When cohesion meets excellence: Analysing the drivers of synergies between EU R&I funding instruments in EU regions

Bachtrögler, Julia; Gabelberger, Fabian; Marques Santos, Anabela; Doussineau, Mathieu Working Paper When cohesion meets excellence: Analysing the drivers of synergies between EU R&I funding instruments in EU regions JRC Working Papers on Territorial Modelling and Analysis, No. 05/2025 Provided in Cooperation with: Joint Research Centre (JRC), European Commission Suggested Citation: Bachtrögler, Julia; Gabelberger, Fabian; Marques Santos, Anabela; Doussineau, Mathieu (2025) : When cohesion meets excellence: Analysing the drivers of synergies between EU R&I funding instruments in EU regions, JRC Working Papers on Territorial Modelling and Analysis, No. 05/2025, European Commission, Joint Research Centre (JRC), Seville This Version is available at: https://hdl.handle.net/10419/322073 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ When cohesion meets excellence: analysing the drivers of synergies between EU R&I funding instruments in EU regions No 05/2025 2025 Authors: Bachtrögler -Unger, J. Gabelberger, F. Marques Santos, A. Doussineau , M. This publication is a working paper by the Joint Research Centre (JRC), the European Commission’s science and knowledge service. It aims to provide evidence-based scientific support to the European policymaking process. The contents of this publication do not necessarily reflect the position or opinion of the European Commission. Neither the European Commission nor any person acting on behalf of the Commission is responsible for the use that might be made of this publication. For information on the methodology and quality underlying the data used in this publication for which the source is neither Eurostat nor other Commission services, users should contact the referenced source. The designations employed and the presentation of material on the maps do not imply the expression of any opinion whatsoever on the part of the European Union concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. The JRC Working Papers on Territorial Modelling and Analysis are published under the supervision of Simone Salotti, Andrea Conte, and Anabela M. Santos of JRC Seville, European Commission. This series mainly addresses the economic analysis related to the regional and territorial policies carried out in the European Union. The Working Papers of the series are mainly targeted to policy analysts and to the academic community and are to be considered as early-stage scientific papers containing relevant policy implications. They are meant to communicate to a broad audience preliminary research findings and to generate a debate and attract feedback for further improvements. Contact information Name: Anabela M. Santos Address: Edificio Expo, C/Inca Garcilaso 3, 41092 Sevilla (Spain) Email: [email protected] Tel.: +34 95 448 71 61 EU Science Hub https://joint-research-centre.ec.europa.eu JRC141964 Seville: European Commission, 2025 © European Union, 2025 The reuse policy of the European Commission documents is implemented by the Commission Decision 2011/833/EU of 12 December 2011 on the reuse of Commission documents (OJ L 330, 14.12.2011, p. 39). Unless otherwise noted, the reuse of this document is authorised under the Creative Commons Attribution 4.0 International (CC BY 4.0) licence (https://creativecommons.org/licenses/by/4.0/). This means that reuse is allowed provided appropriate credit is given and any changes are indicated. How to cite this report: Bachtrögler-Unger, J.; Gabelberger, F.; Marques Santos, A.; Doussineau, M., When cohesion meets excellence: analysing the drivers of synergies between EU R&I funding instruments in EU regions, European Commission, Seville, 2025, JRC141964. 1 Contents Abstract ...................................................................................................................................................................................................... 2 Acknowledgements .............................................................................................................................................................................. 3 Executive summary .............................................................................................................................................................................. 4 1. Introduction ....................................................................................................................................................................................... 6 2. Literature review ............................................................................................................................................................................ 8 2.1. Theorical foundations and policy coherence ........................................................................................................ 8 2.2. The importance of synergies and their potential effects ............................................................................. 8 3. Data and Methods ...................................................................................................................................................................... 11 3.1. Project data as a tool to analyze synergies ...................................................................................................... 11 3.2. Measuring synergies using the cosine similarity index ............................................................................... 16 3.3. Regression analysis considering NUTS-3 socio-economic characteristics ....................................... 18 4. Results ............................................................................................................................................................................................... 20 4.1. Stylized facts on the distribution of ERDF R&I and Horizon 2020 funding across SGCs ........ 20 4.2. The (dis)similarity of the distribution of EU R&I funding among SGCs ............................................. 22 4.3. Socio-economic characteristics as determinants of potential funding synergies ....................... 25 5. Conclusions ..................................................................................................................................................................................... 32 6. Discussion........................................................................................................................................................................................ 33 References ............................................................................................................................................................................................. 34 Annexes ................................................................................................................................................................................................... 38 Annex A: Distribution of Horizon 2020 and ERDF R&I funding among NUTS-3 regions .................... 38 Annex B. Descriptive statistics and robustness checks ......................................................................................... 42 List of figures ....................................................................................................................................................................................... 53 List of tables ......................................................................................................................................................................................... 54 2 Abstract This paper introduces a novel approach to measuring synergies between Horizon 2020 and cohesion policy funding in the field of R&I during the 2014-2020 programming period. Leveraging projectlevel data, we calculate regional cosine similarity indices based on societal grand challenges (SGCs) addressed to assess alignment between the two EU funding instruments in EU NUTS-3 regions. Results indicate that synergies are less likely in rural areas, emerging innovators, and – though not statistically significant – less developed regions, highlighting the role of business environments and innovation ecosystems. Regression analysis reveals that funding alignment is positively linked to the presence of universities and specialization in knowledge-intensive services, though the latter exhibits a nonlinear effect under certain circumstances. However, a higher number of SGCs addressed in smart specialization (S3) policy objectives is negatively associated with thematic funding similarity, likely due to fragmentation and dilution of focus. Regions that prioritize too many SGCs may reduce their ability to develop strong specialization and align different funding sources effectively. 3 Acknowledgements We are thankful for helpful comments by Peter Huber, an anonymous reviewer at JRC Seville, Dimitri Corpakis, Arnault Morrison and Francesco Cappellano. Authors Julia Bachtrögler-Unger* (corresponding author) Austrian Institute of Economic Research (WIFO), Vienna, Austria julia.bachtroegler-[email protected] Fabian Gabelberger* Austrian Institute of Economic Research (WIFO), Vienna, Austria Fabian.gabelberg[email protected] Anabela M. Santos European Commission, Joint Research Centre, Seville, Spain [email protected] Mathieu Doussineau European Future Innovation System Centre, Brussels, Belgium [email protected] * The researchers at WIFO (Julia Bachtrögler-Unger and Fabian Gabelberger) were supported by funds of the Austrian National Bank (OeNB Jubiläumsfonds project no. 18463). Keywords: Synergy; R&I funding; Cohesion policy; Horizon 2020; EU regions JEL Codes: O31; O52; R58. 