The Impact of Endogenous and Exogenous Forces on Innovation: A Logical Analysis of Regional Innovation Systems in Central and Eastern Europe

The modern economic development paradigm relies heavily on the application and commercialisation of innovations. From leading multinational corporations (MNCs) to small and medium enterprises (SMEs), economic agents engage in innovation activities to sustain their competitiveness. Therefore, harnessing innovation has become a focal interest for many actors. European Union (EU) sees it as a primary tool for reaching strategic goals and achieving sustainable social development (European Union n.d.).

However, the situation is not as straightforward. Among the EU members, the developmental gap is acknowledged as an essential issue and obstacle to development. It is especially true when comparing Central and Eastern Europe (CEE) with Western and Northern members. Among the CEE countries, only two regions can be considered innovative leaders (European Commission, Directorate-General for Internal Market, Industry, Entrepreneurship and SMEs 2021a). And even among these, the development is uneven. Given various transition paths, institutional performance, sectoral differences and internal regional interactions (Isaksen 2001; Tödtling and Trippl 2005; Radosevic 2002), the sub-national regions of CEE represent a vast pool of innovation systems.

Emerging as new market economies after the 1990s, many of these regions considered fast integration into the global economy as a strategy for development (Gereffi 2014; Ernst 2002; Ernst and Kim 2001). This factor created additional opportunities associated with technology and knowledge transfer, learning and adoption skills (Giuliani, Pietrobelli, and Rabellotti 2005; Pietrobelli and Rabellotti 2011) and required constant bargaining and negotiations (Gereffi 2001; Dicken 2005). Therefore, it is to acknowledge another set of parameters in the CEE regional innovation.

Combining internal particularities and external conditions creates an avalanche of developmental possibilities. Therefore, the question emerges of what practices were the most useful. And among all the cases, what were the best strategies to follow? What constitutes an efficient innovation process in CEE, and what parameters can be universally applied? To answer these questions, the research will follow the first Research Question (RQ):

RQ1: What endogenous and exogenous factors contribute to innovation performance in CEE regions?

However, observing the impact of separate forces might not be conclusive. Sometimes accumulation of several factors is crucial in creating an adequate environment for innovation (e.g., Pietrobelli and Rabellotti 2011). It is especially true in the context of CEE, with its vast pool of regions and unique developmental paths. Various combinations of regional and extra-regional phenomena can have parallel yet similar effects. Therefore, the article's second aim is to understand what forces complement each other and empower innovation (RQ2).

RQ2: What combinations of endogenous and exogenous forces contribute to regional development and innovation?

Identifying the paths to innovation is imperative to solving the developmental gap in the EU. As such, it will reveal the practices to focus on in formulating new public policy and strengthening existing strategies. Moreover, it can help understand the CEE dynamics exclusively. Researching innovation conditions specific to the area should discourage local governments from copy-pasting ‘western practices’ and focusing more on the solutions that fit their context.

Concerning the sub-national and regional levels of analysis, the innovation process is thought to result from local cooperation and engagement. The Regional Innovation System (RIS) approach highlights the importance of voluntary interactions (social and economic), joint ventures and support forces as its main mechanisms (Asheim, Smith, and Oughton 2011; Doloreux and Porto Gomez 2017). When actors engage in a collaborative activity, it creates conditions for ‘learning by interaction’ (Cooke 1992). Learning builds up competencies that companies can further utilise in their activities. These are means to strengthen their absorptive capacity, contributing to the implementation of innovation (Kallio, Harmaakorpi, and Pihkala 2010; Tödtling and Trippl 2018). Knowledge and information are critical in this matter (Cooke 2003; Asheim and Coenen 2006; Asheim et al. 2005). Through their diffusion, enterprises gain innovative possibilities and replace old technology (Arieli et al. 2020). Moreover, depending on knowledge's practical and scientific utilisation, it can contribute to various learning and innovation patterns (Asheim et al. 2005).

Knowledge and learning are the backbones of innovation activities, so a network is the ‘locus for innovation’ (Spender et al. 2017; Radosevic 2002). It acts as a platform for knowledge and information sharing. Depending on its particularities (structure and links), shared knowledge's speed, degree and volume vary (Burt 2000; Wani and Ali 2015).

