4. Regional Dynamics in EU11

This review is not exhaustive. Among many contributions, an important recent review of theoretical and empirical research is in Breinlich, Ottaviano and Temple (2014). Less recent, but more comprehensive, reviews of empirical research are in Magrini (2004), and, focused on Europe, Monfort (2008). Recent contributions include Ottaviano (2019), Alfaro, Chen and Fadinger (2019), Demertzis, Sapir and Wolff (2019). The definition of convergence, empirical methodology and findings vary across papers. Only the trends that are broadly consistent across papers, sometimes with many exceptions, are reported above.

As a caveat, similar to all administrative units, even NUTS 3 regions may have arbitrary boundaries that do not correspond to actual geographical areas of economic activity.

The database is available at: https://​ec.​europa.​eu/​knowledge4policy​/​territorial/​ardeco-database_​en. Data were downloaded in July 2019.

In order to approximate better the average citizen’s purchasing power across locations, GDP is usually corrected for differences in prices. Price indexes are however available only at national level. Since prices tend to be higher in higher income areas, it is likely that our measure of GDP per capita overestimates differences in real purchasing power at regional level.

An alternative definition of convergence is β-convergence. There is β-convergence when the regions that were initially poorer are growing faster. It has been shown that σ-convergence is a sufficient condition for β-convergence. For a comprehensive review of the definitions and measurements of regional convergence, see Eurofound (2018).

Regions are categorised in four groups: in addition to urban, intermediate and rural (according to European Commission 2017, p. 21) regions, the NUTS 3 regions including the capital cities are treated separately from urban regions.

This seems a long-term trend. Ezcurra, Pascual and Rapún (2007) found multimodality in the regional distribution of income per capita in 1900, which substantially weakened by 2001.

GDP per capita in capitals could be inflated by the concentration of company headquarters in these cities. This concentration allocates a disproportionate share of gross operating surplus to capitals, even though these firms may operate in several regions.

The unemployment rate is the number of persons not working but actively searching for employment, as a percentage of the labour force (unemployed plus those in paid or self-employment). The non-employment rate is the number of those not working (irrespective of whether they are searching for work) as a percentage of the working-age population.

To see why, consider two opposing cases. First, assume that labour is fully mobile and individuals’ utility is determined only by their real income (purchasing power). One would expect people to move to higher productivity places where they reap higher nominal incomes. This migration would continue until the prices of localised goods (e.g., land, property, local services) change sufficiently to equalise real income across locations. Second, assume instead that localities have the same productivity level but different non-monetary amenities. Then, people may accept lower income in exchange for better amenities (e.g. better climate or nicer landscapes). The joint behaviour of local prices and local population dynamics can differentiate between these two cases. If higher income per capita is accompanied by higher population growth, productivity differences would be the main driver of local differences in income per capita. If higher income per capita is instead accompanied by lower population growth, local non-monetary amenities would be the main driver. In this chapter, we find that local population dynamics and incomes are positively correlated, which would suggest that they primarily reflect productivity differences. For a more complete explanation of these mechanisms, see for instance, Breinlich, Ottaviano and Temple (2014).

The position of regions (poor/rich and fast/slow growing) is always relative to their national average. This implies that a Type 1 region in Country A may grow slower than a Type 4 region in Country B, if Country B national average growth is sufficiently higher than Country A national average growth.

Neighbouring regions more than 200 kilometres away are not considered for this analysis, their weights are set to zero. This implies that spillovers are considered negligible beyond this threshold. As a robustness check, various thresholds were tried between 100 and 500 kilometres, but the results did not change materially. The calculations also take into account the GDP per capita in bordering regions in Western Europe.

In 2000, the slope coefficient I = 0.505 with a standard deviation of 0.033 (z = 15.441, significant at all levels). By 2015, the slope had halved (0.238) while remaining statistically significant (standard deviation of 0.033, z = 7.394, significant at all levels).

The calculations take into account the GDP per capita in bordering regions in Western Europe.

Rosés and Wolf (2018) also find evidence of weakening spatial dependency in the EU as a whole over the long term (1900–2010).

See Baldwin et al. (2003) for a detailed analysis of the welfare implications of New Economic Geography models. Ottaviano and Pinelli (2005) provides a summary of the relevant literature.

This is because these models include a non-linear relation between trade costs and agglomeration effects. Agglomeration effects are minimised when transport costs are either very high (firms have to locate close to their customers) or very low (firms’ location becomes irrelevant). For intermediate costs, agglomeration effects are stronger – in this case, market outcomes tend to be inefficient. Combes et al. (2011) find evidence of such bell-shaped curve with regard to French regions over the period 1860–2000.

Breinlich, Ottaviano and Temple (2014), Austin, Glaeser and Summers (2018), Moretti (2012), Glaeser (2013) also provide comprehensive overviews of regional policies and regional growth drivers.

Breinlich, Ottaviano and Temple (2014) provide a good overview of the topic and analyse the related important analytical gaps.