過去台灣的地震保險一直受到忽略，因此缺乏完整的地震損失資料。本研究主要的目的是利用統計方法來模擬地震損失，估計地震損失所造成的直接損失金額，並且進一步討論如何控制地震超額損失保險的預期損失。 本研究推論災害性地震發生頻率是一個負二項分配﹔而損失幅度則利用所求出的房屋倒塌數模型，再假設每一棟倒塌房屋的個設價格作加總而得，因此模擬出地震損失，可以估計出預期地震損失，並且可以進一步建立地震損失機率分布情況。 對於控制地震超額損失保險的預期損失，本研究提出以理賠上限與理賠起始點兩種方式來作損失控制。實證結果發現，利用理賠上限作超額再保險會比較可以控制地震損失。同時也可以利用模擬出的結果來決定應該如何設立理賠上限與理賠起始點。 / The earthquake insurance in Taiwan is ignored for many years, so that the data of earthquake is lacked. This study applied the statistical methods to simulate the earthquake losses in Taiwan, and estimated the loss amount, and discussed how to control the expected loss of excess-of-loss insurance. The frequency of disaster earthquake is a negative binomial distribution. And the severity of each disaster earthquake is to sum up the assumed dollar amount of each damaged houses. Compounding the frequency of earthquake with the severity of earthquake, we can simulate the losses of earthquake. Hence the expected earthquake losses and the probability density function of earthquake losses are built. In order to control the expected loss of excess-of-loss reinsurance, this study compared the results of the trigger level and the capped level of the reinsurance claim, and concluded that the capped level is better than trigger level in controlling the expected loss of excess-of-loss insurance.