Analyzing the root of regional innovation performance in the European territory

Regional innovation performance is an important indicator for decision-making regarding the implementation of policies intended for regional development. However, regional development policies have led economies to very different competitive positions in matters of innovation. To address these issues, this paper aims to identify the variables that most contribute to the positioning of economies in terms of performance innovation in Europe. The data for this study were collected at the Regional Innovation Scoreboard (RIS). We use a quantitative methodology, through a multivariate statistical technique (Discriminant Analysis). The results suggest that specific innovation strategies explain the competitive positioning of economies within each group of countries. It was possible to demonstrate that economies with Leader classification show greater comparative robustness in the variables “SMEs with Product or Process Innovations”, “SMEs with Product or Process Innovations”, “R&D Expenditure Public Sector” and “Population with Tertiary Education”, constituting an effective instrument of innovation policy. Furthermore, it was possible to show that the economies belonging to the Modest group do not have a competitive advantage in any of the variables under study, thus providing a reflection opportunity for policymakers at this level. The present research identifies which variables are most relevant to the classification considering the regional innovation performance in Leader, Strong, Moderate, and Modest. Several suggestions were given to companies, policy makers, and higher education institutions in the sense that the regions where they operate can improve their innovative performance, which may help to a change in their current classification.

classification obtained by each group. Once the values of these variables are known, this research will allow, a priori, classification, of a group belonging to a region not yet classified.
We anticipate that the results of this research will be a valuable source of information for regional stakeholders. The regions will be able to see which variables referring to innovation, have to improve, in order to increase their innovative performance. Knowing which variables contribute most to the classification of regions, regional stakeholders can establish specific measures to increase the performance of these variables. In this way, regions can more easily change/raise their ranking with regard to its performance. We expect to prove that the technique used to classify the four groups (Leader, Strong, Moderate, Modest) regarding the Regional Innovation Performance is valid. This paper commences with the introduction, where the theme under study is outlined. In the second part, an extensive literature review about RIS3 and RIS is carried out. In part three the followed methodology is described. In the fourth part, the results are presented. In part five results, are discussed and compared with the literature. Finally, we highlight the contributions of the present research, as well as the practical implications, concluding with the limitations of the research and future lines of investigation.

