Regional structural heterogeneity: evidence and policy implications for RIS3 in macro-regional strategies

ABSTRACT

In the future of European Union (EU) Cohesion Policy, a critical feature is how to capitalize on the current Research and Innovation Strategy for Smart Specialisation (RIS3), introduced in the 2014–20 programming period as an ex-ante conditionality for accessing European Structural Investment Funds. As a contribution to the current debate, this paper frames the socioeconomic comparative analysis of EU regions, considering subnational structural similarities, and referring to population, labour market and the sectoral composition of the economy. Principal component analysis and cluster analysis are performed on 31 input variables, returning 19 different types of EU regions. With an application to the EU macro-regions, the paper supports policy planning in macro-regions as a meso-level of interventions for the future Cohesion Policy.

KEYWORDS:

• socioeconomic and demographic structural similarity
• Smart Specialisation
• European Union macro-regional strategies

JEL:

• C38
• R12
• R58

ACKNOWLEDGEMENTS

For the discussion on the topics presented in this paper, the authors thank all the EUSALP’s Action Group 1 members.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. EU member states and regions are required: (1) to have a Smart Specialisation in place, based on a strengths, weaknesses, opportunities and threats (SWOT) analysis; (2) to focus the available resources on a limited set of priorities; and (3) to implement a monitoring and review system as a multi-annual plan for the budgeting and prioritization of investments linked to EU priorities (European Commission, 2014b).

2. With regard to RIS3 policy design, benchmarking is a commonly used method, given the availability of harmonized data at the EU level, which makes a comparison across regions possible (Kleibrink & Magro, 2018). In addition to hard data, soft data and peer reviews now play a key role in the support of learning for designing strategic approaches to place-based innovation policies (Kleibrink & Magro, 2018). The review of good practices is also suggested as particularly helpful (Griniece, Panori, Kakderi, Komninos, & Reid, 2017).

3. The database, available on line at Eye@RIS3, has been elaborated by using both free text and coded information which are entered by the regions in the S3 platform to summarize their RIS3 priorities (Pavone et al., 2019).

4. See Camagni et al. (2017) for an example on EUSALP.

5. See http://s3platform.jrc.ec.europa.eu/eye-ris3.

6. See http://s3platform.jrc.ec.europa.eu/regional-benchmarking.

7. Their analysis refers to NUTS-2 regions as statistical units for the analysis, with the only exception of Belgium, Germany and the UK, for which NUTS-1 regions are considered.

8. When there is a single variable, that variable is kept as it is; when there are two variables, aggregation occurs through a simple average; when there are more than two variables, PCA is performed, extracting the minimum number of components (Navarro et al., 2014).

9. Share of employment is considered, with the exception of agriculture, where the share of gross value added is taken. In addition, the touristic sector is included (arrivals per inhabitant).

10. They cover almost the same socioeconomic domains in Navarro et al. (2014), with the exclusion of: inputs that might deliver nation-level biases in the outcome (e.g., multilevel government); average firm size (barely effective in informing on the actual industrial organization at regional level); and trade openness (biased by the location of the subsidiary registering trade activities).

11. PCs explaining at least 70–80% of the total cumulative variance should be taken.

12. By construction, PC averages equal zero and they are ranked in decreasing order according their standard deviations (i.e., the first PC has a larger standard deviation than the second PC, and so on).

13. Industries codes refer to NACE Rev. 2.

14. For the entire set of results, see http://hdl.handle.net/11380/1177860 (doi:10.25431/11380_1177860).

15. This range is by construction for both the HHI and Gini. With regard to the Shannon index, the simple Shannon diversity index is normalized by dividing it by the maximum diversity.

Additional information

Funding

This work is part of the Work Package No: T-3 Enhancing Shared Alpine Governance Project of the project Implementing Alpine Governance Mechanism of the European Strategy for the Alpine Region (AlpGov) of the Interreg Alpine Space Programme – Priority 4 (Well-Governed Alpine Space), SO4.1 (Increase the Application of Multilevel and Transnational Governance in the Alpine Space) [Project Code 379; Subsidy Contract MIN000510A15].