4 Executive summary Synergies in funding refer to the strategic alignment and coordinated use of financial resources to achieve a greater impact than individual programs operating separately. In the European Union (EU), funding synergies have gained prominence as a way to enhance the effectiveness of research and innovation (R&I) investments, particularly through cohesion policy and Horizon 2020 programs. By leveraging synergies, the EU aims to maximize the benefits of funding instruments and accelerate progress in addressing societal challenges. The study examines the extent to which cohesion policy (under the European Regional Development Fund – ERDF) and Horizon 2020 funding were used synergistically in the 2014–2020 programming period to address societal grand challenges (SGCs). The analysis explores: — The similarity in funding distribution across EU regions. — Regional differences based on factors such as agglomeration density, development level, and innovation performance. — Key regional socio-economic characteristics influencing funding alignment. Using a novel data-driven approach, the study employs text mining and latent semantic analysis to classify projects based on their thematic focus. A cosine similarity index measures the degree of alignment between ERDF R&I and Horizon 2020 funding at the NUTS-3 regional level. Regression analysis then identifies the factors influencing the intensity of funding synergies. Key Findings — Synergies between Horizon 2020 and ERDF R&I funding are stronger in regions with welldeveloped innovation ecosystems, particularly those with a high concentration of universities and knowledge-intensive services. — Rural areas, emerging innovators, and less developed regions show weaker funding alignment, indicating potential barriers to synergy, such as fragmented business environments and weaker institutional collaboration. — Regions that include a high number of SGCs in their Smart Specialisation Strategy (S3) tend to exhibit lower funding alignment, suggesting that an overly broad policy focus may dilute specialization and hinder effective synergy. — Population density plays a significant role, with highly urbanized regions demonstrating stronger funding complementarities. Policy Implications To enhance synergies between EU funding instruments, policymakers should: — Streamline legal frameworks and institutional coordination to facilitate collaboration between funding programs. 5 — Strengthen regional innovation ecosystems by investing in universities and knowledge-intensive industries. — Refine S3 priorities to encourage more targeted specializations, preventing the dilution of funding across too many policy areas. — Support tailored funding strategies for rural and less developed regions to bridge the gap in funding synergies. The findings contribute to ongoing policy discussions on optimizing the alignment of EU R&I funding. Future research should further investigate how evolving EU policy frameworks beyond 2020 can build upon these insights to reinforce synergies and enhance the EU’s global competitiveness. 6 1. Introduction The Cambridge dictionary defines synergy as “the combined power of a group (…) when they are working together that is greater than the total power achieved by each working separately” (Combley, 2011:837). The concept is therefore also linked to the benefits and savings of working together to achieve a common goal. In the field of policy evaluation, the academic literature (see, e.g., Van den Bergh et al., 2021) considers that positive synergy is achieved when the combined effect of more than one policy instrument is greater than the sum of their individual effects. Synergies can increase the effectiveness of a policy, but also potentially reduce negative effects and accelerate the achievement of results (Chou, 2018). The European Commission (2022) defines funding synergy as the strategic alignment and combined use of financial resources to achieve more significant impact and efficiency than would be possible if each program operated independently. This involves coordinating efforts across different funding sources, such as the Horizon Framework Programme and the European Regional Development Fund (ERDF), or more broadly under the cohesion policy funds, to support common research, innovation and regional development objectives and thereby maximize their collective benefits. The policy mix, in terms of consistency, coherence, integration and coordination between the different financing instruments, is a key element to achieve synergies (Marques Santos, 2021). In this respect, the OECD (2003) distinguishes policy coherence from policy coordination (getting systems to work together) and policy consistency (avoiding contradictory policies). From a policy coherence perspective, the concept of synergies between EU funding requires more than political intent. Policy coherence is about ensuring that policies are mutually reinforcing to create synergies towards a defined goal (Reid et al., 2007). In the European Union (EU) context, the concept of synergies gained a new dimension with the emergence of the Smart Specialization Strategy (S3) concept in 2009. S3 is a place-based innovation policy that helps regions identify and develop competitive advantages by focusing on their strengths and potential (Foray et al., 2009). The aim is to take the heterogeneity of regions seriously and strengthen economic development by investing in the advancement of technologies that are related to the skills and capabilities prevalent in each region (e.g., Barbero et al., 2024). Since 2014, S3 is a key requirement for cohesion policy funding targeted to R&I through the ERDF in view of ensuring efficient resource allocation and synergies between policies (European Commission, 2011). The European Commission (EC) explicitly encourages the implementation of synergies between cohesion policy and Horizon 2020 (European Commission, 2014) and Horizon Europe (European Commission, 2022). Nevertheless, according to several studies (see, e.g., ECA, 2022; RIMA, 2025; Segal et al., 2025) the full potential of synergies between funding instruments has not been achieved yet. Among the main barriers to achieve synergies are the legal framework and the lack of coordination between the different institutions implementing the programs (Segal et al., 2025). The current political context underscores the urgency of addressing these challenges, as the EU strives to enhance its strategic autonomy and global competitiveness. In this regard, the Draghi (2024) report highlights the critical need to strengthen synergies between EU funding instruments, advocating for a more coherent and coordinated approach to policy implementation. By streamlining legal frameworks and improving institutional collaboration, the report highlights that the EU can optimize resource allocation, foster innovation, and reinforce its economic resilience in an increasingly competitive global landscape. 13 constructed based on the RISIS KNOWMAK ontology (Maynard et al., 2019)8 and supplemented with keywords extracted from Horizon 2020 projects, which are already classified according to SGCs. The latter step ensures the alignment of the categories applied to projects from both funding sources. In a next step, the list of keywords was refined in a trial-and-error approach by checking and cleaning for keywords found to be too general.9 The number of keywords per SGC ranges from around 370 for Security, over 500 for Climate and Society to over 700 for Transport and Health. If an ERDF project is found to relate to more than one SGC, its full financial amount is assigned to each of those categories. ERDF projects that cannot be assigned to at least one societal grand challenge (SGC) using the keyword string search approach are excluded from the following analysis. This yields a sample of 148,000 ERDF R&I projects with total eligible expenditures of 58 billion EUR, representing two thirds of all ERDF R&I projects or, equivalently, 76% of the sum of total eligible expenditures. Table 1 shows the distribution of funding across SGCs for the sample to be analyzed. The share of Horizon 2020 funding allocated to Transport, Energy and Health is notably larger than the respective share of ERDF R&I funding. By contrast, the share of ERDF R&I funding allocated to Climate projects is larger than the one among Horizon 2020 funding. What is striking is the prominent focus of ERDF R&I projects on the SGC Society, which might partly reflect a relatively broad definition of this SGC as compared to other SGCs. However, the number of keywords representing Society is among the lowest and there are substantial cross-country differences: The ERDF R&I funding share for Society ranges from 9.3% in Austria to 45.4% in Sweden; in Lithuania the maximum share of 13.7% of Horizon 2020 funding addresses this SGC (see Table A.1 in the annex). 8 See http://www.knowmak.eu and https://gate.ac.uk/projects/knowmak/ - Final version of the KNOWMAK ontology version 1 for the list of keywords [downloaded 28 July 2021]. 9 A limitation of this method lies in the dependence on the quality and degree of detail of ERDF project names and descriptions as well as potentially linguistic aspects. On the one hand, project names and descriptions in the Kohesio database are translated to English from original lists of operations often provided in national languages, which might trigger the usage of certain words in the texts. On the other hand, certain (key)words might be more often used in certain languages than in others. Synonyms as well as wildcards in the list of keywords applied should in parts tackle that issue. 