The importance of network characteristics and strength of ties has long been a parameter in the socio-economic field (Granovetter 1973, 1985). Therefore, dense connections and strong ties enable deeper embeddedness. With it, a firm's boundaries become ‘porous’ (Chesbrough 2003), leading to more interaction, learning and distribution of knowledge into the network. Therefore, the structure and connectivity in a network can shape (and are shaped) by the processual innovation activities. Stuck, Broekel and Revilla Diez (2014) considered network aspects in articulating RIS typology. According to them, the innovation practices are related to collaboration architecture and partnership preferences. Therefore, it allows us to formulate the first hypothesis:

H1: The network and networking parameters are directly associated with innovation performance.

Besides the interaction dimension, the RIS paradigm considers the available assistance and innovation-facilitating institutions. The theory differentiates them as the Support Subsystem (Tödtling and Trippl 2005; Coenen, Moodysson, and Asheim 2004; Asheim and Isaksen 1997), whose main purpose is to create an efficient environment for innovation (Isaksen 2001). Their activity

These are defined by the functionality and services provided by specialized institutions and organisations.

When discussing the network externalities and support factors, many academics touch on the issue of cultural and moral norms. It can take the form of social capital in a network structure, as means to generate entrepreneurial benefits (e.g., Burt 2000), or as a parameter in the institutional framework (e.g., Zukauskaite 2018). This attention comes from the role that regional mindset plays regarding the other two aspects. Shared norms, visions, problems and attitudes are important factors for developing collaboration (Boschma 2005; Bland et al. 2010). Social, cultural and institutional proximities contribute to easier contact and reduce collaboration costs. By ‘smoothing’ the interaction, it empowers the network learning abilities and therefore contributes to the innovation process (Lin and McDonough 2014).

The second argument to consider is an agent's perception and individual abilities reflected as mental potential. This approach can be traced back to Schumpeter's vision of entrepreneurship (Schumpeter 1983 [1934]; Śledzik 2013). In his first work on innovation, he described the entrepreneur as the actor of development. Through personal traits, he implements innovation and causes Creative Destruction. This is now known as the Mark I model of innovation (Croitoru 2012), which implies introducing new technology through establishing start-ups and continuous new entries.

Given its importance, some scientific works capture cognitive frames as a separate parameter to explain innovation (Modic and Rončević 2018; Cepoi and Golob 2017). Acting as both network and agent characteristics, the mindset contributes to implementing different innovation practices and enables innovation ambidexterity

The ability to engage in knowledge and technology exploration and exploitation simultaneously.

With the liberalisation and globalisation of the economy, endogenous factors explain only a part of the developmental opportunities available to regions. The integration into modular production chains, and their embeddedness in a regional context, brings additional possibilities for value creation, capture and enhancement (Henderson et al. 2002; Yeung and Coe 2015; Coe, Dicken, and Hess 2008). These Transnational Value Chains (TVCs) permeate multiple cultural and institutional frameworks, linking various RIS on a global scale (Dicken 2005). Maintaining such connections is important for avoiding ‘locked-in’ innovation systems (Tödtling and Trippl 2005). The cooperation with TVC ambassadors in regions diversifies the knowledge pool (Bathelt, Malmberg, and Maskell 2004; Martin and Moodysson 2013). In some cases, the TVC representatives become leading actors in the regional network, manifesting control over innovation cooperation and practices (Stuck, Broekel, and Revilla Diez 2014). Therefore, they dictate the innovation vector and the network composition (e.g., local sector and group) to benefit from their information.

Integrating into modular production triggers technology, knowledge and skills transfer, enhancing local learning (Pietrobelli and Rabellotti 2011; Giuliani, Pietrobelli, and Rabellotti 2005). As the relationship progresses, the volume, sophistication and quality of knowledge passing to the region increases, offering further development paths for local RIS (Ernst 2002, 2009). Therefore, TVCs’ adaptation into a regional framework contributes not only as a source of new knowledge but also as pressure and motivation to gradually improve absorptive capacity. Depending on the regional production competencies, the ability of regions to satisfy the demand of production networks, and the governance system within the TVC, regions can develop different models of ‘upgrading’ and value capture (Gereffi 2014; Ernst 2002). This process is associated with continuous bargaining between the TVC leaders and regional stakeholders. As such, the embeddedness conditions and the position in the production process will create different innovation possibilities. Hence, the fourth hypothesis:

H4: The embeddedness conditions of TVC in a regional context can determine regional innovation performance.

