A Study on the Influence Factor of Auction House Price in Tainan

[[abstract]]本研究乃以分量迴歸方法建立住宅大量估價模型，藉以瞭解住宅特徵對於不同價格分量的差異，透過分量的統計可以對於不同類群的投資者提供相對於僅有平均數概念的OLS更精細的評斷準則，投資在低價部位與高價部位之族群不同，投資者的風險趨避程度與風險認知不同，應有不同的分量估計數值。比較分量迴歸與最小平方迴歸的係數，發現在0 1分量（64萬元）低總價住宅對坪數反應較大，13樓以上對於各價位都有明顯溢價效果，區位變數對於不同分量的影響亦有差異。若以最小平方迴歸模型估計時，會低估變數對價格的影響。由於本研究樣本資料為右偏，實證結果亦發現中高價位的影響變數較有明顯差異，分量迴歸模型的確改進兩尾端估計精確度。透過分量迴歸模型確實可以檢測各跨分量間的特徵係數差異，亦可做為估價師在估價時價格調整之依據。[[abstract]]Quantile regression method is employed in this research to establish a large number of residential valuation models and to understand differences in housing characteristics for the components of different prices Through the statistics of components it can provide more precise judgment guideline compared to OLS only has the concept of average to different groups of investors Different groups invest in low-priced position and high-priced position investors’ degrees of risk aversion and risk perception can be defined different; there shall have different estimated values of components Compare to the coefficients of quantile regression and least squares regression it found that the reaction of dwelling of low total price to Ping (ground measurement) is bigger at 0 1 component ($ 640 000) and above 13 floors for each price has an obvious effect of premium There are differences in effects of location variables to the different components If estimate by model of least squares regression it will underestimate the influence on variables to prices Due to this research sample data is skewed to the right the empirical results also found that the effect variables of middle and high price have significant differences and the quantile regression model is indeed improved the estimated accuracy of two ends Through quantile regression model can indeed detect the differences in characteristic coefficients between each cross-componenets and can as a basis for appraiser on the valuation of the price adjustmen