A nonlinear approach to the tourism-led growth hypothesis: the case of the MERCOSUR

Abstract

This paper analyses the tourism-led growth hypothesis for the four countries of the MERCOSUR regional trade block. By applying nonlinear techniques, we explore whether tourism activity leads – in the long run – to economic growth, or, alternatively, economic expansion drives tourism growth, or indeed a bidirectional relationship exists between the two variables. To this end, non-parametric cointegration and causality tests are applied to quarterly data for the period 1990–2011. We show the existence of a cointegrated relationship between real per capita gross domestic product and tourism expenditure for all countries. Moreover, the linearity of this relation is rejected for both Argentina and Brazil (economies with a relatively diversified structure). The non-parametric causality tests confirm in all cases the causality from tourism to growth. Meanwhile, only for Uruguay and Argentina causality also goes in the inverse direction (from growth to tourism). Finally, the paper compares the results of the nonlinear approach with those obtained by using the traditional linear methodology.

Keywords:

• tourism-led growth hypothesis
• nonlinear cointegration
• non-parametric causality tests
• MERCOSUR

JEL Codes:

• C30
• E43
• L83

Notes

1. The degree of tourism specialization is defined by these authors as the receipts from international tourism as a percentage of GDP.

2. This is the latest available Balance of Payment data according to the Central Bank of Paraguay.

3. This is the only current estimation of Tourism Satellite Accounts for the MERCOSUR countries, as far as we know.

4. WTTC, Travel & tourism. Economic impact 2013, country reports.

5. The regional real exchange rate includes Argentina and Brazil.

6. The ADF test null hypothesis considers that the process has a unit root; this hypothesis is accepted unless there is strong evidence against it. On the contrary, the KPSS test has the null hypothesis of stationary process, complementing the ADF test which has low power against stationary when processes are near the unit root. Thus, a stationary process rejects the null hypothesis of the ADF test, while it is accepted by applying the KPSS.

7. We consider two degree of freedom because the score test is applied using three variables.

8. An excellent introduction to cointegration analysis and its interpretation can be found in Hendry and Juselius (2000, 2001).