Endogenous and exogenous factors amplify regional conditions contributing to innovation. Yet, many of these parameters act simultaneously, making it difficult to observe any individual impact. Different contextual combinations create distinct parameters and offer divergent opportunities to regional stakeholders. In such a case, the analysis cannot end with testing the aforementioned hypotheses, as it has to consider an inclusive model where the same outcome can be achieved through different paths.

To analyse the impact of endogenous and exogenous forces on innovation performance in CEE, the research will consider eight regions. The sample includes Continental Croatia (Croatia), Goriska (Slovenia), Moravia-Silesia (Czech Republic), Nord (Republic of Moldova), Sofia (Bulgaria), Zakarpattya (Ukraine), Cluj and Timis (Romania). As the goal is to see general patterns of the innovation process, this sample offers the necessary diversity. It represents a wide spectre from peripheric regions (e.g., Zakarpattya, Nord), semi-periphery (e.g., Timis) to central, capital administrative units (e.g., Sofia). Moreover, most of these regions have dissimilar national contexts, which excludes the observation bias on this level.

Given the small number of cases and the intent to answer the second research question (RQ2), building a statistical model is unsuitable for the study. Therefore, the research design considers Qualitative Comparative Analysis (QCA) (Ragin 2008, 2014). QCA allows the evaluation of the association between variables (the outcome and conditions leading to it) based on a set relationship. It is not as competent to offer generalisations as statistic regression (Seawright 2005), but the asymmetric approach offers the possibility to establish equifinality of the outcome (Wagemann and Schneider 2010) based on Boolean logic (AND, OR, NOT). The QCA operates on identifying Sufficient (when a condition is a subset of the outcome) and Necessary (when the outcome is a subset of a condition) associations with the outcome.

The data for the analysis was collected through the Jean Monnet Centre of Excellence, entitled ‘Technology and Innovations in Regional Development for Europe 2020 (TIR2020)

Jean Monnet Center of Excellence, Technology and Innovations in Regional Development for Europe 2020 (TIR2020), Co-funded by the Erasmus+ programme of the European Union, Key Action: Cooperation for innovation and the exchange of good practices, Action Type: Knowledge Alliances for higher education, Project Reference: 587540-EPP-1-2017-1-SI-EPPJMO-CoE. More information available at https://www.tir2020.net/

Each question offered four labels describing a possible regional scenario (Annex 1). Respondents had to unanimously agree on one score that best describes the region. Besides offering additional validity,

When all the respondents agreed on a score easily, it indicated that the label describes the regional situation adequately. On the contrary, when a debate emerged, it put respondents’ opinions against each other, which decreased the impact of personal or professional bias.

The data collection ended with a database of ordinal scores (1–4). Therefore, the research will rely on the fuzzy-set QCA (fsQCA) that can accommodate a non-binary variation. The fsQCA assesses set relationship based on threshold comparison (Seawright 2005) after calibration – i.e., determination of set membership (Ragin 2008). The lowest score is associated to non-membership (calibration value 0), and a score of 4 represents full membership (calibration value 1). The mid scores indicate a position of ‘more out than in’ (score 2, calibration value 0.33) and ‘more in than out’ (score 3, calibration value 0.67).

To determine sufficient and necessary conditions, the QCA methodology relies on the indicators of Consistency/Inclusion (strength of a relationship) and Coverage (importance of a relationship). In fsQCA, both operate on the premise of calibration variation between the outcome and the condition (Table 2). Auxiliary parameters include Relevance of Necessity (RoN) as indicators of the irrelevance of a condition or combination of conditions. For example, it has low values when a condition is constant (or close to it) and thus unimportant in the analysis (Ragin 2008; Mattke et al. 2022). Further, Proportional Reduction of Inconsistency (PRI) indicates the degree to which a condition is associated with the outcome and its negation (Mattke et al. 2022). This is important to consider by comparing the PRI between cases to understand which has a better (meaningful) relationship.

Given the low number of observations and a vast number of variables in the analysis, the research will deviate from the suggested thresholds of sufficiency and necessity (Mattke et al. 2022). The benchmark is selected based on the sample context. For necessity, the validation criteria are set to accommodate one divergent case out of eight possible (inclusion, consistency and RoN, at 0.875). For sufficiency, the consistency threshold (that validates the empirical evidence of a condition) is set at the same value. However, sufficient conditions require less coverage, based on the possibility of having multiple conditions leading to the outcome. Consequently, the analysis will consider 0.5 as a validation point (4 out of 8 cases). This can be enough to suggest that a condition has a meaningful impact on CEE and should be considered in future policy.

