Tourism-Led Growth Hypothesis in Malaysia: Evidence Based Upon Regime Shift Cointegration and Time-Varying Granger Causality Techniques

Abstract

This article examines the tourism-led growth (TLG) hypothesis in Malaysia based upon quarterly data from 1991:Q1 to 2014:Q1. The Johansen–Juselius cointegration and the regime shift cointegration tests consistently show evidence of cointegration. In addition, we find evidence of unidirectional causality from tourism to economic growth in Malaysia. Furthermore, the rolling Granger causality test confirms that the TLG hypothesis is generally valid and stable in Malaysia. Therefore, tourism is an effective long-term engine of growth. Policies to promote tourism would effectively invigorate Malaysia's long-term economic growth and development in Malaysia.

Key words:

• cointegration
• tourism-led growth hypothesis
• rolling regression
• Malaysia

JEL Classification:

• C32
• O11

Acknowledgements

We would like to thank the two anonymous reviewers for the insightful comments and suggestions on the earlier draft of this research paper. We would also like to acknowledge Kee-Cheok Cheong and Zarinah Yusof for their valuable comments on an earlier draft of this research paper. Our appreciation also goes to Bruce Hansen and Hatemi-J Abdulnasser for sharing their GAUSS programming codes. However, any remaining shortcomings are the sole responsibility of the authors.

Notes

1 According to a comprehensive review of Stock and Watson (2001), VAR modelling is one of the mostly used analytical tools in applied macroeconomics research. One of the advantages of using VAR modelling is that it is simple and flexible compared with the instrumental variables technique and/or the two-stage least-squares estimator in simultaneous equation modelling (SEM). In a VAR model, all variables will be treated as endogenous and the contemporaneous terms will not be included in the right-hand side (RHS) of the equations (Sims, 1980). Therefore, all RHS variables are pre-determined (i.e. lags of endogenous and exogenous variables). This reflects that VAR modelling is a dynamic version of SEM (Maddala, 2001; Stock & Watson, 2001). Hence, the ordinary least squares (OLS) estimator can be applied to a VAR model because the latter is not subject to simultaneity bias.

2 To conserve space, the conventional ADF and KPSS unit root tests results are not reported here, but are available upon request.

3 According to Johansen (1992), the procedure of Pantula principle begins by testing for cointegration with Model 2, then moving to Model 3, and then to Model 4. The process will continue until the first time the null hypothesis cannot be rejected. Based upon the cointegration results presented in Panel A of Table 5, we find that at the 5% significance level, the first time the null hypothesis cannot be rejected is located at Model 3 and there is one cointegrating vector among variables in the system.