Does tourism influence economic growth? A dynamic panel data approach

Abstract

On average, tourism-specialized countries grow more than others. This is not consistent with the core of modern economic growth theory that suggests that economic growth is linked to sectors with high-tech intensity and large scale. In this article, we use appropriate panel data methods to study the relationship between tourism and economic growth. In general, we show that tourism is a positive determinant of economic growth both in a broad sample of countries and in a sample of poor countries. However, contrary to previous contributions, tourism is not more relevant in small countries than in a general sample.

^† This is a fully modified version of a working-paper co-authored by one of the authors.

Acknowledgements

We are indebted to participants at the ‘Second International Conference on Tourism and Sustainable Development – Macro and Micro Issues’, supported by the World Bank, Cagliari, Italy. One of the authors also gratefully acknowledge a financial support from the FCT, under the Project POCTI/EGE/60845/2004. The usual disclaimer applies.

Notes

^† This is a fully modified version of a working-paper co-authored by one of the authors.

¹ The use of panel data techniques in empirical economic growth literature began with Islam (1995).

² This means that we study economic growth empirically through the implementation of a growth regression that evaluates the contribution of different determinants of economic growth. The seminal article from Barro (1991) began this tradition.

³ In the next sub-section, we discuss the composition of the conditioning set and different measures of tourism intensity.

⁴ This estimator is preferable to the difference estimator if the dependent variable is highly persistent, as is the case for output.

⁵ To avoid the over-fitting bias that arises from the high number of instruments in System GMM, we restrict the number of lagged instruments to two by variable. These specification options do not change our conclusions. In Tables A1, A2 and A5 in Appendix A, we also present the number of instruments introduced in each regression. According to these numbers, we must be confident in a quite small ‘overfitting’ bias. However, this would not happen in small samples for which we present results. In face of this, for small samples, we present only the corrected LSDV results.

⁶ We note that ICRG is a more general measure for institutions than BMP because it includes 22 different indicators of country risk. Nevertheless, it is less used in previous empirical contributions. Its introduction comes at the expense of a significant number of observations. It is increasingly used in development (see, e.g. the influential article of Hall and Jones, 1999).

⁷ As in these references, the rate of convergence λ is obtained equating e ^{− λT} to 0.894 (in the case of column 0 in Table A1, Appendix A), where T = 5.

⁸ Although results do not influence our conclusions, they are obviously available upon request.

⁹ Eilat and Einov (2004), in an article on the determinants of international tourism, have discovered that political risk is very important to tourism.

¹⁰ We have also tested a sample of countries with less than 1 million inhabitants. Results showed that tourism specialization has never a positive contribution to economic growth. Results are available upon request.

¹¹ A first attempt to use panel data methods applied to this broad sample led to different results due to the presence of the Nickel (1981) bias in fixed effects estimations (Sequeira and Campos, 2005).