The fsQCA methodology operates on the premise of the association of variables and the alignment of calibrated values. It neither provides proof of a relationship nor indicates the vector of interdependency. The strength of QCA comes through the possibility of validating certain individual conditions and combinations of conditions through empirical examples and theoretical models. Therefore, the associations (especially cases that include conjunction and disjunction of different conditions) will be critically examined through a logical and theoretical lens.

The analysis revealed several sufficient conditions that met the necessary thresholds (Table 3). Among control conditions, R.D and OI showed high inclusion and coverage. Their association with innovation indicates that the degree of regional research and development (R&D) engagement and the adherence to Open Innovation practices is a good way to describe the innovation process, even in CEE. Most numerous are institutional conditions. Institutional support (InstSupp), capacity to retain talented people (Retain) and the implementation of regional innovation policy (InnoPolicy) showed high covariation with innovation output. A possible interpretation of this finding can be the prevalence of Institutional RIS (Cooke 2003; Asheim et al. 2005) in the CEE macro-region. Even in a system where the private sector determines the innovation conditions, the supportive public framework can contribute to institutionalising good practices. The capacity to retain people can be interpreted from the perspective of the brain drain and depopulation crisis in the area. Therefore, the association with innovation can come from the fact that people settle and stay in regions with good living and economic conditions, which usually result from good innovation and competitive performance. Another important factor is considering the human resources associated with educated personnel, which is also a part of the European Innovation Scoreboard (Bielińska-Dusza and Hamerska 2021; European Commission, Directorate-General for Internal Market, Industry, Entrepreneurship and SMEs 2021b).

Regional cooperation (RegCoop) is the only sufficient condition from the networking parameters. Its importance fits well with the argumentation provided in the theoretical chapter. By enhancing ‘learning by interacting’ and knowledge dissemination (Cooke 2003), cooperation leads to new combinations within the network. In the case of regional mindset, Competition is the sole condition that meets the threshold. A good attitude toward competitors can be a factor of collaboration, indirectly influencing innovation performance. Its importance might be perceived in the context of adverse effects when hostile sentiment jeopardises the creation of new links and cooperation patterns or impedes imitation and adoption (Anderson and Semadeni 2015). The same is true for the degree of Trust, which indicated high inclusion and coverage, very close to the selected benchmark. Finally, among the TVC parameters, tvcR.D is the only sufficient condition. The implication of TVCs in the regional innovation process is one of the preconditions mentioned in the Global Production Networks theory (GPN, Ernst 2009; Ernst and Kim 2001; Henderson et al. 2002; Coe, Dicken, and Hess 2008). Its inclusion and coverage scores are slightly lower than the R.D condition, raising the question of whether the regional R&D in CEE is guided by the influence of the value chain integration. Their intent to capitalise not only from production means but also innovation potential (Ernst 2009) creates additional opportunities to accommodate and finance local R&D and support other innovation activities. The regional production competence (tvcTIER) is very close to the set values, which supports this claim. As the production sophistication grows, it increases the attention of the MNCs (Breul and Revilla Diez 2018) and the innovation involvement with it.

In the case of necessity, the list is significantly shorter. Only two conditions met the qualification parameters. Regional cooperation (RegCoop) and institutional support (InstSupp) are both sufficient and necessary conditions and are both crucial aspects of the RIS framework. Showing both super and sub-set relationships indicates a high covariation with the outcome. Such a development enforces the conclusion of the importance of the RIS framework, not only in Western and Northern Europe (e.g., the case of Wales and Norway, in Cooke 1992; Asheim and Isaksen 1997) but also its viability in its developing part.

