The Rise of Second-Rank Cities: What Role for Agglomeration Economies?

Abstract

In the last 15 years, empirical evidence has emerged about the fact that European first-rank cities have not always led national economic performance, and when they did, the difference between first- and second-rank cities in explaining national growth has not been significant. A recent work [Dijkstra, L., Garcilazo, E. & McCann, P. (2013) The economic performance of European cities and city regions: Myths and realities, European Planning Studies, 21(3), pp. 334–354] claims that second-rank cities have in fact outperformed first-rank cities, becoming the main driving forces in national economic performance. In the debate that emphasizes the role of second-rank cities in national growth, a simplified view of the role of agglomeration economies is provided; they are taken for granted in small- and medium-sized cities and only in large cities will the problem of a downturn in urban returns to scale emerge. In this paper, a more complex view is assumed, claiming that the oversimplified interpretation that urban economic performance simply depends on the exploitation of agglomeration economies and that these agglomeration economies merely depend on urban size alone should be abandoned. Some already existing theoretical frameworks in urban economics can help in recalling the role of possible bifurcations in the development path of cities, linked to the capability to attract or develop new and higher-order functions, increase internal efficiency and reach scale economies through cooperation networks with other cities (the city-network theory). All these elements work as conditions for fully exploiting agglomeration economies and ways to overcome urban decreasing returns.

Notes

1 An alternative way to explain the efficiency of second-rank cities goes through the role of size borrowing from nearby large metro areas, hosting high-level functions and being connected in transnational networks, as suggested in Alonso (1973). See Camagni et al. (2013).

2 More details on the empirical approach to the definition of urban rank and the relative performance of second-rank cities with respect to larger metro areas in the last 15 years are provided in Section 3.

3 Richardson (1972) suggests replacing the concept of optimal city size with an efficient interval of urban size in which urban marginal benefits are greater than marginal location costs.

4 The two cities will differ, though, in dynamic terms: the one belonging to the lower rank 1 will not grow further, having reached the maximum size of its interval, while the one having developed the higher functions (linked to rank 2) will grow, due to the presence of new and wide net urban benefits (profits).

5 Camagni (1993) theorised the concept applying it to urban systems. The same concept was already utilised in other fields, such as the behaviour of the firm and microeconomic organisational behaviour. For a review of the concept, see Capello and Rietveld (1998).

6 In another paper, the same authors show that these features are common to cities of different size, demonstrating that cities grow according to the same structural laws, but with some specificities (Camagni et al., 2013).

7 For this reason, in the empirical analysis (Section 5), rank 2 cities will only be analysed in EU15 countries. The trends above discussed allow the identification of four periods within the 1995–2009 time span observed in this paper.

8 Urban rent is usually interpreted as the rent paid to the house owner. However, house prices represent the capitalized rent over time, and for this reason may be chosen as a proxy for urban rent.

9 In dynamic terms, the reasoning requires another important hypothesis. Since the analysis is developed in relative and not absolute terms, between different cities or between core and ring areas, it is assumed that for each relative dimension (large vs. small cities, ring vs. core), the supply curve of houses has the same slope. If this were not the case, a shift upwards of the demand curve, generated by a higher appreciation of location advantages, would give rise to a different increase in prices. This hypothesis does not limit too much the comparison between large and small cities, but could give a heavy bias in a comparison between core and ring areas because of the different potential for the expansion of residential supply in the two areas.

10 Interestingly enough, when a measure for borrowed size is inserted in our estimated model, multicollinearity exists among borrowed size from one side, and functions and networks on the other.

11 A full list of cities surveyed in this paper is available in the appendix.

12 Clearly, this indicator represents only a subset of all possible ways of transnational networking for the surveyed urban areas, and in particular refers to a scientific type of connectivity (thus reflecting an innovation-oriented type of knowledge being exchanged via these connections). However, this is perfectly in line with the scope of this analysis, and fits much better the aims of the empirical work here presented, rather than the use of more classical (and geographically based) networks of physical accessibility.

13 The bar notation indicates a vector.

14 The non-interacted variables “networks” and “functions” have not entered this equation; lacking degrees of freedom, we preferred to keep the specification that is conceptually the most interesting for us.

15 Networks turned out to be insignificant, and for this reason they are not treated in this part. Their role will be highlighted when a diachronic analysis is developed. See Section 5.3.