14 Table 1. Share of ERDF R&I (assigned to at least one SGC) and Horizon 2020 funding by SGC Category ERDF R&I Horizon 2020* Biotechnology 12% 13% Energy 15% 19% Security 5% 6% Transport 15% 24% Climate 15% 12% Health 16% 22% Society 22% 4% Basis: Total eligible expenditure of ERDF projects assigned to at least one SGC (counted more than once if assigned to more than one SGC) 111.2 bn EUR (without multiple counting: 58 bn EUR) Basis: Total H2020 grant 21.4 bn EUR Source: CORDIS database, Kohesio database, own elaborations. Note: ERDF R&I represents the total eligible expenditure assigned to R&I projects co-funded by the ERDF and assigned to at least one SGC. Amounts are, e.g., double-counted if the project is assigned two SGCs. 47% of the 148,000 ERDF R&I projects are assigned only one SGC, 37% of them are assigned two SGCs. 10% of ERDF R&I projects in the sample correspond to 3 SGCs, 4% to 4 SGCs, 2% to 5 and more. Horizon 2020 contains the respective funding amount. *Shares do not add up to 100% because percentage shares are rounded. Distribution of Horizon 2020 and ERDF R&I funding across EU-27 NUTS-3 regions Figure 1 shows the distribution of Horizon 2020 and ERDF R&I funding across NUTS-3 regions in the EU-27. Regions coloured in grey are excluded from the subsequent analysis: First, ERDF project data for Greece, Ireland and Malta is only available for NUTS-2 (not NUTS-3) regions. Second, regions are excluded from the analysis of synergies if they do not receive funding for R&I projects from both instruments. For example, ERDF R&I funding is reported for 72 of the 73 Polish NUTS-3 regions, whereas 25 Polish regions did not attract any Horizon 2020 funding. Thus, the subset of 47 Polish NUTS-3 regions that have received both ERDF R&I and Horizon 2020 funding is considered in this analysis. In sum, this results in 806 out of 1,166 NUTS-3 regions being covered in the analysis. See Table A.2 in the annex for an overview of the number of NUTS-3 regions that did not receive Horizon 2020 and/or ERDF R&I funding by Member State. While the right-hand side figure (1b) mirrors the focus of ERDF funding on Eastern and Southern European Member States also when focusing on cohesion policy-funded R&I activities, Figure (1a) shows that comparatively large amounts of Horizon 2020 funding is attracted to Scandinavian and other Northern European regions (such as in Denmark or the Netherlands) as well as capital regions (such as Madrid or Île de France). Furthermore, Figure 1 points to the fact that the regional funding amount per capita in the 806 NUTS-3 regions attracting both funds is remarkably larger for the ERDF (mean 167 EUR per capita, median 66 EUR per capita) than for Horizon 2020 (mean 34 EUR per capita, median 10 EUR per capita). 10 10 Note that we consider funding in the 2014-2020 programming period as a whole and not year by year in order to capture the overall similarity between the two funding programs. The programs run at different speed (of 15 Figure 1. Distribution of Horizon 2020 and ERDF R&I amounts per capita among NUTS-3 regions (1a) Horizon 2020 funding per capita in € (1b) Total eligible expenditure of ERDF R&I projects (2014-2020) per capita in € Source: CORDIS database, Kohesio database, own elaborations. Note: ERDF project data for Greece, Ireland and Malta is not provided at the NUTS-3 regional level, therefore these regions are excluded from the analysis. The 2021 version of the Nomenclature of Territorial Units for Statistics (NUTS) is used. An analysis of the synergies between the two funding instruments is expected to be more meaningful if extreme cases are excluded. E.g., the NUTS-3 region of Paris is among the top receivers of Horizon 2020 funding per capita and at the same time part of the bottom decile when it comes to ERDF R&I funding per capita received. Overall, 20% of the NUTS-3 regions in the sample (158 of 806) receive more Horizon 2020 than ERDF R&I funding. There are significant differences in the Horizon 2020/ERDF R&I ratio across – in declining order - predominantly urban, intermediate and rural areas (see Table A.3 in the annex). Differences are even more pronounced when comparing NUTS-3 regions part of more developed and less developed regions (Figure 2) as well as across NUTS-2 regions classified according to the regional innovation scoreboard (Figure A.1). absorption) at EU, national and regional level to have a harmonized indicator that is not biased by program characteristics and their implementation (Santos et al., 2025). 16 Figure 2. Horizon 2020/ERDF R&I ratio in different groups of NUTS-3 regions based on their economic development Source: CORDIS database, Kohesio database, own elaborations. Note: The figure shows the relationship between Horizon 2020 and ERDF R&I funding in 806 NUTS-3 regions that receive funding from both instruments. Due to data restrictions, Greece, Ireland and Malta are not considered in the analysis. 3.2. Measuring synergies using the cosine similarity index To quantify the degree of alignment between the thematic priorities of ERDF R&I investments and Horizon 2020 project activities across NUTS-3 regions, we employ the cosine similarity. Cosine similarity measures the similarity between two vectors in terms of the difference between their angles and is defined as: 1 bn 2 bn 3 bn 4 bn 5 bn 17 cos(𝜃𝜃)= 𝐀𝐀∙𝐁𝐁 ‖𝐀𝐀‖‖ 𝐁𝐁‖= ∑𝐴𝐴 𝑖𝑖 𝐵𝐵 𝑖𝑖 𝑛𝑛 𝑖𝑖=1 �∑𝐴𝐴𝑖𝑖 2 𝑛𝑛 𝑖𝑖=1 ∑𝐵𝐵𝑖𝑖2 𝑛𝑛 𝑖𝑖=1 (1) where A and B are the vectors representing the ERDF R&I and Horizon 2020 funding amounts per SGC for a single region and 𝑖𝑖 iterates over the seven SGCs. In contrast to the Euclidean distance, the cosine similarity is always normalized between -1 and 1 since the absolute magnitude of the vectors does not play a role in their angle. Note that in our case, there are no negative values due to the strictly positive funding amounts allocated to NUTS-3 regions (net values). This means the similarity measure will always be between 0 and 1. This metric is advantageous in situations where the primary interest lies in the proportional alignment rather than in absolute volumes. Compared to correlation coefficients (such as Pearson’s), which are primarily designed to detect linear relationships in magnitude, cosine similarity compares the direction of two vectors in a multidimensional thematic space, irrespective of scale. Meaning that correlation coefficients are sensitive to co-variation in magnitude and low if there is no linear relationship present across SGCs, while cosine similarity is high if the SGCs are prioritized similarly, even if the ERDF and Horizon 2020 vectors differ in magnitude or skewness.11Similarity in this context means the similarity of the distribution of ERDF and Horizon 2020 funding attracted or allocated to a NUTS-3 region in terms of the shares attributed to different SGCs. As financial variable of interest, the total eligible expenditure allocated to ERDF R&I projects, including national co-funding, is used. In the case of Horizon 2020 projects, which do not require national co-funding, the respective net EU contribution is considered. The cosine similarity index calculated for a NUTS-3 region has a value of 1 if respective ERDF R&I and Horizon 2020 funding amounts are distributed exactly the same across SGCs. A value of 0 indicates no overlap in the directionality of funding at all, i.e., orthogonal vectors. The cosine similarity index cannot only be used to contrast funding patterns in NUTS-3 regions, but also allows to compare the funding distribution of ERDF and Horizon 2020 in groups of NUTS-3 regions. Therefore, it is calculated for i) all (predominantly) urban, rural or intermediate regions, ii) NUTS-3 regions part of NUTS-2 regions of different economic development level (less developed, transition and more developed regions in the 2014-2020 programming period), and iii) NUTS-3 regions located in NUTS-2 regions with a different innovation performance according to the regional innovation scoreboard (European Commission, 2021). As further metric of interest, we calculate the Gini index as a measure of concentration of EU R&I funding on specific SGCs using the following equation: 𝐺𝐺𝑖𝑖𝐺𝐺𝑖𝑖 = ∑ ∑ �𝑥𝑥𝑖𝑖−𝑥𝑥𝑗𝑗� 𝑛𝑛 𝑗𝑗=1 𝑛𝑛 𝑖𝑖=1 2𝑛𝑛2𝑥𝑥 (2) 11 Compared to using location quotients, or the concept of revealed comparative advantage, to assess synergies (such as in Doussineau and Bachtrögler, 2021), the advantage of the cosine similarity index lies not only in the independence of its interpretation from absolute amounts of funding, but also in no need of choosing a benchmark value and group of regions according to which a specialisation of funding is identified. Also, location quotients would be calculated for each SGC separately and additional assumptions would be necessary to build an aggregate measure of synergies at the NUTS-3 regional level. 18 where x is the vector of the total funding amounts (ERDF R&I or Horizon 2020) per SGC for a single NUTS-3 region and i and j iterate over the seven SGCs. The resulting Gini coefficient is then normalized to values between 0 and 1. 3.3. Regression analysis considering NUTS-3 socio-economic characteristics To explore potential determinants of the synergetic use of EU R&I funding in NUTS-3 regions, a standard OLS regression analysis is employed. Equation (3) depicts the model to be estimated, 𝑦𝑦𝑖𝑖,𝑟𝑟=𝛽𝛽0+𝑋𝑋𝑖𝑖,𝑟𝑟𝛽𝛽+ 𝛾𝛾𝑟𝑟+𝜀𝜀𝑖𝑖,𝑟𝑟 (3) where yi,r is the cosine similarity index for NUTS-3 region i in NUTS-2 region r, Xi,r is a 1 × k vector of explanatory variables and β the corresponding vector of coefficients. β0 denotes the intercept and εi,r is the error term. As we are particularly interested in the relationship between the alignment of ERDF R&I and Horizon 2020 funding and structural characteristics at within NUTS-2 regions, regression results are provided (without and) considering NUTS-2 regional fixed effects (γr) in order to account for unobserved differences across NUTS-2 regions. Given that the dependent variable is bounded between 0 and 1, a (quasi-likelihood) fractional logistic regression approach is applied as an alternative specification to check for the robustness of the results. Furthermore, regression results are provided for a ‘trimmed dataset’ in which we exclude NUTS-3 regions with the largest discrepancies between Horizon 2020 and ERDF R&I funding received to ensure the robustness of findings. The set of explanatory variables consists of a set of socio-economic characteristics at the NUTS-3 regional level that are expected to matter for the usage of EU R&I funding (see Table 2), following the determinants of the geography of innovation (see e.g. Porter, 1998; Cicerone et al., 2023; Marques Santos et al., 2025) and the agglomeration economy (Krugman, 1991; Ottaviano and Puga, 1998; Barbero et al., 2025). In general, pre-programming period values of 2013 are considered to avoid reverse causality bias of funds’ effects on our explanatory variables. The number of SGCs mapped against policy objectives set as S3 priorities for the 2014-2020 programming period is used as an indicator that is expected to shape the thematic orientation of R&I funding. As S3 priorities are mostly defined at the NUTS-2, partly at the NUTS-1, and national level (only Finland and Sweden set priorities for NUTS-3 regions; the Czech S3 is national with regional annexes for particular NUTS-3 regions), they were considered as relevant for all subordinated NUTS-3 regions. For a more differentiated analysis, baseline results consider the number of SGCs addressed excluding national S3 priorities.12 Note that this data analysis revealed that S3 policy objectives are defined very broadly: 454 (or 56%) of NUTS-3 regions in our sample had all SGCs addressed as priorities. 