Literature Review
The development of states, regions and cities is uneven, as such, it causes macroeconomic problems, which are not focused exclusively at the regional, national or international level (Rentková, 2018). According to Stejskal (2009), regional development is characterized by a complex set of processes that take place in the respective regions, in order to influence and manage these processes, a systemic and holistic approach is necessary. Stimson, Stough and Salazar (2005) also addressed the issue of a regional definition for development. According to these authors, regional development is the application of processes and resources available in the region, the result of which translates into sustainable development and the desired economic performance for the region, which meets the expectations of people and organizations.
Thus, the analysis of regional disparities is a fundamental issue, when it is intended to guarantee regional development. The term regional disparities means regional differences in the economic, social and ecological levels of the regions under analysis, which cause uneven development across the EU (Rentková, 2018). In this context, it is necessary to clarify the need to differentiate the notion of disparity and diversity. While diversity is a positive phenomenon. Generally, it has its origins in a historical evolution, where the regions present themselves with unique characteristics, which can be translated into strengths and arguments to be used in their own benefit for the region. These characteristics and strengths make them more competitive and sustainable. The disparity is an undesirable difference between regions, which, due to its simple existence, causes an asymmetric development, becoming the weak side of the region that needs to be faced and mitigated (Zdražil andApplová, 2016, Xiao, Du andWu, 2017). The regional disparities that give rise to social inequalities are economic, cultural, infrastructural, of living conditions, and which, ultimately, can provoke regional polarization.
Aware of this reality, the EU supports the development of its regions through a regional development policy, with a view to mitigating these disparities, using a set of measures called regional development policies, which have evolved over the last few decades. Since the mid-1990s, the European Commission has supported regional development, stimulating innovation processes in the regions. These innovative processes include regional development strategies and an exhaustively developed methodology. Cohesion policy operates through structural, regional and social policies designed to balance the resulting economic and social disparities in EU regions.
Several authors have tackled the subject of regional innovation systems (Lopes and Franco, 2019, Cooke, Uranga and Etxebarria, 1997, Asheim and Coenen, 2005, Cooke, 1992, Leydesdorff and Fritsch, 2006). The Regional Innovation Systems reflects an interactive learning process capable of very quickly producing evidence about institutional reactions (Cooke, 1992). A system is defined as a series of discrete elements, as well as the relationships between them (Georghiou, 1993, Lundvall, 2007. Regional innovation systems consist of interactive learning from formal institutions, public or private, which have common interests (Doloreux, 2011, Fagerberg, 2003. A regional innovation system can be defined as a system of innovative networks and institutions, whose purpose is to develop the innovative capacity of companies in a region (Lopes and Franco, 2019).
Recently, smart specialization and RIS3 has emerged. The smart specialization consists of the promotion of efficient use and the enhancement of the real effects of public investment, in the view of economic growth and prosperity of countries and regions Ortega-Argilés, 2013, McCann andOrtega-Argiles, 2015). RIS3 derives from the EC proposal, aiming to regulate the structural financing program for the period 2014-2020 and 2021-2027. These regulations include innovation strategies for smart specialization (Tiits et al., 2015). RIS3 is the local-based integrated economic transformation political agenda, that starts with the identification of the specific characteristics and assets of each region, enabling a process participated by all interested parties, to establish a vision of a sustainable future for the territory (Panciroli, Santangelo and Tondelli, 2020). Consequently, they are a response to the complex development challenges on policy adaptation to the regional context (Lopes et al., 2019b). RIS3 channels economic development efforts and investments to the relative gains in each region, exploring their respective economic opportunities and emerging trends, and taking measures to enhance their economic growth.
RIS3 ensures that the combination of policies and the available instruments in a given regional environment is effective, whether through grants, loans and other type of support, proving to be effective in achieving the desired goals (Costa, 2020, Iammarino, 2018. In this sense, it is valid to affirm that RIS3 represents the evolution of the RIS concept, which encompasses the concept of intelligent specialization. According to the objectives of the research announced earlier, this paper starts from the idea that the competitiveness of European regions is necessarily linked to regional innovation systems, and that a good knowledge of the relationship between innovation and competitiveness is essential to assess its effects and economic policies capable of increasing the level of competitiveness and growth of European regions (Sabatino andTalamo, 2017, Hajek andHenriques, 2017). Therefore, as industries and companies can base their competitiveness on indicators such as innovation, at the territorial level, the regions are also positioned amongst themselves, which characterizes them as more or less innovative. Ideally, the performance of regional innovation should be measured using the complete measurement table of the Regional Innovation Scoreboard (RIS), that is, using regional data for the same indicators applied to measure innovation performance at the country level. RIS (2016), classifies the regions in terms of their Regional Innovation Performance as: Leader, Strong, Moderate, Modest (Lopes, Farinha, Ferreira and Ferreira, 2018a). RIS (2016), encompasses the following 12 variables (table 1): i) Population with Tertiary Education (%); ii) R&D expenditure in the public sector (%); iii) R&D expenditure in the business sector (%); iv) Non-R&D SME innovation expenditures (%); v) SME innovating in-house (%); vi) Innovative SMEs collaborating with others (%); vii) EPO patent applications (per billion GDP); viii) SME with product or process innovations (%); ix) SME with marketing or organizational innovations (%); x) Employment in medium-high/high tech manufacturing and knowledge-intensive services (%); xi) Exports of medium-high/high technology-intensive manufacturing (%); and xii) Sales of new-to-market and new-to-firm innovation by SMEs (%) (Lopes et al., 2018b, Lopes et al., 2019a Corresponds to the number of people in the age group with some type of post-secondary education. The indicator focuses on the population aged between 30 and 34. R&D expenditure in the public sector It represents expenditure on research and development (R&D) which is one of the main drivers of economic growth in a knowledge-based economy. R&D expenditure in the business sector It captures the formal creation of new knowledge within companies.

Non-R&D innovation expenditures in SMEs
Quantifies non-R&D related innovation expenditures as a percentage of total turnover.

SMEs innovating in-house
It is based on the introduction of new products or production processes in SMEs, whether they are products or production processes that have been significantly improved and that companies have innovated internally. Innovative SMEs collaborating with others Measures the degree of participation of SMEs in cooperation for innovation. EPO patent applications Measures the number of patent applications at the European Patent Office. SMEs with product or process innovations Indicates technological innovation, which is measured by the introduction of new products (goods or services) and processes, which are a key ingredient for innovation in manufacturing activities. SMEs with marketing or organizational innovations It captures the extent to which SMEs innovate through non-technological innovation. Employment in medium-high/high tech manufacturing and knowledge-intensive services Indicates the share of employment in the high-tech manufacturing sectors.

Exports of medium-high/high technology-intensive manufacturing
Measures technological competitiveness in a region, that is, its ability to commercialize R&D and innovation results in international markets.