To assess the equifinality of multiple conditions, QCA methodology consists of three possible combinations: ‘AND/*’ (conjunction of two sets), ‘OR/+’ (disjunction of two sets) and ‘NOT/~’ (inversion of the set value) (Ragin 2008; Dușa n.d.). Each indicates a Boolean algebra operation to set merging and combinations. In fs QCA, AND selects the minimal value out of the presented conditions, OR considers the highest value, and NOT indicates a reversed score (1-condition). In this context, the number of possible outcomes rises gradually. To avoid redundancy, QCA performs a minimisation process, selecting only the conditions that have a meaningful (logical) impact and omitting redundant conditions. However, the possibility of combining multiple conditions creates new boundaries for error. The distribution of values for analysing and selecting particular scores among a group of conditions creates a premise to see false associations. These cases will be considered a calculation excess and explained in the context of the analysis. The results for combinations that met the validation threshold are presented in Table 4. Some of them replicate the findings of previous analyses (Table 3), such as the importance of RegCoop, InstSupp, InnoPolicy, Competition, Retain, R.D and OI. The tvcTIER can no longer be considered sufficient alone, indicating that it has a stronger impact when coupled with other conditions.

Some results are to be considered a calculation excess based on the reversion of scores. When a score is low, after the reversion and conjunction, it can meet the inclusion and coverage marks. Cases 1, 8, 10, 11, 15, 18, 19, 21 and 22 (Table 4) pose doubts about their logical and theoretical relevance. For example, a shortage of information flow in the region (Înformation) contradicts the RIS approach and other empirical observations (e.g., Bathelt, Malmberg, and Maskell 2004). Similarly, a region with an absent external collaboration (~ExtCoop) leads to a ‘locked-in’ situation (Tödtling and Trippl 2005), which is an antipode of high innovation potential. Therefore, these cases present limited explanatory potential and will be omitted from consideration.

Some cases can consider a negation. For example, Trust ‘~tvcR.D and ~R.D’ Trust conditions might indicate a general pattern in the CEE. Suppose the local innovation system is not supported by the TVC anchors and does not rely on R&D. In that case, a substitute can be found in building regional trust and enhancing cooperation (e.g., switching to empower local clusters). In this regard, other conditions (13, 20, 23, 24 and 26) support this conclusion. Trust is presented as an important parameter in activating innovation processes. It combines with endogenous and exogenous forces to activate or boost the impact these have on innovation. Network‘ Trust is basically synonymous with regional cooperation, highlighting the same finding again. In the context of TVC factors, it seems that an embedded TVC becomes a part of local interaction when there is trust (especially in cases of higher production modulation). Therefore, both scenarios lead to knowledge diffusion (Information ’Trust). It can also be stipulated that the best possible way to benefit from information sharing is to build a trustworthy regional environment.

The entrepreneurial mindset seems to thrive when coupled with exogenous conditions (17, 27, 28). When R&D is unavailable, local clustering and reliance on the Mark I innovation model (Croitoru 2012) lead to innovation. A possible conclusion in this scenario is that pushing regional start-ups to supply global demand opens additional perspectives for them and gives them ‘space’ for growth. The TVCs consider the importance of regional innovation, and new entries capitalise on regional knowledge (Audretsch, Grilo, and Thurik 2011). Therefore, the innovation system benefits TVCs, and, in return, the local stakeholders gain more value capture.

The attraction of talented people covariates with innovativeness in regions with a high entrepreneurial spirit, a good network and information-sharing conditions. By opening new businesses, sharing information that was not in the region and integrating into the local network, they contribute to innovation. Similar to Retain, this relationship might be indicative of people moving to regions with better socio-economic conditions, where they exploit their talents.

The combination of exogenous factors among themselves (TVCembed ‘tvcTIER and tvcTIER’ tvcEconom) suggests the importance of the governance positioning within TVC for the innovation process. Depending on the production role, regional stakeholders can feel the pressure to innovate that comes from their collaboration with TVC partners. For higher-tier regions, it is through more robust collaboration with the flagships (Ernst 2002). For lower, the increase in demand and production relationships. In either case, strengthening these connections validates the TVC learning argument (Pietrobelli and Rabellotti 2011) and the GPN framework. This can also indicate paths to upgrading (Gereffi 2014), whereas strengthening relational and economic ties with the TVCs leads to increasing production sophistication and moving to higher manufacturing modules.

Concerning the necessary conditions, 86 combinations passed the validation threshold (Table 5). Many of these conditions can be considered a ‘calculation relationship’ rather than real associations. Multiple observations contain negated outcomes which raise doubt about their validity. Especially disjunction of sets acts as means to have numerous combinations that fit the threshold but are not relevant empirically. Therefore, the analysis will consider a joint interpretation of conditions indicating the same conclusions.

