approach for the case of the Spanish higher education system. Tertiary Education and Management, 23(3), 206-221. Degl’Innocenti, M., Matousek, R., Tzeremes, N. G. (2019). The interconnections of academic research and universities’ “third mission”: Evidence from the UK. Research Policy, 48(9), 103793. Den Hertog, P. (2000). Knowledge-intensive business services as co-producers of inno-vation. International Journal of Innovation Management, 4(4), 491-528. Denis, J. L., Ferlie, E., Van Gestel, N. (2015). Understanding hybridity in public organizations. Public Administration, 93(2), 273-289. D'Este, P., Iammarino, S. (2010). The spatial profile of university‐business research partnerships. Papers in Regional Science, 89(2), 335-350. Drejer, I., Østergaard, C. R. (2017). Exploring determinants of firms’ collaboration with specific universities: employee-driven relations and geographical proximity. Regional Studies, 51(8), 11921205. Elhorst, J. P. (2014). Matlab Software for Spatial Panels. International Regional Science Review, 37(3), 389-405. Escardíbul, J. O., Pérez-Esparrells, C., de la Torre, E. M., Sequera, S. M. (2017). Tuition fees in Spanish public universities: A regional convergence analysis. Estudios sobre Educación, 32, 197221. Etzkowitz, H. (2003). Research groups as ‘quasi-firms’: the invention of the entrepreneurial university. Research Policy, 32(1), 109-121. European Commission (2012). Knowledge-Intensive (Business) Services in Europe. Publications Office of the European Union, Luxembourg. European Commission (2015). Annual report on the ERC activities and achievements in 2014. Publications Office of the European Union, Luxembourg. Feldman, M. P. (1994). The university and economic development: the case of Johns Hopkins University and Baltimore. Economic Development Quarterly, 8(1), 67-76. Fernandes, C. I., Ferreira, J. J. (2013). Knowledge spillovers: cooperation between universities and KIBS. R&D Management, 43(5), 461-472.