Sales of new-to-market and new-to-firm innovation in SMEs
Measures the turnover of new or significantly improved products for the company as a percentage of the total turnover. Source: RIS (2016)

Methodology
In the present research, the followed methodology is quantitative. The quantitative methodology explains phenomena, based on numerical data which is analyzed using mathematical methods, and mainly statistics (Creswell, 1994, Mills andGay, 2019). The present research aims to identify which of the variables allow to significantly discriminate the four groups regarding the Regional Innovation Performance (Leader, Strong, Moderate, Modest), catalogued in RIS (2016). Thus, this methodology is the most appropriate because, as a rule, it requires large representative samples, selected in order to generalize the results achieved, prediction of cause and effect relationships through deductive reasoning (Yilmaz, 2013).
The data for the present research was collected at the Regional Innovation Scoreboard ( According to Marôco (2018), Discriminant Analysis is a multivariate statistical technique whose objectives are: (i) identification of the variables that best differentiate or "discriminate" between two or more structurally different and mutually exclusive groups; (ii) the use of these variables to create a discriminant function that sparingly represents differences between groups; and (iii) the use of this function to classify new individuals in groups a priori. Being so, for this research, Stepwise Discriminant Analysis was used with Wilks Λ method to identify which of the variables under study allow to significantly discriminate the four groups of Regional Innovation Performance (Leader, Strong, Moderate, Modest).
The normality assumptions of each group were tested, respectively with the Kolmogorov-Smirnov test in the group's Leader (N=53), Strong (N=60) and Moderate (N=85) and Shapiro-Wilk test in Modest group (N=22). According to these tests, all variables have a normal distribution in each of the Regional and in Modest group (SW=0.820; p=0,001) and variable Sales of New-To-market and New-To-Firm Innovations in Leader group (KS=0.260; p=0,00), in Strong group (KS=0.188; p=0,00) and in Moderate group (KS=0.121; p=0,004). This is, because Discriminant Analysis is robust to variations from normality (Marôco, 2018) these variables were used in the analysis.
The assumption of homogeneity of the variance-covariance matrices was tested with the Box M test, which proved to be valid (p = 0.365), accepting the hypothesis of homogeneity of the variance-covariance matrices.
Finally, a classificatory analysis with cross-validation was used to obtain classification functions to allow to predict in which Regional Innovation Performance group new case studies could be classified.

Results
The focus on regional innovation policies is seen as a source of competitive advantage for economies (Hajek and Henriques, 2017). Thus, there is an urgent need to make innovation ecosystems more robust, capable of asserting themselves at a competitive global level (Saftescu et al., 2016). In this alignment, Research and Innovation Strategies for Smart Specialization (RIS3) have been achieving political popularity, especially in Europe (Lopes et al., 2019b, Tiits et al., 2015.
In the EU, it appears that the structural cohesion funds have a positive impact on innovation activities, with the need to increase the capacity to use these funds, by less developed countries and with less innovative capacity (Arbolino et al., 2019, Mohl and Hagen, 2010, Wamser et al., 2013. For such, the behavior of the variables of RIS 2016 was analyzed, to study the agglomeration of countries by innovative performance groups: i) Population with Tertiary Education (%); ii) R&D expenditure in the public sector (%); iii) R&D expenditure in the business sector (%); iv) Non-R & D SME innovation expenditures (%); v) SME innovating in-house (%); vi) Innovative SMEs collaborating with others (%); vii) EPO patent applications (per billion GDP); viii) SME with product or process innovations (%); ix) SME with marketing or organizational innovations (%); x) Employment in medium-high / high tech manufacturing and knowledge-intensive services (%); xi) Exports of medium-high / high technology-intensive manufacturing (%); and xii) Sales of new-tomarket and new-to-firm innovation by SMEs (%) (Lopes et al., 2018b, Lopes et al., 2019a. Thus, Analysis of Variance or ANOVA was used to compare the distribution of three or more groups in independent samples. In this perspective, the individual ANOVA for each of the independent variables considered shows that only the variable Non-R & D Innovation Expenditures cannot be considered as discriminant (p = 0.98) for the usual levels (α = 0.05) (see table 2). In each step of the Stepwise method, the selected variable is the one that minimizes the value of Λ of Wilks, that is, the one for which the greatest differences between groups occur. The process is repeated until there are no significant differences in Λ.   Function 1 is essentially defined by the EPO Patent Applications and Innovative SMEs Collaborating with other variables, positively discriminates the four groups (Λ = 0.091; χ ^ 2 = 507.53; p = 0.000) and explains 89.1% of the variability between the groups (λ = 5,272).
Function 2 is essentially defined by the variables Sales of New-To-market and New-To-Firm Innovations, SMEs with Product or Process Innovations and R&D Expenditure Business Sector, positively discriminates against the 4 groups (Λ = 0.572; χ ^ 2 = 118.28 ; p = 0.000) and explains 6.1% of the variability between groups (λ = 0.364).
Function 3 is essentially defined by the variables SMEs Innovating In-House and Innovative SMEs Collaborating with Others, positively discriminates the 4 groups (Λ = 0.781; χ ^ 2 = 52.49; p = 0.000) and explains 4.7% of the variability between groups ( λ = 0.281).   Table 4 exhibits the coefficients of the classification functions defined for each of the Regional Innovation Performance groups. The competitiveness of European regions is necessarily linked to regional innovation systems and a good knowledge of the relationship between innovation and competitiveness is essential to assess the effects of public policies on increasing the competitiveness and growth of the European regions (Hajek andHenriques, 2017, Sabatino andTalamo, 2017). According to RIS (2016), economies are grouped into four groups, depending on their level of performance in terms of regional innovation (Leader, Strong, Moderate, Modest) (Lopes et al., 2018b).
These functions serve only as an a priori classification for new individuals in the groups, without any discriminative interpretation. Depending on the values of the variables of a new subject and its substitution in the respective equation, these coefficients allow the determination of the group to which the individual belongs as being the one whose classification function allows to obtain a higher score. The pace of development in the regions is uneven and causes macroeconomic problems, according to a complex set of processes related to each territory or country (Rentková, 2018, Stejskal, 2009). Generally, we can affirm that Leader-rated economies show greater comparative robustness in the variables "SMEs   As we can see in table 7

