Impacts of tourism and fiscal expenditure on remote islands in Japan: a panel data analysis

Abstract

Japan consists of many small inhabited islands in addition to four main islands. We examine the impact of fiscal expenditure and the number of tourists on per capita taxable income in remote islands using panel data analyses. The results show that both fiscal expenditure and population size have significant positive impacts on per capita taxable income, whereas the number of tourists does not have statistically significant impact. They indicate that tourism development would not work as a substitute for financial support from the government. In other words, continuous financial support may be needed to maintain the islands' economies.

Acknowledgements

Financial support from the Japan Society for the Promotion of Science (Grant-in-Aid for Scientific Research A(2) 16203020 and (B) 19330045) is gratefully acknowledged. We are also grateful to Toru Nagahashi for his helpful comments.

Notes

¹ The 6847 ‘islands’ other than the mainland are defined as remote islands in Japan in the Handbook of Remote Islands Development, which is published by the National Institute for Japanese Islands. This says that ‘island’ has to satisfy the following requirements based on the survey conducted by the Hydrographic and Oceanographic Department in 1986: (1) they must have a girth of over 0.1 km; (2) they must have a narrow structure such as bridge or pier if they have a fixed link to the mainland; and (3) they must not be reclamation lands.

² We, hereafter, refer to these three Special Laws and Law for Development of Remote Islands as ‘Remote Islands Development Laws.’

³ According to the Annual Report on the Japanese Economy and Public Finance 2005 published by the Cabinet Office, the total amount of outstanding national and local long-term debts was estimated at 774 trillion yen (the budget deficit in relation to GDP is almost 150%) at the end of 2005.

⁴ The merger of municipalities has been promoted under the Special Law on the Merger of Municipalities, enacted on 29 March 1965 and amended on 26 July 1999. The number of municipalities decreased from 3232 at the end of March 1999 to 1822 at the end of March 2006.

⁵ Most financial aid to remote islands from the central government and prefectures is provided to each municipality that controls remote islands. We therefore equate the fiscal expenditure of each municipality with financial aid to remote islands. It is difficult to calculate expenditure on each island as each has some projects that the central government subsidizes directly.

⁶ These data are from the Handbook of Remote Islands Development, published by the National Institute for Japanese Islands.

⁷ Many studies have been conducted to determine a tourism multiplier, such as Archer (1982), Lundberg et al. (1995) and Vanhove (2005). Balaguer and Cantavella-Jorda (2002) estimated the long-run effect of tourism on economic growth by applying some recent time-series techniques, and Baaijens and Nijkamp (2000) and Baaijens et al. (1998) adopted a meta-analytic approach to empirical research on this topic. We also refer to Murinde and Rarawa (1996) who examined the effectiveness of stabilization policy on the Solomon Islands’ economy using a macroeconomic model, and Johnson and Thomas (1990), who estimated the impact of a major tourist attraction on local employment in the north of England.

⁸ We can set different lag lengths to each variable. In this article, however, we set all the lag lengths equal (M) for simplicity.

⁹ This model is considered a distribution lag model, which is more flexible than the Koyck distributed lag model (Koyck, 1954) given by a partial adjustment model.

¹⁰ All parameter estimates were calculated using TSP 5.0.

¹¹ Recent developments in time-series analysis suggest to test whether the series are stationary or not. According to Levin and Lin (1992), the augmented Dickey–Fuller type test statistic with time dummies is distributed standard normal asymptotically. The estimated test statistic for logarithmically transformed per capita taxable income is –7.83. This result means that we can reject the unit root hypothesis at the 5% level.

¹² If we assume that the error term in the partial adjustment model follows an error component structure with an individual effect, there exists a correlation between the error term and the explanatory variables. In this article, a model can be estimated using the maximum likelihood estimation method because the error term is supposed to be composed of only a time effect. The dynamic panel data model, which is seen in Baltagi (2005) also has an individual effect on the error term. In terms of estimation of tourism demand, Song and Witt (2000) comment on the model and provide some applications. For more information on the random effect model, see Hsiao (2002) or Baltagi (2005).