12 Note that S3 priorities for EU Member States consisting of one NUTS-2 region (Cyprus, Estonia, Luxembourg, Latvia, Malta) were considered as set at the regional (not national) level. Data on S3 priorities was downloaded from the Eye@RIS3 database (https://s3platform.jrc.ec.europa.eu/map), adapted to fit the NUTS 2021 classification and assigned to SGCs (which is trivial in most cases). 19 Table 2. Explanatory variables at the NUTS-3 regional level: variable description and source Name Description Source Capital region Dummy variable equal to 1 if the respective country’s capital is located in the NUTS-3 region; 0 otherwise Own elaboration based on city location within NUTS-3 region. University13 Dummy variable equal to 1 if there is an university in the region; 0 otherwise Own elaboration based on data from The European Higher Education Sector Observatory (2022) LQ Knowledge-intensive services Concentration of knowledgeintensive services estimated using the location quotient (LQ), i.e. the share of gross value added (GVA) in NACE sectors J-N in region 𝑖𝑖 over the share of GVA in sectors J-N in the EU-27 Own elaboration based on ARDECO (SOVGZ) GDP per capita (ln) Real GDP per capita, EUR2015, by inhabitant, in logs (ln) Own estimation based on ARDECO (GDP - SOVGD; Population - SNPTD) Investment rate Ratio of Gross Fixed Capital Formation (GFCF) over Gross Domestic Product (GDP) Own estimation based on ARDECO (GFCF - ROIGT; GDP - SOVGD) Competition Competition level = 1 – Herfindahl-Hirschman-Index (10 NACE sectors) Own estimation based on ARDECO (SOVGZ) Population density (ln) Inhabitant per km 2 , in logs (ln) Own estimation based on ARDECO (Population - SNPTD) and EUROSTAT (area - reg_area3) No. of SGCs in S3 priorities (excl. national priorities) Number of SGCs mapped against policy objectives set as part of S3 priorities for the 2014-2020 programming period. Note that this variable is available at NUTS-3 level only for Sweden and Finland as well as certain Czech NUTS-3 regions. For the other Member States, S3 priorities are set at the NUTS-2, NUTS-1 and/or national level and considered as such for subordinated NUTS-3 regions. Eye@RIS3 database, own elaboration Source: own elaboration. Summary statistics are provided in Table B.1 in the annex. 13 For a robustness check, we consider the number of universities and the number of students (ISCED levels 5-7), respectively, per NUTS-3 region in 2013 (see Tables B.5 and B.6 in the annex). 20 4. Results 4.1. Stylized facts on the distribution of ERDF R&I and Horizon 2020 funding across SGCs Figure 3 shows the distribution of the EU’s two main R&I funding instruments in different groups of NUTS-3 – and corresponding NUTS-2 – regions. As already discussed in the Data section, ERDF projects are more strongly focused on the SGC Society, whereas Horizon 2020 projects are in general more targeted at Transport and Energy. Comparing especially Horizon 2020 funding shares across groups of regions, however, reveals some patterns: The funding shares of Horizon 2020 for Transport and Health decline with the level of development as well as with innovation performance. By contrast, the share allocated to Biotechnology is highest for NUTS-3 regions located in transition and emerging innovator regions, respectively. For ERDF funding (in the field of R&I), systematic differences across groups of regions appear to be less pronounced. All in all, these stylized facts motivate an analysis of socio-economic indicators that contribute to the extent of alignment in the thematic orientation of the two funding schemes at NUTS-3 level. Figure 3. Funding shares allocated to different SGCs in NUTS-3 regions part of … Continued on the next page … 21 Source: CORDIS database, Kohesio database, own elaborations. Note: First classification based on the economic development level of corresponding NUTS-2 regions in the 2014-2020 programming period. Second classification of corresponding NUTS-2 regions based on regional innovation scoreboard (European Commission 2021). The S3 concept motivates regions to focus their funding on specific strategic areas, such as policy objectives. Therefore, the Gini index, ranging from one (only one SGC addressed) to zero (equal distribution of funding across the seven SGCs), is calculated to assess concentration of ERDF R&I and Horizon 2020 funding on a number of SGCs in NUTS-3 regions. The Gini index indicates high variation in the concentration of regional EU R&I funding among NUTS3 regions and as well across funding instruments (see Table A.4). Horizon 2020 funding is, on average, much more focused on certain SGCs than ERDF R&I funding (the mean of the Gini for ERDF projects is 0.54, for Horizon 2020 projects 0.77). While this is partly driven by design (ERDF projects can be attributed to more than one SGC, see notes of Table 1), the higher concentration of Horizon 2020 funding is not surprising given its excellence-based allocation mechanism. Another contributing factor may be that already S3 priorities in smart specialization strategies that govern ERDF allocations in the field of R&I, i.e., policy objectives, are chosen very broadly: More than half of the NUTS-3 regions in the sample chose S3 policy objectives that correspond to all the seven SGCs. In this respect, recent literature finds for technological domains that actual R&I funding is more selective (in terms of the relatedness to regional capabilities) than S3 priorities set in smart specialization strategies (Hellinga et al., 2025). This motivates to check whether the number of SGCs prioritized through S3 priorities contributed to the (dis)similarity in the distribution of EU R&I funding in (different groups of) NUTS-3 regions. Interestingly, predominantly rural NUTS-3 regions perform better in channeling both their ERDF R&I and Horizon 2020 funding to specific SGCs than intermediate and even more than urban areas (see 22 Table A.4), which will partially be due to a broader variety of types of organizations (companies, research institutes, universities, etc.) and expertise of actors located in more densely populated regions. A similar pattern is found for Horizon 2020 funding in groups of NUTS-3 regions by development level and innovation performance of corresponding NUTS-2 regions: concentration in less developed and transition regions is higher than in more developed ones, and concentration in regions classified as emerging innovators is significantly higher than in innovation leaders and strong innovators. However, the differences in mean values across groups of regions are relatively small (Table A.4). By contrast, concentration of ERDF R&I funding on certain SGCs is significantly higher in more developed and transition regions as compared to less developed ones. Also, the degree of concentration of increases remarkably (also in terms of mean values) with improving innovation performance. 4.2. The (dis)similarity of the distribution of EU R&I funding among SGCs To explore the synergetic use of the two main EU R&I funding instruments in terms of addressing the same societal challenges, the cosine similarity index measuring the alignment of Horizon 2020 and ERDF R&I funding is calculated for the level of NUTS-3 regions. This index amounts to one if the two funding instruments are allocated or have been attracted -to address the same SGCs in the same proportions. A value of zero indicates no overlap at all. The cosine similarity index varies remarkably across the 806 European NUTS-3 regions that received EU R&I funding from both ERDF and Horizon 2020 (see Figure 4). The average cosine similarity index amounts to 0.51 (the median is 0.55, see Table 3). 46 of the 806 NUTS-3 regions do not orient their EU R&I funding towards the same SGCs at all (most of them are part of more developed NUTS-2 regions), while the distribution of EU R&I funding across SGCs is fully aligned in two rural NUTS-3 regions in Germany.14 Table 3 provides summary statistics on the differences of the similarity of the distribution of ERDF R&I and Horizon 2020 funding across groups of NUTS-3 regions. First, the alignment of the two funding instruments in terms of SGCs addressed is found to be significantly higher in predominantly urban regions than in intermediate and, even more, rural regions. In the previous section we discussed the stylized fact that concentration of both funding instruments on certain SGCs is relatively low in urban areas, which is likely to be explained by the fact that the number and variety of beneficiaries and their expertise is higher in more densely populated regions. Therefore, the finding that both funding streams are nevertheless used in a more similar fashion in urban areas, indeed points to synergies in the attraction of R&I project grants. This could be supported by a good innovation ecosystem and network, potentially centered around a university, research institutes or innovative companies. 14 If NUTS-3 regions that receive most (10th decile) and lowest (1st decile) of ERDF and Horizon 2020 funding, respectively, are excluded from the sample (518 NUTS-3 regions left), the average cosine similarity amounts to 0.53 (median 0.56). In this case, the cosine similarity index amounts to 0 only in 20 NUTS-3 regions; the two regions with the totally similar funding distribution remain in the sample. 