Regional paths towards Europe 2020 targets: a spatial approach

ABSTRACT

Europe 2020 is a 10-year EU strategy, promoting smart, sustainable and inclusive growth. Despite ambitious goals, its spatially blinded approach might seriously threaten its success. Actually, large territorial disparities still affect the EU, being the basis for a strong EU-wide cohesion policy and suggesting a general re-framing of sectoral policies on a regional basis. In this respect, the paper tackles the issue of regional disparities in achieving Europe 2020 Strategy targets. As the Strategy involves different targets, principal component analysis is applied to disentangle Europe 2020 domains and to describe major differences in EU-27 NUTS 2 regional performances. In particular, two components are returned: high-employment inclusive growth and smart growth. Territorial patterns of both components are analysed, by jointly considering some geographical features that may affect them. Both a rural and a spatial effect occur: rural and remote regions show poor performances whilst the presence of spatial autocorrelation may actually lock-in negative outcomes. When considering urban rural divides, also within-regions disparities matter. Results strengthen the idea that Europe 2020, as other sectoral policies, should be translated into a regional setting according to a place-based approach: although requiring large efforts, this represents the only way to fully achieve its own targets.

KEYWORDS:

• Europe 2020
• PCA
• spatial analysis
• regions

Disclosure statement

No potential conflict of interest was reported by the author.

ORCID

Francesco Pagliacci http://orcid.org/0000-0002-3667-7115

Notes

1. Being both heterogeneous and ambitious, the effectiveness of these targets pose major questions. Many Member States have not achieved Lisbon Strategy’s goals, so it is unrealistic to think that they would be able to achieve Europe 2020 ones, also because of the ongoing international economic crisis. In fact, especially EU peripheral Member States are still far apart from expected targets.

2. According to Natali (2010), two groups of criticisms occur. Firstly, the Strategy’s political and economic foundations have been considered as mostly wrong for EU integration (Hopner & Schafer, 2007), prompting a wrong political agenda (Amable, 2009; Rodrigues, 2002). Secondly, its governance has been ineffective. The EU has never developed proper economic policy institutions to foster its own growth and to stimulate national engagement (Copeland, 2012). Indeed, Member States’ participation has been uneven, the role of OMC being too weak (Smismans, 2008; Tucker, 2003; Zeitlin, 2007, 2008; Copeland, 2012).

3. As claimed by many theoretical economic models, this tendency is expected to continue even in the future: for instance, the New Economic Geography clearly suggests the persistency of divergent core-periphery paths (Krugman, 1991).

4. PCA has been widely adopted in order to solve classification problems in regional studies (Vidal, Eiden, & Hay, 2005), as it does not require any ex-ante assumption: original data determine transformation vectors. According to these properties, PCA is here used as an exploratory analysis.

5. The author is aware that such a territorial level may not fit well into the scope of this manuscript, as some EU Member States (such as the Baltic States or Luxembourg) just comprise one single NUTS 2 region. Nonetheless, such a choice relies on two major aspects: lack of available data at a greater territorial disaggregation and expected homogeneity – in terms of population – among NUTS 2 regions throughout the EU.

6. Population standard deviation returns a finite value (namely, zero), even when a NUTS 2 region comprise just one NUTS 3 region.

7. Selected variables are suitable for PC extraction, according to the Kaiser–Meyer–Olkin (KMO) test. It is a sampling adequacy test, which may range from 0 to 1 (Kaiser, 1974). Here, KMO test on the variables under study is fully satisfactory (0.784).

8. Differences among typologies are tested through one-way analysis of variance (ANOVA), which uses F statistics to test if all groups share the same mean or not. As a major assumption of one-way ANOVA is homoschedasticity in population variances, the Levene’s test has been preliminary computed. When group variances are equal, simple F test for the equality of means is performed. In the opposite case, the method of Welch (1951) is used. Here, differences are all proved to be statistically significant at 5% (2-tailed).

9. Nordic countries: Denmark, Finland and Sweden. Continental Western Europe: France, Luxembourg, Belgium, the Netherlands, Germany and Austria. Mediterranean Europe: Portugal, Spain, Italy, Greece, Malta and Cyprus. Eastern EU: Estonia, Latvia, Lithuania, Poland, the Czech Republic, Slovakia, Slovenia, Hungary, Romania and Bulgaria.

10. Results are quite robust. Similar findings are obtained when a five nearest-neighbour matrix is considered. Results are available upon request.