Sustainable competitiveness: a spatial econometric analysis of European regions

Abstract

The paper complements the few regional studies on the sustainability–competitiveness nexus by providing a novel composite index of sustainable competitiveness for 272 European regions in 28 European countries. Principal component factor analysis is combined with a variance-based structural equation model to create a statistically reliable index, which overcomes the methodological issues of previous studies. Especially, the use of the latter also allows estimation of the cause–effect relationships between the different pillars of sustainable competitiveness, where empirical evidence is scarce. The paper shows that favorable ecological, social, and economic environments can jointly contribute to facilitating long-term sustainable competitiveness outcomes. Thereby, the progress in one dimension is not necessarily at the expense of another dimension of sustainable competitiveness. The proposed index reveals important insights for policymakers into the sustainable competitiveness trajectory of European regions. Region-specific plans for action can be derived and new policy conclusions can be drawn from the index.

Keywords:

• sustainable regional competitiveness
• composite index
• Beyond-GDP
• structural equation modeling
• European studies

Supplemental data

Supplemental data for this article can be accessed here.

Notes

1 Thereby, the argumentation is mainly against conceptions of competitiveness based on low wages, low production costs, as well as beggar-thy-neighbor policies.

2 For example, Porter (1998)’s cluster theory and Krugman (1991)’s new economic geography identify local agglomeration of economic activities as significant sources for external economies.

3 Moreover, the idea of sustainable competitiveness is not only increasingly applied at the national and regional level but also to the more disaggregated level of cities. For instance, Ni and Wang (2017) provide a framework of urban sustainable competitiveness for 500 cities in 130 countries. They identify seven sub-components to be important for urban sustainable competitiveness. While six pillars are measured by a single indicator, one pillar is constructed by applying arbitrary aggregation rates to two indicators. PCA is then used in the final step to relate the pillars to each other.

4 In particular, the indicators are employment, R&D spending, education, and fighting against poverty and social exclusion. Due to data availability, the environmental dimension of the Europe 2020 targets is omitted from the index formation.

5 Apart from the endowment measure based on population, total land area has been used to proxy the endowment pillar. However, as the results do not significantly change, Table 4 displays only the results for endowment reflected by population.

6 Due to the relatively sparse availability of regional ecological and social data, the decision was made towards a cross-sectional sample. If there was no data available for the year 2014 at the European NUTS-2 level, the previous year was used. In some exceptional cases, NUTS-1, NUTS-0 data, or a 3 year average was used to close the gaps and to smooth the dataset. In addition, missing data was interpolated by utilizing an inverse-distance interpolation considering the geographical 5-nearest neighbors.

7 The current version of the nomenclature of territorial units NUTS 2013/EU-28 (Eurostat, 2015) is used to classify the European regions.

8 In some cases, sufficient data is not available on the NUTS-2 level. While this is not an issue when measuring conditions that do not change within the country, such as macroeconomic characteristics, it becomes problematic when measuring, for example, microeconomic or ecological conditions, which vary across regions. In order to create a latent variable, which measures the underlying dimension of sustainable competitiveness as precisely as possible, a mixture of both NUTS-2 and NUTS-0 indicators have been used for the construction of some components. The utilized regional and national indicators aim to reflect one common underlying dimension, which is not directly observable. For this purpose, the PCA extracts from the indicators at the NUTS-2 and NUTS-0 only the information, which describes these underlying common dimensions. The factor loadings for the indicators are then used to weight their contribution according to their relevance for the respective sustainable competitiveness component. After this procedure, a composite component has been created where, despite combining data at different levels of aggregation, the factor scores are regionally distributed and vary across regions.

9 One exception is the monetary and fiscal policy component, which is not compiled using PCA. In adopting [24]’s approach, the component is constructed as follows: (1) define a neutral zone: government deficit/surplus $[<-3\mathrm{\%}],$ general government gross debt $[<60\mathrm{\%}],$ inflation $[0.5-3.0\mathrm{\%}];$ (2) indicators within the neutral zone: 0, outside: $-\mathit{Ln}(1+\mathit{deviationneutralzone})$ (3) $\mathit{MFP}=0.25\mathrm{*}\mathit{governmentdeficit}+0.25\mathrm{*}\mathit{governmentdebt}+0.5\mathrm{*}\mathit{inflation}.$

10 In particular, CA is used to evaluate the composite reliability as well as to check for internal consistency and the KMO criterion is utilized to measure sampling adequacy.

11 One interesting exception is the study by Aiginger and Firgo (2017), who use multivariate statistics in the first step of aggregation to overcome the statistical challenges in constructing their sub-components. For the other European regional indices, the weighting and aggregation of the sub-components into the overall index score is not based on empirical relationships.

12 In this case, the term dependent variable means here either composite constructs (Aiginger and Vogel 2015; Aiginger and Firgo 2017; Despotovic et al. 2016) or productivity measures (Delgado et al. 2012).

13 The notation is largely adopted from Henseler, Ringle, and Sinkovics (2009)

14 The rationale is that both consist of three interdependent dimensions, which measure different aspects. Hence, extracting enough commonality to identify the latent variable is rather difficult. A possible solution to this identification problem is a formative-reflective outer measurement model. In a first step, this higher order construct estimates, with the help of a formative outer measurement model, three different components. In a second step, a reflective outer measurement model, which depends on the relationships among the whole path model, defines the final latent variable sustainable competitiveness outcomes or intermediate outputs, respectively.

15 To check the robustness of the results, different inner structural model specifications have been tested in Table D2 in Appendix D (online supplemental material). However, the reported model in Table 4 is the preferred specification with the highest explanatory power.

16 For example, the calculation of the total effects of microeconomic competitiveness on sustainable competitiveness outcomes can be done as follows: total effects${}_{micro\to outc}=\underset{direc{t}_{micro\to outc}}{\underbrace{0.124}}+\underset{indirec{t}_{micro\to outc}}{\underbrace{0.569\text{*}0.170}}=0.221.$ As can be seen, a rise in microeconomic competitiveness by one standard deviation is, on average, associated with a 0.221-point increase in the composite sustainable competitiveness outcomes. The latent variable intermediate outputs are, in this case, the mediating construct by which the indirect effect is channeled. According to Hair (2017), two types of non-mediation exist: direct-only non-mediation and no-effect mediation as well as three types of mediation: complementary, competitive, and indirect-only mediation. Decomposing the total effect into a direct and indirect allows a more comprehensive analysis of the cause–effect relationships of the different components of sustainable competitiveness, which is, according to previous econometric analysis, (Aiginger and Firgo 2017; Delgado et al. 2012) a methodological improvement.

17 The VIF-values for the inner structural model can be provided upon request.

18 Figure 1 and Table 10 in Appendix E (online supplemental material) show that regions with capitals have, in most cases, higher values of GDP per inhabitant than rural regions. However, urban areas do not have the same green potentials as rural areas. As a consequence, low performance in ecological fundamentals corresponds with high performance in the economic dimension of sustainable competitiveness outcomes. This observation also holds true for the other indicators of the economic dimension in sustainable competitiveness outcomes: long-term unemployment, compensation of employees, gross value added.

19 Figure E2 represents a graphical visualization of that comparison and plots the SCO ranking (y-axis) against the other regional rankings (x-axis). Thereby, the relative position of the PIIGS countries, which are represented by Portuguese, Italian, Irish, Greece, and Spanish regions, are highlighted with a cross.