This study aimed to determine whether the simulated marginal tax rate discussed in Shevlin (1990) and Graham (1996) is the most superior proxy for the marginal tax rate, using Korean tax regulations and tax rate data, and to identify which variables can serve as proxies for the marginal tax rate when the simulated marginal tax rate is unavailable. The research results, based on a sample period of five years from 1989 to 1993, showed that the simulated marginal tax rate is the most superior proxy for the marginal tax rate, consistent with the findings of Graham (1996). In the analysis using the perfect-foresight marginal tax rate as the benchmark, the simulated marginal tax rate demonstrated superior predictive ability compared to other proxies. Furthermore, when the simulated marginal tax rate is unavailable, the statutory tax rate was found to be the closest proxy to the perfect-foresight marginal tax rate, followed by the trichotomous variable and the taxable income dummy as usable proxies for the marginal tax rate. In contrast, the net operating loss carryforward dummy and the Manzon effective tax rate variable exhibited low explanatory power for the perfect-foresight marginal tax rate, and were thus analyzed as inappropriate proxies for the marginal tax rate.