本研究之主要貢獻為發展適合台灣營造綜合險保單（本體損失險）之風險曲線。 先前由Salzmann (1963)及Ludwig (1998)之相關研究，係採用美國市場之損失經驗，考量不同保單之承保範圍、不同保險標的、及不同之承保風險，直接引用美國市場風險曲線於台灣市場之通用性值得釐清。 本研究回顧台灣營造綜合險保單之承保範圍及除外條款，依據西元一九九六年至兩千年之損失經驗，發現台灣營造綜合險之損率深受天然災害如地震、颱風、豪雨所影響。 再保險安排可分成兩大類，一是比例再保險; 另一是非比例再保險。風險曲線是非比例再保險常用之計價方式之一。 比較先前之研究，本研究之風險曲線分佈集中在低層其主要之原因如下:- 一、 分析標的之可能最大損失(PML)偏低: PML代表某一特定危險所可能造成之可能最大損失。以道路工程為例，其可能最大損失小於30%之保額。由於道路工程大多綿延數十公里甚或更長，其工程本身呈線狀之分佈。換言之，非常不可能會有單一損失其損失金額超過保額之30%。 因此，道路工程之損失分佈不平均且分佈集中在低層。 二、 營造綜合險保期偏長: 營造工程其中一個特性為多年期的保險期間。 因為工程之進行往往需要花超過一年的時間來完成，因此，營造綜合險全損(損失金額等於保額)之機率，遠小於財產險因火災造成全損之機率。 三、 分析之承保風險不同: Salzmann (1963)僅針對火災此一危險因子來從事她的研究。就我們所知，美國大多數住宅皆是木造，屬可燃性建材，火災造成之影響相當大。 相反地，本研究標的大多利用不可燃性建材來興建，如鋼鐵或混凝土，故火災之影響性相對較小。 關鍵字: 非比例再保險、風險基礎計價、營造綜合保險 / The major contribution of this research is to develop the exposure curves suitable for the Taiwan Contractor’s All Risks (CAR) Insurance (Material Damage Cover). The exposure curves generated in Salzmann (1963) and Ludwig (1998) were originated from the loss experience in the United States. Considering the different types of policies, different types of risks, and different types of perils, the applicability of those curves in Taiwan market is discussable. This study reviews the scope of coverage, insuring clause and exclusions of the Taiwan Contractor’s All Risks Insurance. According to the loss records between 1996 and 2000, it’s understood that the result of the Contractor’s All Risks Insurance is highly influenced by the natural catastrophic perils, such as earthquake, typhoon and heavy rains. The reinsurance placement can be categorized into proportional and non-proportional reinsurance or called excess of loss reinsurance. Exposure rating is one of the common pricing methods for the excess of loss reinsurance. Compared with those two previous researches done by Salzmann (1963) and Ludwig (1998), the outcome of this analysis shows an unbalanced loss distribution which leans to the lower layers that is mainly because of the following reasons: - (1) Low PML percentage of analyzed risks: The PML represents the Probable Maximum Loss subject to a certain type of peril. Taking a road construction project as an example, the PML of a road construction project is less than 30% of the total sum insured. Usually, the length of a road construction project is tens of kilometers or even much longer. In consideration of its linear shape, it’s very unlikely to have a loss exceeding 30% of its total sum insured. Therefore there was very few or nil loss data above the 30% of the total sum insured that’s the reason why the loss distribution is unbalanced and leans to the lower layers. (2) Long Construction Period of a CAR Project: A character of a CAR project is having a multiple-year policy period because the construction work usually takes more than one year to be completed. Therefore the possibility to have a total loss case, loss amount equal to the total sum insured, is much less than a property risk with a fire scenario. (3) Types of Perils Analyzed: Salzmann (1963) carried out her research by analyzing the peril of fire covered under the homeowners’ policies. As we know, most of the houses in the United States are the wooden structures, which are combustible. Oppositely, the risks in this research are mostly constructed by using the non-combustible materials, such as steel or reinforced concrete. Keywords: excess of loss, non-proportional reinsurance, exposure rating, and contractor’s all risks insurance.