Some important observations can be drawn for the RIS approach in CEE. First, the combination Information ‘RegCoop reinforces the argument that interactions contribute to knowledge circulation. Moreover, it suggests other practical conclusions, such that the cooperation should target information sharing and that any policy focusing on knowledge generation should also encourage its utilisation (Audretsch, Grilo, and Thurik 2011). The combination InstSupp’ Trust suggests a precondition for an efficient support subsystem. Considering RIS in developed countries, this precondition might be easily overlooked, given their high degree of general trust. In CEE, however, national institutions face a trust deficit (Eurofound 2018). Therefore, organisations should be investing in their image to efficiently implement decisions and mobilise stakeholders.

The necessity configurations also confirm some previous conclusions. This is related to the importance of R&D in the CEE area. The R.D+tvcR.D combination acts as proof that the implication of TVC anchors supports regional R&D. This replicates the interest of production chains to build their Global Innovation Networks, in which Easter European counties are gradually integrated (Ernst 2009).

Several groups of combinations (5–10, 13–28) represent a conjunction of endogenous and exogenous conditions. InstSupp, Trust, tvcEconom, tvcTIER, and tvcEmbed create different configurations highly associated with innovation. This proves that endogenous and exogenous factors complement each other rather than acting in parallel. On a conceptual level, it can represent the validation of GPN theory (Ernst 2002; Yeung and Coe 2015; Henderson et al. 2002, etc.). Its argument supports various interaction patterns, leading to distinct outcomes on a regional level. The equifinality of so many combinations might indicate the circumstantiality of any systemic innovation. The bargaining potential of institutions and local and global factors determine how these forces will interact. The interest from the TVC, regional competencies, the ability of regional institutions to accommodate TVCs and create preferential conditions, the degree to which supply chain anchors are embedded in a regional context, and the economic and innovation reliance (of TVCs) on regional production will shape the regional innovation process (Yeung and Coe 2015; Rainnie, Herod, and McGrath-Champ 2011; Nilsen 2019). This can also explain why most TVC factors were not considered sufficient or necessary individually. Their impact comes from interacting with local conditions, either leading them or creating various symbiotic relationships (Stuck, Broekel, and Revilla Diez 2014).

The disjunction (OR) configurations will not be considered, given the risk of redundant association. These cases resemble a mathematical excess of the fsQCA, which was triggered by a small number of observed regions and a big number of conditions (Mattke et al. 2022).

Regional innovation represents a very complex concept that shows distinctive parameters in different types of regions based on their internal and historical background (Radosevic 2002; Tödtling and Trippl 2005). The research sought to see developmental patterns in CEE by testing the impact of network configuration (H1), institutional framework (H2), regional mindset (H3) and influence for integration into global value chains (H4) on innovation performance. Additionally, the study identified various combinations of parameters that are associated with innovation in the European macro-region.

The fsQCA analysis partially confirmed all the initial assumptions. Regional cooperation, the potential of support institutions and the efficiency of innovation policy create the basis to consider the first two hypotheses and act in defence of the RIS approach. The results also confirmed that enabling mindset, associated with a positive attitude to local competition and a trustworthy environment, contributes to developing innovative practices. Similarly, the implication of TVC partners in local R&D and innovation can help understand the regional potential.

The study also found that it might be wrong to consider endogenous and exogenous conditions separately. These are compatible and circumstantial to local predetermined factors. Their impact is conjunctive, and the (unique) combinations that emerge in regional contexts create (unique) regional innovation pressures, conditions and opportunities.

In theoretical terms, research provides additional validation of RIS and GPN theories. The regional cooperation and institutional framework are the only sufficient and necessary conditions in the analysis. Their strong association with innovation suggests that regional stakeholders in CEE consider this model. The range of configuration with regional and global factors should urge local authorities to also analyse the global context and enhance their bargaining potential (Coe et al. 2004). Another practical implication is that regions should search to increase the degree of general and institutional trust. It was part of multiple configurations, including regional and global factors leading to innovation. Trust offers necessary network flexibility (the ability to easily switch to the right partner at the right time), which boosts the efficiency of regional cooperation, institutional impact and collaboration/integration into TVCs.

The research offers an extensive range of conclusions and, through fsQCA, a vast list of combinations relevant to innovation. However, it does not offer the possibility to generalise these findings. Many of the configurations reconstruct previously researched models, but others must be tested and proved empirically. The research on innovation is not easy and therefore requires continuous confirmations. Further analysis should verify the validity of these configurations.