Network infrastructure spillover in private productive sectors: evidence from Spanish high capacity roads

Abstract

Interest in public infrastructure research has been a subject of increasing concern to economists and policy-makers. This article aims at analysing the locational impact of the high capacity roads (HCR) on the Spanish private economic activity from 1970 through 1998, given that this is one of main infrastructure-based development strategy undertaken in Spain in this period. In a stochastic frontier production-function framework, we allow for modelling of provincial heterogeneity through the existence of different efficiency levels in the territorial units. Results show HCR spillovers between geographically close provinces and between provinces displaying similar socio-demographic characteristics and government size. To the extent that the magnitude and sign of these impacts differ across sectors a reasonable explanation of the limited impact of HCR on aggregate private production could be made. The presence of a negative spillover effect in the industrial and business service sector suggests that Spanish provinces may have used HCR capital as a competitive tool for attracting factors of production leading to a rearrangement in these economic activities.

Acknowledgements

This research has been carried out with the financial support of the Spanish Ministry of Public Works (BOE-23 December 2004, BOE-22 November 2002). This support is gratefully acknowledged. We wish to thank Federico Pablo Martí for his helpful support with mapping software. An anonymous referee provided valuable comments on an earlier version of the article.

Notes

¹ See De la Fuente (2000) and Mikelbank and Jackson (2000) for revisions of state of public capital research.

² See Fraumeni (1999) for a description of capital stock measures and revision of studies.

³ After Spain's entry into the EU, the European institutions adopted a series of measures to achieve real integration of peripheral areas within the Community. Spain benefited from growing funds assigned to finance infrastructure projects in less-developed regions in order to promote growth and hence cohesion within EU territories.

⁴ Differences in the scope of the areas under consideration may play a role in the different results obtained in literature. It is argued that the larger the region, the more masking will occur of negative effects found in more disaggregated analyses.

⁵ Most of the Spanish evidence on spillover has been obtained from regional studies, NUTS II level (Avilés et al ., 2003; Pereira and Roca-Sagalés, 2003; Cantos et al., 2005). Studies carried out with provincial data were on specific sectors. This is the case of Moreno and López Bazo (2007), which involved the industrial sector.

⁶ The difference between the high capacity network (motorways, highways and dual-lane roads) and other roads is relevant in an analysis of spillover effects, given that their main purpose is to connect different provinces.

⁷ See Murillo-Zamorano (2004) for a survey of efficiency frontiers.

⁸ See Cuesta (2001) for a revision of these models.

⁹ These arguments have been usually presented in the literature. See for example Martín and Suárez (2000).

¹⁰ It is also possible to obtain indexes to compute the predictions of technical efficiency based on the assumptions for the model. These indexes are not shown in this study, as they were not subject to separate analysis.

¹¹ The log-likelihood function of this model is presented in the appendix (Battese and Coelli, 1993). Estimation of the likelihood function also requires the specification of a relationship between variance parameters such as  =  + σ² and γ = σ²/. The contrast of the relative importance of technical inefficiency effects relating to error in specifying frontiers is carried out by means of parameter γ. A value of γ equal to zero means that the frontier deviations are exclusively due to specification error effects, it thereby being pointless to include them in the estimate of factors explaining technical inefficiency.

¹² Some studies have used a translog function, which differs from the Cobb–Douglas function in that it does not imply various restrictions on the production structure. The substitution elasticities of the production factor are, for instance, by definition equal to one in a Cobb–Douglas production function.

¹³ Several recently developed models such as Greene (2005, 2004) have addressed this issue.

¹⁴ Instead of this proposal, a practical alternative to modelling spillover effect consists of building a sole variable which is added to public capital and includes the region's public capital together with the capital of remaining regions. This modelling of the spillover effect has been applied to Spain in Cantos et al. (2005) and Mas et al. (1994). Nevertheless, Álvarez et al. (2003) consider it a problematic option as it requires the marginal productivity of an additional public capital unit within the area studied to be the same as in the others.

¹⁵ The symmetric character of the contiguity and competitive matriz is debatable, because it supposes that the influence that province j receives from province i is the same as that received by i frorm j, whereas the influence between two regions is not always reciprocal in its intensity. In order to asses a sensitive analysis of the spillover effect of HCR, two additional definitions of the W matrix were used: W _ave: The average endowment of HCR in the neighbouring provinces. W _dist: The inverse of the square of the distance bewtween any par of provinces. The results obtained for each one do not vary considerably to the following. It is often highlighted that the spillover effect is transmitted more intensely via commercial channels between different regions, in such a way that the weighting assigned to regions should be directly related to the magnitude of such channels. In the case of Spain, this modelling has been used by Ezcurra et al. (2005) and Avilés et al . (2003), amongst others. These studies employed data provided by the Permanent Survey on Road Transported Goods, carried out by the Ministry of Public Works. This information is only available on a regional basis, which is why it was not used in this research.

¹⁶ Although this source of information allows other forms of public capital also included as inputs (ports, airports, railways, water infrastructures, urban structures, health and education) to be analysed, they have been excluded as the objective of this study is centred exclusively on HCR. Highways are seen as being one of the main public sector investment in much of the theoretical literature on economic growth; see for example, Barro and Sala-i-Martin (1992).

¹⁷ The Ministry of Public Works classifies roads according to ownership (Public Administration network and Toll Network), authority (Network controlled by the State, Autonomous Communities and Provincial and Local Governments) and their physical characteristics (HCR and Other Networks).

¹⁸ In the estimate carried out, calculations on average life, survival functions and depreciation as expressed in the BBVA-Ivie Foundation study are maintained. For road capital, an average life of 60 years up to 1965 is taken and 40 years as of 1966. In calculating capital, the Winfrey S-3 survival function was used, as is often the case in this type of study.

¹⁹ This is the most comprehensive database available on Spanish net public capital. In order to prepare the database, the OECD method (OECD, 1993) was used. Review of this method (OECD, 2001) distinguishes between capital wealth, comparable to net capital estimated by the FBBVA, and productive capital, associated to the value of capital services and considered relevant for productivity analyses. The BBVA-Ivie Foundation is currently preparing new data series based on the new methodology, however it was not avavailable while this research was undertaken.

²⁰ According to Battese and Coelli (1995), maximum likelihood estimation is performed using FRONTIER 4.1 (Coelli, 1996).

²¹ This is a common line of explanation to justify the existence of spillover effects: studies performed on a lower than national scale (states, regions or metropolitan areas) obtain infrastructure elasticity lower than those at national levels. Recent empirical studies study also the importance of the spillover effects across countries (Owyong and Thangavelu, 2001).