29 Table 6. Correspondence between the cosine similarity index and socio-economic characteristics in NUTS-3 regions by innovation group OLS regression by innovation group (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) VARIABLES Emerging innovator Moderate innovator Strong innovator Innovation leader Capital region -0.212 -0.196 -0.311 0.159 0.120 0.085 -0.220* -0.197 -0.140 -0.008 0.763*** 0.475** (0.315) (0.389) (0.364) (0.255) (0.228) (0.235) (0.122) (0.223) (0.205) (0.134) (0.200) (0.178) University 0.148** 0.187* 0.192** 0.115** 0.103** 0.106** 0.075* 0.074 0.096** 0.045 0.147* 0.160* (0.064) (0.098) (0.090) (0.046) (0.050) (0.048) (0.044) (0.049) (0.044) (0.071) (0.080) (0.081) LQ knowledge-intensive -0.069 -0.312 -0.132 0.781 0.983* 0.705 0.554 0.742 0.628 0.765*** 0.937** 0.661 services (KIS) (0.421) (0.584) (0.499) (0.519) (0.579) (0.576) (0.399) (0.535) (0.508) (0.237) (0.414) (0.420) LQ KIS squared 0.375 0.357 0.309 -0.399 -0.526 -0.333 -0.282 -0.414 -0.310 -0.202*** -0.256* -0.116 (0.354) (0.437) (0.395) (0.314) (0.343) (0.330) (0.172) (0.272) (0.250) (0.064) (0.142) (0.138) ln_GDP per capita 0.045 0.199 0.163 0.187 0.172 0.217 0.540*** 0.454** (0.200) (0.258) (0.116) (0.133) (0.134) (0.154) (0.136) (0.221) Investment rate 1.186 4.079 5.088*** 5.384*** 1.016 1.860 1.367 2.996 (1.398) (2.872) (1.941) (1.989) (2.838) (3.077) (2.270) (2.979) Competition level -0.571 -0.063 -1.201 0.460 0.590 0.546 -0.475 -0.212 -0.747 0.332 -0.807 -1.665* (1.169) (1.377) (1.043) (0.880) (0.944) (0.827) (1.142) (1.058) (1.002) (0.776) (1.092) (0.871) ln_Population density -0.002 0.048 0.046 0.024 0.013 0.037 0.060* 0.060* 0.074*** -0.118*** -0.202*** -0.123** (0.044) (0.055) (0.054) (0.030) (0.033) (0.031) (0.031) (0.031) (0.028) (0.042) (0.054) (0.051) No. of SGC in S3 priorities -0.037 -0.102* -0.056 0.040 -0.016 -0.011 -0.048 -0.056 -0.053 -0.079*** -0.062** -0.068** (excl. national S3 priorities) (0.038) (0.057) (0.047) (0.046) (0.015) (0.015) (0.046) (0.049) (0.047) (0.026) (0.028) (0.027) Observations 169 114 114 236 214 214 274 242 242 127 75 75 NUTS-2 FE YES YES YES YES YES YES YES YES YES YES YES YES TRIMMED dataset NO YES YES NO YES YES NO YES YES NO YES YES R-squared 0.592 0.710 0.692 0.492 0.495 0.470 0.378 0.407 0.398 0.482 0.575 0.516 Adjusted R-squared 0.328 0.414 0.399 0.263 0.237 0.211 0.160 0.169 0.166 0.258 0.301 0.239 VIF 7.50 11.91 8.60 19.37 21.20 19.95 6.47 6.83 5.29 9.61 8.54 7.49 VIF > 10 (excl. squared terms) S3 prios, ln_GDP_pc, Invrate, LQ_KIS S3 prios, ln_GDP_pc, Invrate, LQ_KIS, ln_PopDen s S3 prios, ln_PopDen s (10.6) S3 prios, InvRate, ln_GDP_pc S3 prios, InvRate, ln_GDP_pc S3 prios S3 prios InvRate S3 prios InvRate S3 prios - - - Notes: The dependent variable is the cosine similarity index. Significance level: *** p < 0.01, ** p<0.05, * p<0.1. Heteroskedasticity-robust standard errors (clustered at the NUTS-3 regional level) are in parentheses. The constant as well as the specification excl. the no. of SGC in S3 priorities is not reported here. For the latter, as in baseline results, coefficients of the other variables do only change marginally. As discussed in previous sections, differences in the cosine similarity index across groups of regions seem to be linked to the concentration of funding. As it would induce endogeneity issues, it is not possible to include the Gini coefficient as a measure for concentration in the regressions, though. Therefore, we run a seemingly unrelated regression (SUR) model for all observations as well as subsets of the data, with the cosine similarity index and the Gini indices measuring the concentration of ERDF R&I funding and Horizon 2020 funding, respectively, as dependent variables that are jointly determined by regional characteristics. Table 7 presents the results of the SUR model. What strikes out is that the statistically significant positive relationship between the similarity of EU R&I funds usage and the presence of a university in the NUTS-3 region goes along with, at the same time, a significant negative relationship between university location and the concentration of both funding instruments on certain SGCs. At first sight, more alignment through a thematically broader funds usage appears not to mirror the idea of S3. However, given the different objectives and allocation principles of Horizon 2020 and ERDF funding, it seems more intuitive that NUTS-3 regions able to attract funding in more policy areas each face more potential for synergies. In the same fashion, when considering variables that were de-meaned at the NUTS-2 regional level (to consider NUTS-2 fixed effects in this setting), a higher specialization in knowledge-intensive services is found to be associated with less concentration of funding and, at the same time, a better alignment in the thematic orientation of the two funds. Results of the SUR model run for different groups of regions and robust across all specifications (Tables B.3 and B.4) indicate in line with OLS results in Table 5 that the specialization in knowledgeintensive services plays a role in NUTS-3 part of less developed regions, while the presence of a university matters in transition and more developed regions. Furthermore, a higher number of SGCs addressed via S3 priorities (policy objectives) is statistically significantly associated with less alignment of the funding instruments in more developed regions, when considering variables demeaned by NUTS-2 averages (in the specification without de-meaned variables, it is statistically significantly negatively linked with alignment in less developed regions). By innovation group, the result that the presence of a university corresponds to less concentration, but better thematic alignment holds across all groups (using the trimmed dataset). In the case of innovation leaders, the same result is found for a higher specialization in knowledge-intensive industries, even if there is a maximum specialization level up from which no better alignment is achieved. 31 Table 7. Seemingly unrelated regression (SUR) results All observations Trimmed dataset VARIABLES Cosine_sim Gini_Horizon Gini_ERDF Cosine_sim Gini_Horizon Gini_ERDF Capital region 0.124** -0.282*** -0.159*** 0.166** -0.286*** -0.151*** (0.059) (0.036) (0.045) (0.077) (0.047) (0.057) University 0.134*** -0.154*** -0.122*** 0.144*** -0.160*** -0.138*** (0.020) (0.012) (0.015) (0.022) (0.013) (0.016) LQ knowledge-intensive -0.024 -0.090 -0.162** 0.069 -0.077 -0.324*** services (KIS) (0.104) (0.063) (0.079) (0.145) (0.089) (0.108) LQ KIS squared -0.009 0.021 0.069** -0.058 0.012 0.125** (0.043) (0.026) (0.033) (0.069) (0.042) (0.051) ln_GDP per capita 0.025 -0.013 0.111*** 0.017 0.007 0.089*** (0.020) (0.012) (0.015) (0.024) (0.015) (0.018) Investment rate -0.432 0.063 1.033*** -0.397 0.153 1.058*** (0.267) (0.161) (0.203) (0.294) (0.181) (0.219) Competition level 0.536* -0.696*** 0.030 0.744** -0.745*** -0.128 (0.275) (0.166) (0.209) (0.337) (0.207) (0.251) ln_Population density 0.019* -0.015** 0.009 0.015 -0.016** 0.016* (0.010) (0.006) (0.007) (0.011) (0.007) (0.008) No. of SGC in S3 priorities -0.008 0.005 0.005 -0.011* 0.004 0.009** (0.005) (0.003) (0.004) (0.006) (0.003) (0.004) Observations 806 806 806 645 645 645 R-squared 0.104 0.376 0.193 0.111 0.338 0.216 De-meaned variables VARIABLES Cosine_sim Gini_Horizon Gini_ERDF Cosine_sim Gini_Horizon Gini_ERDF Capital region -0.023 0.023 0.041 0.048 0.021 0.154** (0.087) (0.048) (0.053) (0.115) (0.064) (0.068) University 0.094*** -0.079*** -0.087*** 0.111*** -0.085*** -0.095*** (0.020) (0.011) (0.012) (0.022) (0.012) (0.013) LQ knowledge-intensive 0.324*** -0.495*** -0.373*** 0.335** -0.469*** -0.601*** services (KIS) (0.123) (0.068) (0.074) (0.160) (0.089) (0.094) LQ KIS squared -0.087* 0.143*** 0.095*** -0.114 0.131*** 0.186*** (0.045) (0.025) (0.027) (0.069) (0.039) (0.041) ln_GDP per capita 0.212*** -0.118*** -0.089*** 0.188*** -0.125*** -0.085** (0.054) (0.030) (0.032) (0.065) (0.036) (0.038) Investment rate 2.357*** -0.857** -1.104** 3.184*** -1.172** -1.286** (0.779) (0.429) (0.469) (0.942) (0.526) (0.552) Competition level 0.111 -0.398** 0.233 0.178 -0.420* 0.439* (0.328) (0.181) (0.198) (0.387) (0.216) (0.227) ln_Population density 0.019 -0.048*** 0.007 0.025 -0.049*** 0.013 (0.014) (0.008) (0.008) (0.016) (0.009) (0.009) No. of SGC in S3 priorities -0.043 -0.031** 0.002 -0.049 -0.029* -0.010 (0.028) (0.016) (0.017) (0.031) (0.017) (0.018) Observations 806 806 806 645 645 645 R-squared 0.176 0.459 0.212 0.178 0.458 0.261 Notes: Significance level: *** p<0.01, ** p<0.05, * p<0.1. Standard errors in parentheses. The constant is not reported to save space. 32 5. Conclusions This study provides a novel quantitative approach to assessing synergies between Horizon 2020 and cohesion policy (ERDF) funding in the field of research and innovation (R&I) during the 20142020 programming period. By leveraging project-level data, we develop a cosine similarity index based on societal grand challenges (SGCs) to evaluate the thematic alignment between these two major EU funding instruments in EU NUTS-3 regions. Our findings highlight significant disparities in synergies across different regional typologies, with rural areas, emerging innovators, and less developed regions exhibiting lower alignment levels. The analysis underscores the crucial role of regional innovation ecosystems, particularly the presence of universities and specialization in knowledge-intensive services, in fostering stronger funding complementarities. The results also indicate that a higher number of SGCs addressed in Smart Specialization Strategy (S3) policy objectives is negatively linked to thematic funding similarity. This suggests that an overly broad policy focus may hinder effective alignment between Horizon 2020 and ERDF R&I investments. Moreover, the observed co-existence of higher similarity and lower concentration of funding schemes implies that Horizon 2020 and ERDF R&I projects often address different SGCs within a region, likely reflecting the diversity of actors involved in the local innovation landscape. Crucially, the study identifies robust patterns across regional innovation capacities. The presence of a university matters for thematic alignment of EU R&I funding in more developed and transition regions, as well as every group of NUTS-3 regions when distinguished by innovation performance of corresponding NUTS-2 regions. By contrast, specialization in knowledge-intensive services is found to play a role only for moderate innovators and innovation leaders, and in less (and more) developed regions, whereby excessive specialization appears to have the opposite effect. Furthermore, in less developed and transition regions, as well as in strong innovator regions, a higher population density is linked to a more synergetic use of the two funding instruments. These insights contribute to ongoing policy discussions on improving the strategic alignment of EU R&I funding instruments. The evidence suggests that refining S3 priorities to foster more targeted specializations and strengthening regional innovation ecosystems, particularly through universities and knowledge-intensive industries, can enhance synergetic funding use. Future research should further explore the mechanisms underpinning these relationships, particularly in light of evolving EU policy frameworks beyond 2020. 