治山防洪工程成本分析-以台大實驗林為例

Abstract

本論文的主要研究目的在於分析治山防洪工程成本的影響因素。首先，本研究蒐集2001年至2005年間台大實驗林執行的治山防洪工程資料，接續藉由文獻回顧與成本理論建構成本模型。為減緩資料中異質變異與離群值的問題，本研究採以加權最小平方法及穩健迴歸方法估計成本模型，之後再以拔靴法評估成本模型的預測精確性。 實證結果顯示，治山防洪成本與施工規模、施工方法、施工區位、及施工年度顯著相關。成本模式明確顯示施工規模為最重要的因素，而施工規模達到某一面積時，工程將享有明顯的規模經濟報酬。此外，除施工規模外，加入其它因素能改善成本估計的準確水準。若施工地點遭受嚴重的自然災害，施工成本顯著升高，且成本與施工區位密切相關。研究顯示溪頭與和社營林區有較高的工程成本。最後，拔靴法模擬的結果顯示，本研究所建構的成本模型具有良好的預測精確性。

The main purpose of this thesis is to analyze affecting factors of the cost of watershed conservation and flood control projects. First of all, we collected data of the projects carried out in National Taiwan University experimental forest from 2001 to 2005, then set up a cost model by reviewing literatures and cost theory. To alleviate problems of heteroskedasticity and outliers embedded in the data, we applied weighted least squares (WLS) and robust approaches to estimate the cost model. We then assessed prediction accuracy of the cost model with bootstrapping technique. The result of the empirical study showed that the cost of watershed conservation was significantly related to construction scale, the type of construction, the location and time period of projects implemented. Obviously, the cost model indicated that scale to be the most important factor; a project had significant increasing return to scale if its scale over a certain acreage. Furthermore, the additional factors other than scale could improve the level of accuracy in cost estimation. The cost increased significantly if the project site damaged severely by natural disasters, and related positively to the location of project. It showed that the Chi-Tou and Ho-She districts had higher project cost. Finally, the result of bootstrapping approach suggested that the cost model developed by this study had good accuracy in prediction.