Small sample evidence on the tourism-led growth hypothesis in Lebanon

Abstract

This paper investigates the contribution of tourism to economic growth in Lebanon for the time period of 1995–2010. The presence of long-run and causal relationships is investigated applying the bounds testing approach to cointegration and Granger causality tests. Because of the small sample (T = 16), econometric approaches and critical values used for testing receive special attention. Additionally, a number of diagnostic tests are utilised to ensure that the model is suitable and correct. Interestingly, our results reveal that tourism and economic growth are cointegrated. The Granger causality test indicates that the tourism-led growth hypothesis is valid for Lebanon. Therefore, policy initiatives promoting tourism ought to be further developed and implemented to stimulate economic growth and development for the economy of Lebanon.

Keywords:

• causality
• tourism
• growth
• Lebanon
• small sample

Acknowledgements

The authors thank the three anonymous reviewers for their valuable comments and suggestions to the earlier draft of this research. Any shortcomings that remain in the paper are solely our responsibility.

Notes

1. It is interesting to point out here that international tourism is a source of export earning, but tourism is different from commodity exports in the sense that the consumer (or the visitor) must consume the product in the exporting (or the visiting) countries. Therefore, the tourism sector has important implications for other sectors in the economy as illustrated by the computable general equilibrium analysis stressed in Blake, Gillham, and Sinclair (2006).

2. The diagnostic tests indicate that the residuals of the ARDL models are normally distributed, free from serial correlation and without the presence of ARCH. In addition, the Ramsey RESET (regression equation specification error test) tests indicate that both ARDL models are free from specification errors. The cumulative sum (CUSUM) and cumulative sum of squares (CUSUMSQ) tests for parameter stability demonstrate that the estimated parameters of the ARDL models are stable at the 5% significance level.

3. It could be argued that there are statistical techniques to increase the number of observations by increasing the frequency of the data from annual to quarterly. However, Hakkio and Rush (1991) articulated cointegration as a concept of a long-run relationship; thus the cointegration results may not change by merely changing the frequency of the data. Moreover, Tang (2008) affirms the idea that interpolated data do not enhance the statistical power of a test. Therefore, the original data will speak better than interpolated data.

4. We appreciate an anonymous reviewer's comment noting that the meaning of Granger causality is different from the meaning of causality in the theoretical sense. Granger causality only shows the predictability behaviour of one variable based on the past values of another group of variables. Hence, Masih and Masih (1998) claimed that Granger causality is actually a predictability test. On the other hand, causality in a theoretical sense is difficult to examine through statistical approaches because the statistical relationship itself cannot logically imply the meaning of causation. Furthermore, Kendall and Stuart (1979) revealed that the statistical relationship is strong and suggestive, but it can never show the direction of causality because a causal relationship must be supported by theories, logics and/or universal laws (see also Hoover, 2001).

5. Because of the finite sample utilised by this study, we additionally performed the Toda-Yamamoto-Dolado-Lütkepohl (TYDL) Granger causality tests proposed by Toda and Yamamoto (1995) and Dolado and Lütkepohl (1996) in association with the leveraged bootstrap critical values to confirm the causal relationship between tourism and economic growth. On the basis of the Monte Carlo experiment, Mantalos (2000) and Hacker and Hatemi-J (2006) found that the bootstrap tests improved the robustness of the causality test, particularly for a small sample. Similar to the conclusion provided in Table 4, the leveraged bootstrap TYDL causality tests indicate that tourism and economic growth in Lebanon are bi-directional Granger causality in nature. Hence, the Granger causality results given in Table 4 are valid. To conserve space, the entire results of the leveraged bootstrap TYDL causality tests are not reported here, but are available upon request from the authors.

6. To confirm the robustness of the estimation results, we re-tested the cointegration and Granger causality results by dropping the first observation (using only 1996–2010) as suggested by an anonymous reviewer. Dropping the first observation (1995), we find that the conclusions for cointegration and Granger causality remain unchanged. Hence, we affirm the robustness of the results. To conserve space, the full results are not reported here, but are available upon request from the authors.