Essays on the Inter-Governmental Interdependence of Tax Policy

This dissertation consists of the following three chapters. In the first chapter, the interrelationship between the capital tax policies of local governments is empirically investigated using a data set from Japan, which is constructed in the second chapter of this dissertation. Reaction functions of local governments, which relate their capital tax rates to those in competing governments and their characteristics, are estimated through both the instrumental variable method and maximum likelihood estimation. The results suggest that a positive relationship exists among local governments. In the second chapter, using established methods, two types of original panel data sets on the effective tax burden of 47 prefectures in Japan are constructed. One is an average tax rate (ATR), which is based on the method of Mendoza et al. [Mendoza1994], and the other is an average marginal tax rate (AMTR), which is based on the method of Joines [Joines1981]. Generally speaking, a positive relationship between AMTR and ATR can be observed. The time series of calculated AMTR and ATR are similar, although AMTR is generally larger than ATR due to their definitions. However, cross sectional characteristics are not consistent with each other due to the limitations of regional data for estimating AMTR. In the third chapter, I estimate the gasoline demand function while considering spatial correlation. A recent work estimates the price elasticity of gasoline consumption without spatial error term, but it has been shown that the hypothesis of no spatially autocorrelated error term is rejected by use of the same data set. Furthermore, when the gasoline demand function is estimated by the spatial error autocorrelation model, the price elasticity of gasoline consumption is estimated to be -1.186. This figure is larger than the figure obtained from a model without spatial error.