33 6. Discussion Several limitations of the analysis should be taken into account. The results are only valid for the sample considered, meaning that it should be considered that one third of ERDF R&I projects in the enriched Kohesio dataset could not be assigned to a SGC based on project names, descriptions, and the list of keywords used. Furthermore, limitations arising from the text mining approach and linguistic aspects (e.g., certain keywords used more often in some national languages or text translated to English) have to be kept in mind. This research does not evaluate the successful implementation of smart specialization policies in regions. In order to contribute to such an evaluation, a follow-up paper should analyze the individual SGCs addressed by ERDF R&I and Horizon 2020 projects before the background of S3 priorities defined in smart specialization strategies. In addition, it would also be necessary to study the outcomes of the projects, which requires a set of outcome indicators mirroring progress in addressing prevalent SGCs at the regional level. 34 References Almudi, I.; Fatas-Villafranca, F.; Fernández-Márquez, C.M; Potts, J.; Vazquez, F.J. (2020). “Absorptive capacity in a two-sector neo-Schumpeterian model: a new role for innovation policy”, Industrial and Corporate Change, 29(2):507–531. https://doi.org/10.1093/icc/dtz052. Bachtrögler-Unger, J., Arnold, E., Doussineau, M. and Reschenhofer, P. (2021a), UPDATE: Dataset of projects co-funded by the ERDF during the multi-annual financial framework 2014-2020, Publications Office of the European Union, Luxembourg, ISBN 978-92-76-43735-2, doi:10.2760/440159, JRC125008. Bachtrögler-Unger, J., Marques Santos, A. and Conte, A. (2021b), ERDF BENEFICIARIES DATASET 2014-2020: AN OVERVIEW FOR POLICY-MAKERS, European Commission, JRC127403. Barbero, J., Diukanova, O., Gianelle, C., Salotti, S., & Santoalha, A. (2024). “Technologically related diversification: One size does not fit all European regions”, Research Policy, 53(3), 104973. Borrás, S. and Schwaag Serger, S. (2022). “The design of transformative research and innovation policy instruments for grand challenges: The policy-nesting perspective”, Science and Public Policy, 49(5): 659–672. https://doi.org/10.1093/scipol/scac017. Cappellano, F.; Molica, F. and Makkonen, T. (2024). “Missions and Cohesion Policy: is there a match?”, Science and Public Policy, 51(3):360–374. https://doi.org/10.1093/scipol/scad076. Chou, T.C (2018). “The combination index (CI < 1) as the definition of synergism and of synergy claims”, Synergy 7: 49-50. https://doi.org/10.1016/j.synres.2018.04.001. Cicerone, G.; Faggian, A.; Montresor, S. and Rentocchini, F. (2023). “Regional artificial intelligence and the geography of environmental technologies: does local AI knowledge help regional green-tech specialization?”, Regional Studies 57, no. 2 (2023): 330-343. Cincera, M. and Santos, A. M. (2015). Innovation and Access to Finance – A Review of the Literature, Working Papers TIMES² 2015-16, ULB - Universite Libre de Bruxelles. Combley, R. (2011). Cambridge Business English Dictionary. Cambridge University Press. Crescenzi, R., de Blasio, G., & Giua, M. (2020). “Cohesion Policy incentives for collaborative industrial research: evaluation of a Smart Specialisation forerunner programme”, Regional Studies, 54(10): 1341–1353. https://doi.org/10.1080/00343404.2018.1502422. D’Adda, D., Iacobucci, D., & Perugini, F. (2021). “Smart Specialisation Strategy in practice: have regions changed the allocation of Structural Funds?”, Regional Studies, 56(1), 155–170. https://doi.org/10.1080/00343404.2021.1890326. Deegan, J., Broekel, T., & Fitjar, R. D. (2021). “Searching through the Haystack: The Relatedness and Complexity of Priorities in Smart Specialization Strategies”, Economic Geography, 97(5): 497– 520. https://doi.org/10.1080/00130095.2021.1967739. Deleidi, M. and Mazzucato, M. (2021). "Directed innovation policies and the supermultiplier: An empirical assessment of mission-oriented policies in the US economy", Research Policy, 50(2), 104151. https://doi.org/10.1016/j.respol.2020.104151 35 Di Cataldo, M., Monastiriotis, V., & Rodríguez‐Pose, A. (2021). “How ‘Smart’ are smart specialization strategies?”, Journal of Common Market Studies, 60(5): 1272–1298. https://doi.org/10.1111/jcms.13156. Doussineau, M. and Bachtrögler-Unger, J. (2021). Exploring Synergies between EU Cohesion Policy and Horizon 2020 Funding across European Regions: An analysis of regional funding concentration on key enabling technologies and societal grand challenges, Publications Office of the European Union. https://data.europa.eu/doi/10.2760/218779. Draghi, M. (2024). “The Draghi report: A competitiveness strategy for Europe”. Available here: https://commission.europa.eu/topics/strengthening-european-competitiveness/eu-competitivenesslooking-ahead_en. ECA (2022). Special Report 23/2022: Synergies between Horizon 2020 and European Structural and Investment Funds - Not yet used to full potential, European Court of Auditors. Available at: https://www.eca.europa.eu/en/publications/SR22_23 European Commission (2011). Ex-ante Conditionality in Cohesion Policy, Third Meeting of the Conditionality Task Force, 8 April 2011. European Commission (2014). Enabling synergies between European Structural and Investment Funds, Horizon 2020 and other research, innovation and competitiveness-related Union programmes. Available here: http:// ec.europa.eu/regional_policy/sources/docgener/guides/synergy/synergies_en.pdf. European Commission (2016). EU funds working together for jobs & growth – Examples of synergies between the framework programmes for research and innovation (Horizon 2020) and the European Structural and Investment Funds (ESIF), Directorate-General for Research and Innovation, Publications Office, 2016, https://data.europa.eu/doi/10.2777/678944. European Commission (2021), Regional innovation scoreboard 2021, Publications Office. https://data.europa.eu/doi/10.2873/674111. European Commission (2022). Communication to the Commission, Approval of the content of a draft Commission Notice on the synergies between ERDF programmes and Horizon Europe. Brussels, 5.7.2022, C(2022) 4747 final. Foray, D., David, P., and Hall, B. H. (2009). Smart Specialisation – The Concept. Knowledge Economists, Policy Brief Number 9, June. Brussels: European Commission, DG research. Fratesi U., Gianelle, C. & Guzzo, F. (2021). Assessing Smart Specialisation: Policy Implementation Measures, EUR 30758 EN, Publications Office of the European Union, Luxembourg, 2021, ISBN 97892-76-40034-9, https://doi.org/10.2760/71769, JRC123821. Gianelle, C., D. Kyriakou, C. Cohen and M. Przeor (2016). “Implementing Smart Specialisation: A Handbook”, Brussels: European Commission, EUR 28053 EN, doi:10.2791/53569. Gianelle, C., Guzzo, F., & Mieszkowski, K. (2020). “Smart Specialisation: what gets lost in translation from concept to practice?”, Regional Studies, 54(10): 1377–1388. https://doi.org/10.1080/00343404.2019.1607970. Hellinga, Z., Bachtrögler-Unger, J., Balland, P., & Boschma, R. (2025). Beyond the Blueprint. From Smart Specialization Strategies to R&I Funding, Papers in Evolutionary Economic Geography (PEEG) 2502, 36 Utrecht University, Department of Human Geography and Spatial Planning, Group Economic Geography, http://econ.geo.uu.nl/peeg/peeg2502.pdf. Huggins, R. and Thompson, P. (2015). “Entrepreneurship, innovation and regional growth: a network theory”, Small Business Economics, 45:103–128 (2015). https://doi.org/10.1007/s11187-015-96433. Kramer, J. P. et al. (2021). Study on prioritisation in Smart Specialisation Strategies in the EU. Report, European Commission, Luxembourg: Publications Office of the European Union. https://doi.org/10.2776/60867. Krugman, P. (1991). “Increasing returns and economic geography”, Journal of Political Economy, 99, no. 3: 483-499. Marques Santos, A. (2021). Linking the ‘Recovery and Resilience Plan’ and Smart Specialisation. The Portuguese case, JRC Working Papers on Territorial Modelling and Analysis, No. 05/2021, available at: https://publications.jrc.ec.europa.eu/repository/handle/JRC126178. Marques Santos, A.; Molica, F. and Torrecilla-Salinas, C. (2025). “EU-funded investment in Artificial Intelligence and regional specialization3, Regional Science Policy & Practice, 17(7). https://doi.org/10.1016/j.rspp.2025.100190. Marrocu, E., Paci, R., Rigby, D. L., & Usai, S. (2022). “Evaluating the implementation of Smart Specialisation policy”, Regional Studies, 57(1), 112–128. https://doi.org/10.1080/00343404.2022.2047915. Maynard, D., Petrak, J., Song, X. & Funk, A. (2019), KNOWMAK Report on Ontologies and Tagging, available at https://gate.ac.uk/projects/knowmak/D2.4-final.pdf. Mazzucato, M. (2018). “Mission-oriented innovation policies: challenges and opportunities”, Industrial and Corporate Change, 27(5):803–815. https://doi.org/10.1093/icc/dty034 McCann, P. and Soete, L. (2020). “Place-based innovation for sustainability”, Publications Office of the European Union, doi:10.2760/250023, JRC121271 Molica, F. and Marques Santos, A. (2024). “In search for the best match. Complementarities between R&I funds across EU regions”, Territorial Development Insights