Towards regional renewal: a multilevel perspective for the EU

ABSTRACT

The 2008 financial crisis reopened the debate on regions’ ability to deal with shocks within the European Union. We identify the spatial dimension of the renewal capacity, among the dimensions of economic resilience, and estimate its main drivers. We investigate the variables that determine the regional renewal capacity using different model specifications focusing upon several socioeconomic factors at two geographical scales: national and regional. The results highlight the fact that regional renewal has to be analysed including both local and contextual (national) factors. This multilevel perspective is useful for policy strategies in terms of reorienting their targets to the proper geographical and socioeconomic dimensions.

KEYWORDS:

• regional renewal
• multilevel
• resilience
• European Union

JEL:

• C38
• P25
• R11
• R12

ACKNOWLEDGEMENT

A previous version of this paper was presented at the Regional Studies Association Central and Eastern Europe Conference 2017. The authors are grateful to the seminar participants and to two anonymous referees for useful comments.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors. The scientific output expressed in this paper does not imply a policy position of the European Commission. Neither the European Commission nor any person acting on behalf of it is responsible for the use that might be made of this paper.

Notes

1. Among the latest studies on resilience at the national level, see: Martin, Sunley, Gardiner, and Tyler (2016) and Kitsos and Bishop (2018) for the UK regions; Di Caro (2017, 2018), Sedita, De Noni, and Pilotti (2017), and Faggian, Gemmiti, Jaquet, and Santini (2018) for the Italian regions; Giannakis and Bruggeman (2017b) and Psycharis, Kallioras, and Pantazis (2014) for the Greek regions; Diodato and Weterings (2015) for the Dutch regions; and Cuadrado-Roura and Maroto (2016) and Angulo, Mur, and Trívez (2018) for the Spanish regions.

2. The trend has been computed as follows: (1) we regress the time period on the log of the selected variables and (2) we keep the coefficient associated with them. If it is positive (negative) and significant, it means that the slope rises (falls). If the coefficient is zero or not significant, the trend is not statistically different from zero.

3. Owing to missing data, Croatia has been excluded.

4. This contiguity scheme guarantees that there are no isolated regions, namely the islands. Alternative weighing schemes were used and the result did not change substantially.

5. Moran's I and the spatial econometric analysis were carried out using R library spdep (Bivand, Pebesma, & Gomez-Rubio, 2013; Bivand & Wong 2018).

6. The central tendencies of a distribution, that is, the middle 50% of the distribution, are described in the middle of each box plot. The solid thick line locates the median; the top and bottom edges are the 75th and 25th percentiles, respectively. The height of the box is the interquartile range. The upper (lower) adjacent value is the largest renewal value observed no greater than the 75th (25th) percentile plus 1.5 × r, where r is the interquartile range.

7. The models were estimated using the R function lmer of library lme4 (Bates, Machler, Bolker, & Walker, 2015) and the R function SpatialFiltering from library spdep.

8. We follow Farhauer and Kröll (2012) and focus on the absolute, not relative, specialization to avoid distortions that may arise while using the relative specialization index (RZI). A region being largely employed in a nationwide small branch could have a higher value of RZI than a region with a high share of employment in a sector with high total national employment – even though the latter is much more specialized. Distortion then arises in the sense that the regional concentration of a sector would be confused with specialization. This rationale is also applied to diversification.

9. The indexes are based on 15 NACE-1 sectors from the Cambridge Econometrics European Regional Database.

10. There is no agreement on the effect of specialization and diversification on growth. De Groot, Poot, and Smit (2009), in a meta-analysis, find strong positive evidence of sectoral diversity and competition on growth but contrasting evidence on specialization effects. The same authors, in an updated version of their paper (De Groot, Poot, & Smit, 2016), find that specialization is more important in lower density areas and that more recent studies support less the importance of diversity externalities.

11. ‘Lag’ refers to the spatially lagged dependent variable, while ‘error’ stands for the spatial autoregressive process for the error term. If only one is significant, lag or error, we choose the correspondent model. If both are significant, then we check the robust LM tests. As for the LM tests, if only one is significant, we choose the correspondent model; alternatively, if both are significant, we choose the model with the biggest test value associated. We used the R function lmtest in library spdep.

12. A flat curve generally suggests that investors are unsure about the future (Harvey, 1988).

13. Chinn and Kucko (2010) proved yield spread's ability in forecasting future industrial production growth and recession for the United States as well as for European countries.

14. The OECD (2016) found that gender differences in working hours are driven by disproportionately high rates of part-time employment among female employees in OECD countries. Part-time employment rates for women reach four or five times the size of those for men in Austria, Belgium, Germany, Luxembourg and the Netherlands. Furthermore, a European Parliament (2016) report found that ‘men are more likely to work on a full-time and permanent basis than women (65 % compared with 52 %), whereas women are much more likely than men to work on a part-time basis. A total of 12 % of women work on a part-time basis, compared with 2 % of men, and 15 % of women work on a marginal part-time basis, compared with 4 % of men’.