Contributions and Practical Implications
The Regional Innovation Scoreboard (RIS) includes a broad comparative analysis, with regard to the performance of innovation in different regions/countries of the EU, as well as other regions/countries neighboring the EU. Specifically, it shows the impacts on the economy resulting from innovation activities (Kijek and Matras-Bolibok, 2019). RIS classifies countries/regions into four distinctive groups of innovation performance: Leader, Strong, Moderate, Modest (RIS, 2016). Thus, the present paper aims to identify which of the variables that allow to significantly discriminate the four groups regarding the Regional Innovation Performance (Leader, Strong, Moderate, Modest).
With the present research, we concluded that by comparing the Regional Innovation Performance groups originally assigned with those determined by the use of the classification function coefficients, the percentage of countries correctly classified in each group was 83.6%. This points out that with the discriminant functions achieved and without knowing the classifications of the various groups, we would be able to classify 83.6% of them, which represents a high value. We also concluded that the correct classifications by group, regarding Regional Innovation Performance, are 84.5% in Leader regions, 85% in Strong regions, 78.8% in Moderate regions, and 95.5% in Modest regions. The statistical results of the present research also suggest that the technique used to classify the four groups (Leader, Strong, Moderate, Modest) regarding the Regional Innovation Performance is valid and robust. In Table 7, we summarize the results obtained in the present research, clearly identifying which variables contributed the most to the classification obtained by each group. They are ordered from the most significant to the least significant.
Having identified the variables that have the greater weight in the classification of regions, Leader, Strong, Moderate and Modest, it is important to postulate some suggestions for policy makers, companies, and universities to take into consideration, in order to improve the innovation performance of regions where they operate. Recently, EU regional policy makers have had to adjust their actions regarding regional innovation. Regional development strategies, or RIS3, were implemented in 2014 and will be adjusted for the period 2021-2027 (Bilas, 2020). Within this framework, regional policy makers in the EU were urged to adopt regional currencies policies, in order to guarantee concentration and prioritization in certain areas of smart specialization, thus leading to the growth of entrepreneurship and innovation, with the aim of promoting regional economic development (Saftescu et al., 2016, Simion, Paul andMitroi, 2016) That said, for Strong regions to increase their innovative performance, eventually becoming Leader regions, the results suggest that they implement measures in the variables EPO Patent Applications, SMEs with Product or Process Innovations, and Population with Tertiary Education. Regarding the EPO Patent Applications variable, it is essential that local governments support/encourage more financially higher education institutions/research centers when applying for patents (Verspagen, 2006). Universities should be able to mobilize their professors, researchers, and students to develop new products which may be patented. Another important measure is to grant publications in journals indexed to Scopus (Maraut and Martinez, 2014) and Web of Science. On the other hand, in Europe, it is regularly verified that university investigations are listed by the inventors, with universities being excluded from patent ownership. Thus, universities are left without bargaining power with regard to patent ownership, which in turn can lead to a lack of commitment by higher education institutions in joint projects with companies and may even lead to little valuable innovations (Verspagen, 2006). In such a manner, the SMEs with Product or Process Innovations variable are also affected. It is in this variable that the greatest difference lies between the regions classified as Leader and Strong. Higher education companies and institutions must have the capacity to cooperate and establish win-win agreements. Companies need higher education institutions because they have the knowledge. On the other hand, higher education institutions also need companies, as they can receive extra income for the research contracted by companies, as well as receiving the information from companies, necessary to develop their research (Lopes et al., 2018a). Finally, it is essential to raise the values of Population with Tertiary Education, as human resources with higher degrees they have a positive impact, regarding labor and productivity, thus, increasing growth regions. That said, the educational policy in each country should be adapted to the real needs of the labor market (Spisakova et al., 2016).
The weak interactions between companies, government, and universities are pointed out in the literature as one of the main reasons why regions are unable to pay dividends from higher education and R&D activities (Sterlacchini, 2008).
For Moderate regions to increase their innovative performance, and possibly becoming Strong regions, the results suggest that they should implement measures in the EPO Patent Applications and Innovative SMEs Collaborating with other variables, as there is a great disparity in these variables comparing the two regions in particular. Despite the efforts of regional policy makers to strengthen public funds for innovation, they are still not enough, it is recommended that they should be strengthened Švihlíková, 2020, Stanculescu, 2015). Being so, for Moderate regions it is important to reinforce tailored consultancy, which can be done by higher education institutions, or by specialized companies, to increase financial support from regional, specialized, and public organizations. Financing conditions for companies that want to invest in innovation can be made more flexible and easier, thus meeting demand (Lewandowska and Stopa, 2019). The regulatory framework, education, and training of a greater number of specialists and researchers in information and communication technology can also be improved, reinforcing the efforts of the EU institutions to stimulate the international acceptance of innovation results, fostering external associations with institutions active in innovation, supporting investors in the patent application process, significantly increasing the existing R&D infrastructure, including the creation of innovative clusters and platforms, supporting SMEs and higher education institutions in their efforts to invest in innovation and R&D, creation of clusters and innovative platforms to strengthen the connection between companies and higher education institutions (Stanculescu, 2015).
In summary, according to Ponsiglione, Quinto and Zollo (2018) the relevant factors that affect the regional innovation capacity, and differentiate the leading innovation regions from the less developed regions (Modest and Moderate innovators) are: 1) the propensity cooperation; 2) exploration capacity; and competence. These three factors can contribute to reduce the disparities between European regions with regard to the performance of innovation.
The present research is original, pertinent and clearly identifies which variables are most relevant to the classification considering the regional innovation performance in Leader, Strong, Moderate, and Modest. Several suggestions were given to companies, policy makers, and higher education institutions in the sense that the regions where they operate can improve their innovative performance, which may help to a change in their current classification. This research also contributes to a better understanding of the Regional Innovation Scoreboard, as well as clarifying the current literature on the subject under study.

