Does trade creation by social and business networks hold in services?

Abstract

Recent literature on the border effect has shown that the intensity of trade in goods is positively correlated with the migration stocks between any pair of countries/regions. The number of references for business networks is more limited, but they are also related with a reduction on information costs. In this article we investigate whether such a relation holds also for Spanish domestic trade flows in consumer services. To this end, we use a gravity model rooted in the Dixit–Stiglitz–Krugman theoretical framework and a unique data set on interregional trade flows for some of the main tourism service sectors, namely, accommodations and restaurants. Our industry-specific analysis finds a large positive effect for restaurants but a more limited effect for accommodations. Forces driving the demand in each sector explain this result and suggest the idea that although social networks can act as a substitute of firms in some sectors at the same time they can enhance trade flows in other sectors. We perform the same kind of analysis with a data set (obtained by a similar method) for domestic trade in goods and discover a different response to social and business networks. Finally, we treat the potential endogeneity by taking the instrumental-variable approach of the Poisson pseudo-maximum likelihood estimator and thus obtain consistent results.

Keywords:

• gravity model
• social and business networks
• internal migration
• services trade
• bilateral exports

JEL Classification:

• F12
• F14
• L14
• L83
• R23

Acknowledgements

I am grateful to Aitor Garmendia for kindly supplying the business-network variables, to Pierre-Philippe Combes and Miren Lafourcade for their very helpful comments, and to Andrés Rodríguez-Pose and the London School of Economics, where the first draft of this article was written during a short-term visit to the Department of Geography and Environment. This article has been developed in the context of various research projects: the C-intereg Project (www.c-intereg.es); the TransporTrade S2007/HUM/497 (www.uam.es/transportrade), funded by the Education Department of the Madrid Regional Government; the Project (ECO2010-21643/ECON), funded by the Spanish Ministry of Science and Innovation; and the DESTINO Project (Ministerio de Fomento). All errors that remain, are, of course, our responsibility.

Funding

I also gratefully acknowledge the financial support of the Universidad Autónoma de Madrid (Programa FPI-UAM).

Notes

¹ One of the arguments uses Jensen’s inequality – E(ln y) ≠ ln E (y) – to show that the standard interpretation of the parameters of log-linearized models as elasticities can be misleading in the presence of heteroscedasticity.

² In a previous analysis, immigration and emigration stocks and the business network variable were included first separately and then simultaneously in the regression analysis. We have carried out the same analysis with the addition of origin-time fixed effects and destination-time fixed effects. The results are basically the same and available upon request. The results will be available upon request.