조세곡선의 누진성에 대한 통계적 추론

The Lorenz curve had been used essentially as a descriptive device and not as a tool for formal statistical inference to describe and compare income inequality due to the mathematical complexity of its sampling distribution. Beach and Davidson (1983) derived the asymptotic joint variance-covariance structure for Lorenz curve ordinates and income shares, and suggested statistical inference procedures with sample data to compare alternative Lorenz curve ordinates and income shares. Bishop et al. (1989) extended the statistical framework to the generalized Lorenz curves, while Bishop et al. (1994) and Klavus (2001) applied it to the relationship between the Lorenz and associated concentration curves. By applying the methodologies into the tax curves, Iyer and Seetharaman(1997) developed inference procedures to test the progressivity of tax systems. The asymptotic distribution of sample Lorenz curve ordinates is multivariate normal whose ordinates are determined by cumulative population shares, the conditional means and unconditional overall mean of incomes. Like the Lorenz curve, the ordinates of the tax curve are determined by the cumulative population shares, the conditional and unconditional means of taxes. So, flawed is Iyer and Seetharaman(1997)'s suggestion to use the cumulative income shares instead of the cumulative population shares for estimating the ordinates and variance-covariance structure of the tax curve. This paper applied the revised inference procedures of Iyer and Seetharaman (1997) to evaluate statistically the progressivity or equity issue of Korean corporate income taxes using listed companies' corporate income and tax data between 2001-2005. The procedures involved comparing estimated tax curve ordinates and docile tax shares to hypothesized ordinates and shares, and comparing two independent ordinates and shares. The empirical test results show that no inference results were found statistically significant. The conclusions inferred from the results are that Korean corporate tax system is actually not progressive but proportional and there are no significant progressivity changes during the test periods. Statistical inference on the Suits index showed only the largest firms are taxed regressively.