Research limitations and future lines of investigation
With regards to limitations, primarily, the paper only covers the data from countries/regions present in RIS 2016 and new data are available every year.
These data are from EU countries/regions and some countries in the vicinity. Having only data from countries in Europe, it is not possible to carry out an analysis on a global level.
Secondly, the paper covers data from a pre-pandemic context, and probably the results will be different as the pandemic deaccelerates, we cannot stop seeing the emergence of regional and local economies where small geographical and political areas will try to be independent economically withing several sectors (Samarathunga and Weerathunga, 2020). Furthermore, as countries and regions start looking into reopening the economy, guide decisions not only give the pace of lifting mitigation policies but also leverage other measures that may be needed to restore confidence and trust for people to return to pre-COVID-19 behaviors (Chen, Igan, Pierri and Presbitero, 2020). Afar from the immediate crisis response, policymakers will also have to think about sustained measures after the epidemic subsides. What kind of measures will be needed to reboot the economy and ranks in the regions? This is not only a local question for the EU but also for the global economy, as different parts of the world might be in different phases of fighting the virus (Demertzis, Sapir, Tagliapietra and Wolff, 2020).
As for future lines of research after the pandemic, we suggest that qualitative data on policy actions and performance of EU Member States should be collected, with the aim of identifying best practices on innovation, consequently contributing to the development of increasingly intelligent, more intelligent, sustainable policies and greens. Comparative studies can be made between the different years of the RIS. It should be noted that as of RIS 2017, new variables were inserted in the RIS, as previously indicated in the present study, and it is pertinent to investigate these new variables and verify their impact on the measurement of the innovative performance of the regions and what are their practical effects.