The Spatial Distribution of the Internet in the European Union: Does Geographical Proximity Matter?

Abstract

This paper examines the spatial distribution of the Internet in the European regions. To achieve this aim, our analysis combines a set of non-parametric techniques proposed in the context of the economic growth literature, with various spatial econometric instruments. The results reveal that regional disparities in Internet adoption are greater than territorial inequalities in gross domestic product (GDP) per capita. In addition, our findings show that the distribution under consideration is characterized by the presence of positive spatial dependence, which implies that physically adjacent regions register a similar degree of Internet adoption. Finally, the analysis carried out allows us to assess the role played by variables such as GDP per capita, unemployment rate, stock of human capital and population density, in explaining the spatial distribution of the Internet in the European Union.

Acknowledgments

The authors would like to thank two anonymous referees for their helpful comments and suggestions as well as D. Antonio Vazquez Barquero for his comments. In addition, the second author wishes to acknowledge the financial support from the Spanish MEC (Projects SEJ 2005-08738-C02-01).

Notes

1. The digital divide refers to “the gap between individuals, households, businesses and geographic areas at different socio-economic levels with regard to their opportunities to access information and communication technologies and to their use for a wide variety of activities” (OECD, 2001, p. 5).

2. Milicevic and Gareis (2003), in the context of the BISER project, consider exclusively 28 European regions. Demunter (2005), using national figures, distinguishes only between Objective 1 regions and Non-Objective 1 regions.

3. A review of the evidence regarding the role of different types of proximity as well as some empirical works on this issue can be found in Boschma (2005a) and in Regional Studies, volume 39, February 2005, respectively.

4. NUTS is the French acronym for “Nomenclature of Territorial Units for Statistics”, a hierarchical classification of sub-national spatial units established by Eurostat. In this classification, NUTS-0 corresponds to country level, and increasing numbers indicate higher levels of sub-national disaggregation.

5. In any event, when interpreting the conclusions of our empirical analysis, it should be noted that the level of territorial disaggregation considered in the study may affect the results. Nevertheless, this occurs in any analysis based on spatial data drawn from different geographical units, and it has to do with the “modifiable areal unit problem (MAUP)”, well known by geographers (Arbia, 1989; Openshaw & Alvanides, 1999).

6. To check the robustness of our findings, we use various different spatial weight matrices. In particular, we construct two additional matrices W based on the five and 15 nearest neighbours, which in all cases yield results similar to those discussed later.

7. Inference is based on the normal approximation (Cliff & Ord, 1981). In this respect, it is worth noting that when the Wald normality test is applied to the variable under consideration, we obtain a statistic of 2.768, with an associated probability value of 0.251.

8. The data on regional GDP used were drawn from Eurostat and were referred to 2002, since Eurostat does not provide more recent information.

9. The data on the set of explanatory variables were drawn from Eurostat and Cambridge Econometrics.

10. It should be pointed out that, as an alternative to the variable used so far to control for existing differences in regional demographic structures, we considered the percentage of population aged 70 and over. Nevertheless, the coefficient corresponding to this new variable continues to be non-significant when we estimate this new version of equation (5). The remaining results, however, are in both cases very similar.

11. Note that we do not include the virtual distribution corresponding to the case where there were no regional differences in terms of population and population aged 55 and over, since the coefficient associated with these variables did not prove statistically significant in the preceding analysis (Table 1).