Taiwan is one of the few countries where a split-rate or differential-rate property tax is applied. Under the split-rate taxation, land is taxed at a higher rate than the structure. The scarce transactions of vacant land in the cities have led the government assessors to separating land and structure values of a built property. The present valuation practices in separating these two value elements heavily rely on the cost approach. However, the accuracy of the assessed land values has been largely eschewed examination. This article employs a Cobb-Douglas type regression model to estimate the respective values of land and structure against a data set of 17469 transactions of residential properties in Taipei City between 1994 to 2003. A variable that accounts for the spatial effects on property value is constructed through the trend-surface method instead of employment of the often-used district dummy variables. We contrast the government-assessed and market-revealed land values primarily from the equity perspective. A government-assessed land value used for taxation purpose is generally acceptable to differ from its true market value. However, dissatisfaction, even legal disputes, often arises when the differences between government-assessed and market-revealed values vary significantly among properties. The differences will violate the equity principle of taxation. We also attempt to provide explanations for the detected value differences in the hope to improve the present valuation system.