Series, JRC136780. OECD (2003). Policy coherence: Public governance and territorial development directorate. OECD (2023). Driving Policy Coherence for Sustainable Development: Accelerating Progress on the SDGs. OECD Publishing. Ottaviano, G.IP., and Puga, D. (1998). “Agglomeration in the global economy: a survey of the ‘new economic geography’”, The World Economy, 21, no. 6 (1998): 707-731. Porter, M. E. (1998). Clusters and the new economics of competition. Vol. 76, no. 6. Boston: Harvard Business Review. Potters, L. (2009). “R&D in Low-Tech sectors”, IPTS Working Paper on Corporate R&D and Innovation, 08/2009, European Commission. Reid, A., Miedzinski, M., Bruno, N., Le Gars, G., (2007) Synergies between the EU 7th Research Framework Programme, the Competitiveness and Innovation Framework Programme and the Structural 37 Funds, Technical report for the European Parliament's committee on Industry, Research and Energy (ITRE) IP/A/ITRE/ST/2006-16, DOI: 10.13140/RG.2.2.25650.20169 RIMA, Research & Innovation and Cohesion Managing Authorities Network (2025). Report on European Synergies of Funds, available at: https://horizoneuropencpportal.eu/repository/10d06752371b-47ff-89ab-54265cf2ef13. Santos, A. M.; Cincera, M. and Cerulli, G. (2024). “Sources of financing: Which ones are more effective in innovation–growth linkage?”, Economic Systems, 48(2), 101177, https://doi.org/10.1016/j.ecosys.2023.101177 Santos, A., Edwards, J., and Neto, P. (2022). “Does Smart Specialisation improve any innovation subsidy effect on regional productivity? The Portuguese case”, European Planning Studies, 31(4), 758–779. https://doi.org/10.1080/09654313.2022.2073787 Santos, A.M. and Conte, A. (2024). "Regional participation to Research and Innovation programmes under Next Generation EU: The Portuguese case", Papers in Regional Science, 103(1), 100006. https://doi.org/10.1016/j.pirs.2024.100006. Santos, A. M., Conte, A., & Molica, F. (2025). “Financial absorption of cohesion policy funds: how do programmes and territorial characteristics influence the pace of spending?”, JCMS: Journal of Common Market Studies, 63(1), 227-245. Segal L. J., Griepink M., Boschetti A., Marchioni M., Vilajosana Guillén X., Riera Duran M., and LupiáñezVillanueva F. (2025). “Exploring synergies between Horizon Europe and the EU Cohesion Policy”, Publication for the Committee on Industry, Research and Energy (ITRE), Policy Department for Transformation, Innovation and Health, European Parliament, Luxembourg. Sörvik, J. & Kleibrink, A. (2015). ‘Mapping Innovation Priorities and Specialisation Patterns in Europe’, JRC Technical Reports JRC95227. Van den Bergh, J.; Castro, J.; Drews, S.; Exadaktylos, F.; Foramitti, J.; Klein, F.; Konc, T. and Savin, I. (2021). “Designing an effective climate-policy mix: accounting for instrument synergy”, Climate Policy, 21(6):745-764, DOI: 10.1080/14693062.2021.1907276 Annexes Annex A: Distribution of Horizon 2020 and ERDF R&I funding among NUTS-3 regions Table A.1. Minimum and maximum shares of ERDF R&I (assigned to at least one SGC) and Horizon 2020 funding by SGC in 806 NUTS-3 regions that receive funding from both instruments Category ERDF R&I Horizon 2020 Minimum share Maximum share Minimum share Maximum share Climate 5.7% (Bulgaria) 20.1% (Cyprus) 6.2% (Luxembourg) 18.7% (Finland) Energy 7.4% (Sweden) 42.9% (Luxembourg) 13.8% (Luxembourg) 35.9% (Bulgaria) Food 0.0% (Luxembourg) 21.8% (Slovenia) 6.2% (Luxembourg) 41.8% (Slovakia) Health 3.1% (Austria) 31.4% (Bulgaria) 6.4% (Bulgaria) 36.7% (Netherlands) Security 1.2% (Belgium) 20.0% (Bulgaria) 1.9% (Denmark) 20.1% (Cyprus) Society 9.3% (Austria) 45.4% (Sweden) 2.0% (France) 13.7% (Lithuania) Transport 5.9% (Luxembourg) 40.8% (Austria) 5.1% (Estonia) 39.8% (Czechia) Source: CORDIS database, Kohesio database, own elaborations. Note: ERDF R&I represents the total eligible expenditure assigned to R&I projects co-funded by the ERDF and assigned to at least one SGC. Horizon 2020 contains the respective funding amount. This table proves a considerable variation in the distribution of funding across SGC in the Member States. Table A.2. Number of NUTS-3 regions (NUTS version 2021) that receive both Horizon 2020 and ERDF R&I funding Member State No. of NUTS -3 regions … without Horizon 2020 … without ERDF R&I funding … considered in this analysis Member State No. of NUTS -3 regions … without Horizon 2020 … without ERDF R&I funding … considered in this analysis AT 35 3 9 26 IE 8 8 n.a. 0 BE 44 7 20 22 IT 107 5 7 96 BG 28 10 0 18 LT 10 4 0 6 CY 1 0 0 1 LU 1 0 0 1 CZ 14 1 0 13 LV 6 0 0 6 DE 401 73 128 231 MT 2 2 n.a. 0 DK 11 0 1 10 NL 40 0 4 36 EE 5 0 0 5 PL 73 25 1 47 EL 52 46 n.a. 0 PT 25 1 1 23 ES 59 5 5 54 RO 42 4 10 30 FI 19 0 1 18 SE 21 0 0 21 FR 101 5 7 90 SI 12 0 0 12 HR 21 7 0 14 SK 8 0 0 8 HU 20 2 0 18 Total 1,166 208 194 806 Source: CORDIS database, Kohesio database, own elaborations. Note: Project funding data for Irish, Greek and Maltese regions is not provided at the NUTS-3 (but only NUTS-2) level and therefore is not considered in this analysis. Note that, e.g., in Romania, every NUTS-3 region receives ERDF grants according to the Kohesio database. The same is true, e.g., for Belgium, however, only few ERDF projects are reported for East Flanders (NUTS-2 region BE23) which explains why 20 NUTS-3 Belgian NUTS-3 regions have not received any ERDF R&I funding. Transition regions (capital region and no. of SGC in S3 priorities omitted) All observations Trimmed dataset DE-MEANED VARIABLES (instead of NUTS-2 FE) Cosine_sim Gini_Horizon Gini_ERDF Cosine_sim Gini_Horizon Gini_ERDF University 0.207*** -0.128*** -0.172*** 0.223*** -0.130*** -0.182*** (0.053) (0.025) (0.032) (0.054) (0.025) (0.032) LQ knowledge-intensive -0.284 -0.684 -1.050* -0.192 -0.898* -1.143* services (KIS) (1.057) (0.493) (0.638) (1.061) (0.495) (0.637) LQ KIS squared 0.005 0.427 0.642 -0.051 0.516 0.675 (0.713) (0.332) (0.430) (0.707) (0.330) (0.424) ln_GDP per capita -0.433** -0.180* 0.195 -0.508** -0.203** 0.194 (0.218) (0.102) (0.132) (0.222) (0.104) (0.133) Investment rate 1.356 2.243 0.800 -0.279 2.598 1.462 (3.571) (1.665) (2.156) (3.583) (1.671) (2.151) Competition level 2.181 -0.131 -0.001 1.778 0.978 0.869 (1.584) (0.739) (0.956) (1.769) (0.825) (1.062) ln_Population density 0.137*** -0.031 -0.050* 0.155*** -0.019 -0.043 (0.045) (0.021) (0.027) (0.048) (0.023) (0.029) Constant 0.000 0.000 -0.000 0.004 -0.001 0.005 (0.020) (0.009) (0.012) (0.020) (0.009) (0.012) Observations 115 115 115 108 108 108 R-squared 0.234 0.430 0.345 0.246 0.460 0.364 Continued on the next page … More developed regions All observations Trimmed dataset DE-MEANED VARIABLES (instead of NUTS-2 FE) Cosine_sim Gini_Horizon Gini_ERDF Cosine_sim Gini_Horizon Gini_ERDF Capital region 0.013 -0.064 0.046 0.167 -0.146* 0.174* (0.104) (0.056) (0.061) (0.155) (0.084) (0.089) University 0.074*** -0.077*** -0.066*** 0.093*** -0.080*** -0.073*** (0.027) (0.014) (0.016) (0.029) (0.016) (0.017) LQ knowledge-intensive 0.325** -0.442*** -0.450*** 0.271 -0.433*** -0.746*** services (KIS) (0.160) (0.086) (0.094) (0.213) (0.115) (0.122) LQ KIS squared -0.092* 0.123*** 0.123*** -0.091 0.114** 0.233*** (0.055) (0.030) (0.032) (0.086) (0.047) (0.049) ln_GDP per capita 0.314*** -0.086** -0.146*** 0.281*** -0.099** -0.125*** (0.067) (0.036) (0.039) (0.082) (0.044) (0.047) Investment rate 2.814** -1.520*** -0.087 3.521*** -2.018*** -0.433 (1.095) (0.587) (0.645) (1.233) (0.667) (0.705) Competition level 0.165 -0.214 0.058 0.189 -0.191 0.375 (0.427) (0.229) (0.251) (0.494) (0.267) (0.283) ln_Population density -0.000 -0.055*** 0.014 0.003 -0.054*** 0.024** (0.018) (0.009) (0.010) (0.019) (0.010) (0.011) No. of SGC in S3 priorities -0.057* -0.025 0.008 -0.056* -0.024 -0.001 (excl. national priorities) (0.031) (0.017) (0.019) (0.034) (0.018) (0.019) Constant -0.000 0.000 0.000 -0.001 -0.004 -0.004 (0.009) (0.005) (0.006) (0.011) (0.006) (0.006) Observations 480 480 480 393 393 393 R-squared 0.184 0.448 0.221 0.166 0.450 0.261 Notes: The trimmed dataset is excluding the top and bottom decile of NUTS-3 regions according to the ratio between Horizon 2020 and ERDF R&I funding. Significance level: *** p < 0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses. SUR results by innovation group Table B.4. Seemingly unrelated regression results by innovation group DE-MEANED VARIABLES All observations Trimmed dataset Emerging innovators (no. of SGCs in S3 priorities omitted) Cosine_sim Gini_Horizon Gini_ERDF Cosine_sim Gini_Horizon Gini_ERDF Capital region -0.212 0.575*** -0.139 -0.136 0.456** -0.022 (0.275) (0.176) (0.199) (0.303) (0.203) (0.216) University 0.148*** -0.085*** -0.137*** 0.175*** -0.124*** -0.166*** (0.045) (0.029) (0.033) (0.052) (0.035) (0.037) LQ knowledge-intensive -0.069 -0.109 -0.123 -0.106 -0.145 -0.032 services (KIS) (0.405) (0.259) (0.292) (0.428) (0.287) (0.305) LQ KIS squared 0.375 -0.366* 0.029 0.288 -0.278 0.005 (0.322) (0.206) (0.233) (0.343) (0.230) (0.244) ln_GDP per capita 0.045 0.023 -0.136 0.119 0.061 -0.211** (0.130) (0.083) (0.094) (0.146) (0.098) (0.104) Investment rate 1.186 -0.368 -2.301*** 2.674* -0.599 -2.805** (1.226) (0.786) (0.886) (1.567) (1.052) (1.118) Competition level -0.571 0.200 -0.431 -0.351 0.465 -0.540 (0.704) (0.451) (0.509) (0.865) (0.581) (0.617) ln_Population density -0.002 -0.066*** 0.034 0.019 -0.075*** 0.010 (0.032) (0.021) (0.023) (0.040) (0.027) (0.028) Constant 0.000 0.000 -0.000 0.025 -0.022** 0.005 (0.013) (0.008) (0.009) (0.016) (0.011) (0.011) Observations 169 169 169 114 114 114 R-squared 0.261 0.533 0.233 0.319 0.541 0.352 Continued on the next page … Moderate innovators Cosine_sim Gini_Horizon Gini_ERDF Cosine_sim Gini_Horizon Gini_ERDF Capital region 0.159 -0.229 0.093 0.123 -0.232 0.049 (0.262) (0.146) (0.154) (0.271) (0.153) (0.155) University 0.115*** -0.102*** -0.080*** 0.104*** -0.096*** -0.083*** (0.034) (0.019) (0.020) (0.035) (0.020) (0.020) LQ knowledge-intensive 0.781 -0.687*** -0.756*** 0.886* -0.739** -1.032*** services (KIS) (0.475) (0.265) (0.279) (0.528) (0.298) (0.302) LQ KIS squared -0.399 0.299* 0.319* -0.463 0.340* 0.485** (0.313) (0.174) (0.184) (0.343) (0.194) (0.197) ln_GDP per capita 0.163 -0.100* -0.007 