本文以200 年及2008年北部地區9 筆交易案例，由估價人員逐筆以成本法估計之成本價格，與成交價格比較其差異。發現成本價格經統計檢定分析，皆顯著高於實際成交價格。推論成本價格有相對較高之系統性偏誤現象，與以往文獻認為成本價格可作為估計值上限之論述相同。此外，屋齡老舊的確會使估價偏誤提高；惟不同分量點之效果不同。本文實證影響成本法偏誤之變數，發現單位地價愈低、總樓層數愈高、移轉樓層愈低、建物面積愈小、建築工期愈長及利潤率愈小者；偏誤愈小。結構、房屋類型、分區、區域之不同對偏誤亦有影響。經由分量迴歸分析，高低偏誤之影響因素亦存在差異。營造施工費求取方式方面，應用工程造價比較法較單位面積比較法偏誤小。 This paper compares the sale price and cost value, as appraised by valuers applying the cost approach for a sample of 974 transactions for properties sold in 2007 and 2008 in Northern Taiwan.We discovered that the sale price is on average lower than the cost value. The reliability of thecost approach was comparatively questionable due to the systematic bias of the overestimation.The previous literature claimed that cost value could serve as the upper limit for value conclusion and that the cost approach was unreliable. Moreover, the bias associated with cost value increased as buildings aged and this bias differed for different quantiles. This paper demonstrated the bias influencing variables based on the cost approach. We found that the bias will be smaller when the land unit value is smaller, the building is higher, the story is lower, the area is smaller, the period of construction of the property is longer and the profit rate is smaller. The bias was affected by structure, building type, zoning, and region. There were differences between the high and low bias on some factors in regard to the quantile regression. The bias was greater based on the construction items comparison method compared to the unit square method when estimating the construction fees for properties.