ABSTRACT
A growing body of literature investigates the bi-univocal relationship between tourism and economic growth. Even if sub-national studies are deemed relevant, these are rare: we provide a first analysis of such relationship for Japanese regions and prefectures, using Granger Causality tests in a Bayesian VAR model, from 2007 to 2014. Both the tourism-led growth and the economic-led tourism hypotheses are supported in 4 out of 8 regions and in 19 out of 47 prefectures (either univocally and/or bi-univocally): it is possible to use tourism as a policy instrument to stimulate economic growth, even if regional discrepancies are to be expected.
Acknowledgements
We would like to show our gratitude to Chiba Manami, Hirai Marina, Miura Saki and Shimizu Kenta for their support during the data collection phase from the System of National Account (SNA) of Japan. A special thanks to Luca Corazzini, Luigi Vena, Andrea Venegoni and Simone Zardi for their more than helpful comments.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 We use only 1 lag due to the small sample of regional/prefectural data, as we cannot estimate longer lag models and reliably test them, without first estimating a BVAR
2 For each prefecture, we conducted a KPSS test on the two series to assess their stationarity; 103 out of 108 proved showed a p-value higher than 10%, 3 prefectures showed a p-value lower than 10% in the GPP series (0.99 for Yamaguchi, 0.088 for Kagawa and 0.085 for Shimane) and 2 prefectures showed a p-value lower than 5% in the Overnight series (0.0483 for Iwate and 0.0401 for Aichi). All of these tests fail to reject the H0 of the KPSS test, i.e. our series are all trend stationary.
3 To assess the robustness of our results, we estimated more comprehensive B-VARs, including Japanese CPI and the Effective Exchange Rate (Japanese Yen versus a weighted basket of foreign currencies), with non-informative Minnesota priors, due to impossibility of estimating informative ones as in the main model. These results, available upon requests, largely confirm the one presented here, with only 8 prefectures and 2 regions showing a difference in the significance of estimated GC tests.