Service industry and cumulative growth in the regions of Europe

Abstract

European regions have experienced a greater presence of service producers in their economy over the last few decades. Indeed, the manufacturing sector increasingly contracts out many activities to intermediate producer services. This is mostly because they are located close to each other and because services experience increasing returns to scale which reduce their marginal costs. In this paper, we propose to measure the extent to which productivity in services has converged across European regions. The model we use, originally developed by Verdoorn (1949), takes the increasing returns to scale explicitly into account. We apply spatial econometric techniques and control for border effects by introducing two different spatial weights matrices under the assumption that economic interactions decrease very substantially when a national border is passed. Furthermore, we take proper care of the presence of both types (spatial and non-spatial) of endogeneity by using spatial two stages least squares (Kelejian and Prucha 1998). Our conclusions bring new insights in the identification of regional productivity differentials.

Keywords:

• regional convergence
• Europe
• services
• spatial econometrics

Notes

Notes

1. See Schettkat and Yocarini (2003) for an extensive literature review.

2. Generally, the producer services are defined as including a combination of banking and finance, insurance, real estate, engineering, architecture, accounting, marketing, information services, and legal services (Goe 1990).

3. McCombie and Thirlwall (1994) also put Verdoorn's Law at the centre of the well-known model of cumulative causation growth which should be considered as the base for successive models à la Krugman (1991).

4. As a referee correctly pointed out, an alternative way to define the model would be to use dynamic panel data specification with simultaneous equations. However, using panel data, where the growth rates should reflect variations over (at most) five years, would fail to capture the long run dynamics. As a consequence, the interpretation of estimated coefficients would be different and would not correspond to Verdoorn's Law. A more detailed explanation is provided in McCombie and Roberts (2007) who demonstrate the substantial fallacy of time-series and panel models in estimating Verdoorn's Law as a long-run relationship.

5. For a review of the different approaches to the estimation of Verdoorn's Law, see Leon-Ledesma (2000).

6. For a detailed explanation of the Spatial Two Stages Least Squares procedure and the proper choice of instruments we refer the interested reader to Kelejian and Prucha (1998, 1999).

7. There is no formal proof in the literature that the extension of the S2SLS procedure to the case of a model with two weights matrices leads to a consistent estimator of the spatial autoregressive parameters.

8. We prefer working with statistical regions than functional regions for the following reasons: (1) homogenous data are not available at a smaller spatial scale than NUTS-3 regions; (2) the boundaries of functional regions may change over time; (3) the definition of functional regions is not unanimous (for instance, Andersson 2006, uses accessibility while Cheshire and Hay 1989, use employment and commuting); and (4) statistical regions are still the official definition of space according to the European Commission reports.

9. We are aware that our empirical results could be affected by the modifiable area unit problem (Openshaw and Taylor 1979; Arbia 1986, 1989; Anselin and Cho 2000). However, the choice of this disaggregation level is driven by the preference of the European Commission while assessing convergence in its official reports.

10. We refer to internal spillovers to indicate the spillovers deriving from neighbours within the same state. Consistently, we define external spillovers (for each region) those deriving from the spatial relationships with regions belonging to different states.

11. We cannot calculate any test for residual spatial autocorrelation or for spatial heteroskedasticity simply because there is no evidence in the literature on how the tests should be performed in a model with two spatial lags like the one we are specifying.

12. We use a weight matrix based on five neighbours. Note that we are not adopting the same distinctions as before so that the five nearest neighbours can either be all from the same or from a different country (and any combination of the two extreme possibilities). Results are confirmed with other weight matrices. Complete results are available from the authors upon request.

13. This phenomenon is what Barro and Sala-i-Martin (1995) define as the catch-up effect.