European Corridors as Carriers of Dynamic Agglomeration Externalities?

Abstract

Transport corridors are viewed as a promising way forward in European Union (EU) transport policy, assumed to contribute positively to regional economic development. However, the validity of this assumption is not evident. The aim of this paper is to empirically test whether agglomeration economies in European transport corridor regions are positively related to indicators of regional economic development compared to regions outside the scope of corridors. The results build on the notion that the type of agglomeration economy in combination with the structure of the economy matters for prospects of structural economic growth in different regions. In this way, the analysis not only contributes to enhancing the empirical scrutiny of the corridor concept in EU transport policy, but also provides new insights into how corridors contribute to regional economic growth. We find only limited evidence for a corridor effect across European regions on productivity and employment growth externalities. Instead, we find a large degree of spatial heterogeneity interacting with corridors—a heterogeneity that has been little recognized in EU policies. We suggest that recent attention to place-based development strategies may accord well with the kinds of agglomeration effects related to corridor development observed in this study.

Acknowledgements

This research is part of the INTERREG IV-B-funded project “CODE24”, which intends the interconnection of transport, spatial and economic development along the TEN-T Corridor 24 axis, ranging from Rotterdam to Genoa (http://www.code-24.eu).

Notes

1. De-concentration need not involve the qualitative separation of functions in which more innovative firms remain in city centres and productive firms move out. Instead, suburbanization of economic activity may involve all types of activities (Green Leigh & Blakely, 2013; Phelps, 2004).

2. Data are used for Austria, Belgium, the Czech Republic, Germany, Denmark, Spain, Finland, France, Greece, Hungary, Ireland, Italy, the Netherlands, Poland, Portugal, Romania, Sweden, Slovakia and the UK. For reasons of optimal data comparability, small modifications were made to the regional divisions in Belgium, Sweden and the UK. Data from regions in Norway, Switzerland, Slovenia and Luxembourg are missing.

3. Mimicking the distribution as suggested by the OECD and imposing it on the NUTS2-regional level in the EU implies that some capital regions (e.g. Bratislava, Stockholm) do not appear as large urban regions, while some large-surface regions (e.g. regions in Romania, France and Spain) do. Because much of our data is merely available at the NUTS2-level, we cannot distinguish cities on a lower spatial scale. Combining population data with density and functions may be an alternative way to capture the urban structure of Europe—but for now we wish to test the logic suggested by the OECD. In addition, density is one of our main explanatory variables in all models—imposing it on the regimes would introduce it on the left-hand side of our equations.

4. Three corridors have a north–south orientation; the other corridors are orientated from West to East. Among the corridors identified is Corridor A, connecting Rotterdam (the Netherlands) to Genoa (Italy). The second corridor (B) ranges from Stockholm (Sweden) to Napoli (Italy); the third north–south corridor (C) runs from Antwerp (Belgium) to Lyon (France). The fourth corridor (D) has a west–east orientation, stretching from Valencia (Spain) to Budapest (Hungary). The fifth corridor (E) connects the urban regions of Dresden (Germany) and Constanta (Romania). The final corridor (F) ranges from Aachen (Germany) to Terespol (Poland).

5. Included in the comparison are the TEN-T Priority Axes, RNE corridors, CER Business Cases for a Primary European Rail Freight Network and TREND. The recently defined Core Network Corridors from the Connecting Europe Facility of the European Commission are not yet included. We intend to incorporate these in follow-up empirical research.

6. Explanation of the variables: Employment growth = ln(emp10–emp00); Productivity growth = ln(prod10–prod00); Employment level 2000 = number of employed persons in 2000; Productivity level = labour productivity level in 2000 (all economies); Specialization-diversity = entropy measure over 59 location quotients of sectoral production (==specialized, −=diversified), 2000; Private R&D = percentage of GDP spent on R&D in firms in 2000; Public R&D = percentage of R&D spent on R&D in universities and non-profit institutes in 2000; Openness economy = (exports + imports) as percentage of total trade, 2000; Market potential = gravity value on production with travel time distances, 2000; Educational level = share of tertiary education in working population, 2000.

7. We realize that endogeneity may still be an issue in the models, as some explanatory variables (like educational level) may be partly determined by employment growth and productivity growth.

8. Recall that the core and peripheral European regimes are based on this potential value as well. When using those regimes in models, the market potential variable is left out.

9. Squared inverse distance weights do not capture the spatial dependence in our models for productivity growth and employment growth any more effectively.

10. Productivity levels are generally higher in European cities (Bosma & Van Oort, 2012; OECD, 2012), but growth and level analysis show different outcomes in Europe.