0.179 -0.095 -0.041 (0.107) (0.059) (0.063) (0.116) (0.065) (0.066) Investment rate 5.088*** 0.036 -2.527** 5.143*** -0.014 -2.845*** (1.680) (0.936) (0.985) (1.751) (0.988) (1.002) Competition level 0.460 -0.498 0.581 0.614 -0.511 0.912** (0.683) (0.380) (0.401) (0.732) (0.413) (0.419) ln_Population density 0.024 -0.061*** -0.004 0.017 -0.064*** 0.004 (0.026) (0.014) (0.015) (0.028) (0.016) (0.016) No. of SGC in S3 priorities 0.040 -0.086** -0.007 0.017 -0.074 -0.042 (excl. national priorities) (0.071) (0.039) (0.041) (0.087) (0.049) (0.050) Constant 0.000 0.000 -0.000 0.002 -0.003 0.003 (0.012) (0.007) (0.007) (0.013) (0.007) (0.008) Observations 236 236 236 214 214 214 R-squared 0.214 0.518 0.251 0.199 0.499 0.289 Continued on the next page … Strong innovators Cosine_sim Gini_Horizon Gini_ERDF Cosine_sim Gini_Horizon Gini_ERDF Capital region -0.220 0.096 0.257** -0.209 0.074 0.302** (0.200) (0.105) (0.109) (0.255) (0.133) (0.138) University 0.075** -0.041** -0.055*** 0.075* -0.047** -0.063*** (0.037) (0.019) (0.020) (0.039) (0.020) (0.021) LQ knowledge-intensive 0.554 -0.638*** -0.948*** 0.686 -0.865*** -0.913*** services (KIS) (0.389) (0.203) (0.213) (0.468) (0.245) (0.253) LQ KIS squared -0.282 0.222** 0.341*** -0.378 0.374*** 0.315** (0.184) (0.096) (0.101) (0.241) (0.126) (0.131) ln_GDP per capita 0.172 -0.216*** -0.143** 0.210* -0.246*** -0.145** (0.108) (0.056) (0.059) (0.116) (0.061) (0.063) Investment rate 1.016 -2.860** -0.154 1.308 -3.080** -0.006 (2.231) (1.166) (1.221) (2.376) (1.242) (1.285) Competition level -0.475 -0.441 0.974** -0.369 -0.469 0.780* (0.789) (0.412) (0.431) (0.832) (0.435) (0.450) ln_Population density 0.060** -0.039*** 0.016 0.062** -0.041*** 0.018 (0.024) (0.013) (0.013) (0.025) (0.013) (0.013) No. of SGC in S3 priorities -0.048 -0.014 -0.020 -0.054 -0.007 -0.017 (excl. national priorities) (0.054) (0.028) (0.029) (0.055) (0.029) (0.030) Constant -0.000 0.000 0.000 -0.004 -0.002 0.000 (0.013) (0.007) (0.007) (0.014) (0.007) (0.008) Observations 274 274 274 242 242 242 R-squared 0.152 0.430 0.283 0.166 0.445 0.309 Continued on the next page … Innovation leaders Cosine_sim Gini_Horizon Gini_ERDF Cosine_sim Gini_Horizon Gini_ERDF Capital region -0.008 -0.039 -0.162* 0.753 -0.208 0.108 (0.153) (0.081) (0.093) (0.472) (0.266) (0.303) University 0.045 -0.090*** -0.080** 0.141** -0.089*** -0.118*** (0.053) (0.028) (0.032) (0.060) (0.034) (0.039) LQ knowledge-intensive 0.765*** -0.379** 0.007 0.914* -0.515* -0.672** services (KIS) (0.286) (0.151) (0.174) (0.478) (0.269) (0.306) LQ KIS squared -0.202** 0.103** -0.001 -0.263* 0.121 0.194* (0.083) (0.044) (0.050) (0.155) (0.087) (0.099) ln_GDP per capita 0.540*** -0.110* -0.205*** 0.497*** -0.155 -0.114 (0.116) (0.061) (0.071) (0.171) (0.097) (0.110) Investment rate 1.367 -0.534 1.744 3.690 -1.054 0.475 (1.912) (1.011) (1.166) (2.421) (1.363) (1.553) Competition level 0.332 -0.397 -0.857** -0.281 -0.160 0.361 (0.702) (0.371) (0.428) (0.909) (0.512) (0.583) ln_Population density -0.118*** -0.030 0.032 -0.169*** -0.001 0.069** (0.037) (0.019) (0.022) (0.050) (0.028) (0.032) No. of SGC in S3 priorities -0.079* -0.030 0.015 -0.061 -0.033 0.008 (excl. national priorities) (0.041) (0.022) (0.025) (0.044) (0.025) (0.028) Constant 0.000 -0.000 -0.000 -0.002 -0.007 -0.009 (0.018) (0.010) (0.011) (0.025) (0.014) (0.016) Observations 127 127 127 75 75 75 R-squared 0.313 0.455 0.266 0.335 0.450 0.234 Notes: The trimmed dataset is excluding the top and bottom decile of NUTS-3 regions according to the ratio between Horizon 2020 and ERDF R&I funding. Significance level: *** p < 0.01, ** p<0.05, * p<0.1. Standard errors are in parentheses. Robustness checks: OLS regression considering the number of universities and the number of students instead of the University indicator. Table B.5. Considering the number of universities per NUTS-3 region in 2013 - Correspondence between the cosine similarity index and socio-economic characteristics (OLS regression) OLS regression (1) (2) (3) (4) (5) (6) VARIABLES All Observations Trimmed All Observations Trimmed Trimmed Trimmed Capital region 0.014 0.039 -0.069 0.014 -0.085 -0.079 (0.042) (0.052) (0.089) (0.098) (0.082) (0.081) ln_Number of universities 0.117*** 0.128*** 0.095*** 0.121*** 0.124*** 0.126*** (0.017) (0.018) (0.024) (0.027) (0.025) (0.025) Location Quotient (LQ) -0.032 0.087 0.309** 0.355* 0.310* 0.292 knowledge-intensive services (KIS) (0.113) (0.157) (0.136) (0.185) (0.182) (0.182) LQ KIS squared -0.009 -0.070 -0.080 -0.135 -0.090 -0.086 (0.048) (0.078) (0.049) (0.088) (0.086) (0.086) ln_GDP per capita 0.028 0.023 0.172** 0.126 (0.021) (0.026) (0.070) (0.087) Investment rate -0.511** -0.495* 2.388*** 3.753*** (0.255) (0.280) (0.855) (1.186) Competition level 0.491* 0.690** -0.081 -0.099 -0.380 -0.380 (0.298) (0.342) (0.403) (0.503) (0.429) (0.429) ln_Population density 0.010 0.005 0.013 0.015 0.024 0.023 (0.010) (0.011) (0.019) (0.021) (0.020) (0.020) No. of SGC in S3 priorities -0.007 -0.011** -0.047** -0.061** -0.062** (excl. national S3 priorities) (0.005) (0.005) (0.023) (0.026) (0.026) Constant -0.129 -0.238 -1.552 -1.350 1.024** 0.725 (0.337) (0.419) (0.999) (1.208) (0.472) (0.459) Observations 800 642 800 642 642 642 NUTS-2 FE NO NO YES YES YES YES R-squared 0.104 0.113 0.454 0.498 0.485 0.482 Adjusted R-squared 0.094 0.100 0.240 0.262 0.247 0.244 Wald-test for joint significance (p-value) 0.000 0.000 0.000 0.000 VIF 4.22 4.59 3.55 3.66 3.12 2.48 VIF > 10 (excl. squared terms) - - S3 prios, ln_GDP_pc, Invrate S3 prios, ln_GDP_pc, Invrate S3 prios - Notes: The dependent variable is the cosine similarity index. Instead of the University dummy in the baseline regression, the (natural log of the) number of universities per NUTS-3 region in 2013 is used. Significance level: *** p < 0.01, ** p<0.05, * p<0.1. Heteroskedasticity-robust standard errors are in parentheses. Table B.6. Considering the number of students (ISCED 5-7) per NUTS-3 region in 2013 - Correspondence between the cosine similarity index and socio-economic characteristics OLS regression (1) (2) (3) (4) (5) (6) VARIABLES All Observations Trimmed All Observations Trimmed Trimmed Trimmed Capital region 0.124*** 0.168*** 0.023 0.121 0.012 0.017 (0.037) (0.048) (0.097) (0.127) (0.097) (0.096) ln_Number of students 0.017*** 0.018*** 0.012*** 0.015*** 0.015*** 0.015*** (0.002) (0.002) (0.003) (0.003) (0.003) (0.003) LQ knowledge-intensive -0.087 -0.029 0.248* 0.294 0.259 0.246 services (KIS) (0.101) (0.152) (0.141) (0.199) (0.193) (0.192) LQ KIS squared 0.012 -0.019 -0.065 -0.116 -0.072 -0.068 (0.038) (0.070) (0.047) (0.090) (0.087) (0.087) ln_GDP per capita 0.025 0.022 0.200*** 0.154* (0.021) (0.027) (0.069) (0.087) Investment rate -0.480* -0.469 2.403*** 3.701*** (0.260) (0.288) (0.897) (1.301) Competition level 0.569* 0.784** 0.160 0.151 -0.217 -0.226 (0.300) (0.333) (0.396) (0.500) (0.423) (0.421) ln_Population density 0.019* 0.014 0.017 0.023 0.036* 0.035* (0.010) (0.012) (0.019) (0.021) (0.019) (0.019) No. of SGC in S3 priorities -0.008* -0.012** -0.028 -0.048 -0.044 (excl. national priorities) (0.005) (0.005) (0.028) (0.033) (0.032) Constant -0.204 -0.327 -2.124** -1.903 0.776* 0.571 (0.343) (0.427) (0.971) (1.197) (0.463) (0.447) Observations 779 622 779 622 622 622 NUTS-2 FE NO NO YES YES YES YES R-squared 0.124 0.134 0.461 0.502 0.489 0.488 Adjusted R-squared 0.113 0.121 0.244 0.260 0.244 0.244 Wald-test for joint significance (p-value) 0.000 0.000 0.000 0.000 Mean VIF 4.18 4.56 3.78 4.06 3.51 2.47 VIF > 10 (excl. squared terms) - - S3 prios, ln_GDP_pc, Invrate S3 prios, ln_GDP_pc, Invrate S3 prios - Notes: The dependent variable is the cosine similarity index. Instead of the University dummy in the baseline regression, the (natural log of the) number of students (ISCED levels 5-7) per NUTS-3 region in 2013 is used. Significance level: *** p < 0.01, ** p<0.05, * p<0.1. Heteroskedasticity-robust standard errors are in parentheses. List of figures Figure 1. Distribution of Horizon 2020 and ERDF R&I amounts per capita among NUTS-3 regions 15 Figure 2. Horizon 2020/ERDF R&I ratio in different groups of NUTS-3 regions based on their economic development .................................................................................................................................................................... 16 Figure 3. Funding shares allocated to different SGCs in NUTS-3 regions part of … .................................. 20 Figure 4. Overall similarity of the ERDF R&I and Horizon 2020 funding distribution across SGCs by NUTS-3 region (cosine similarity index) ................................................................................................................................. 23 List of tables Table 1. Share of ERDF R&I (assigned to at least one SGC) and Horizon 2020 funding by SGC ........ 14 Table 2. Explanatory variables at the NUTS-3 regional level: variable description and source ........... 19 Table 3. Cosine similarity index by NUTS-3 regions that receive both ERDF R&I and Horizon 2020 funding ...................................................................................................................................................................................................... 24 Table 4. Correspondence between the cosine similarity index and socio-economic characteristics in EU NUTS-3 regions (OLS regression) ....................................................................................................................................... 26 Table 5. Correspondence between the cosine similarity index and socio-economic characteristics in NUTS-3 regions by development group ................................................................................................................................. 28 Table 6. Correspondence between the cosine similarity index and socio-economic characteristics in NUTS-3 regions by innovation group ....................................................................................................................................... 29 Table 7. Seemingly unrelated regression (SUR) results .............................................................................................. 31