Cigarette tax revenues and tobacco control in Japan

Abstract

The hypotheses of non-addiction, myopia and rational addiction are tested using annual, quarterly and monthly data. Changes in the prices of Japanese cigarettes can be viewed as exogenous from the point of view of consumer behaviour, because the Japanese government controls cigarette prices. The empirical results of this paper support the addiction hypothesis. The short-run and long-run price elasticities range from −0.338 to −0.421, and from −0.679 to −0.686, respectively; thus, increases in tax revenues in the long run are likely to be smaller than those in the short run. As a result, tax increases would be an effective means of curbing smoking and reducing its social cost. Furthermore, the debt compensation programmes for the Japan Railway and the National Forestry will not go according to plan, unless revenues are increased in the future.

Acknowledgements

This paper is Chapter 2 of the author's doctoral dissertation for Osaka University in December 2004.

The author wishes to thank Kazuo Ogawa for guidance and advice provided while he was writing this paper, as well as Charles Yuji Horioka, Shinsuke Ikeda, Fumio Ohtake, Wataru Suzuki for their valuable advice and comments. In addition, Dariusz Stanko, Midori Wakabayashi, and seminar participants at Osaka University are thanked for their helpful comments. Any remaining errors are the responsibility of the author.

Notes

¹ According to Merriman (2002) and Goel (2004), higher tobacco tax rate may cause cigarette smuggling. Cigarette smuggling in Japan is not analysed here.

² Bask and Melkersson (2004) analyse two addictive goods, alcohol and cigarettes. Here, only cigarette smoking is analysed.

³ The price elasticity of demand for Japanese cigarettes was estimated by Saito (1991). In the book, National Accounts, Saito estimated the price elasticity of demand for cigarettes to be −0.46, using data from the Japanese Household Consumption Expenditure survey. However, some aspects of his analysis left room for improvement, such as parts of the model and the sample, among other things. First, the analytic model used was static, because cigarettes were considered to be non-addictive goods. If smoking addiction were to be considered, the analytic model would have to be dynamic. Second, the sample included only worker households, not all households. Furthermore, the time-series data were for the period from 1954 to 1984; thus, cigarette consumption after 1985 was not analysed.

⁴ For example, Hsieh (1998) analyses this point.

⁵ There seems to be a misprint in BGM (1994). According to my calculations, the last multiplication term in the numerator of the formula for θ ₃ should be u _{2 y}u_ey instead of u _{2 y}u _{2 e} .

⁶ See BGM (1994) for details.

⁷Auld and Grootendorst (2004) point that aggregate data tend to yield spurious evidence in favour of the rational addiction hypothesis. But it is also pointed in their paper that instrumental variable estimates of the coefficients on the lag and lead of consumption are consistent if prices are exogenous.

⁸ See The History of Japanese Tobacco Monopoly (in Japanese, Nihon Tabako Senbaishi).

⁹ The results based on yearly data are reported here. Other results, which are based on quarterly and monthly data, are very similar to those derived from the yearly data and are thus not reported here. The omitted results are available upon request from the author.

¹⁰ Following another paper of the author's, the rates of tax increase are used as proxies for inventory in the estimated equation. The estimates of the coefficients on the proxies are not reported here, though there is a large inventory effect. The omitted results are available upon request from the author.

¹¹ Following Wan (2005), dummies are used as proxies for inventory in the estimated equation. The coefficient estimates for the proxies are not reported here but are available upon request. The inventory effect is large.