隨著科技日新月異，人口有更好的醫療資源，壽命也逐漸延長，各個國家皆有相同趨勢，透過良好的動態模型預測死亡率有助於減少錯誤訂價之機率。Milidonis and Efthymiou (2015)研究亞太地區之死亡率改善發現經濟發展較好的國家對經濟發展較差之國家具有領先關係，且驗證存在短期的預測能力。本研究以Lee-Carter (LC)模型 (Lee and Carter, 1992)為架構，延伸考慮其他影響死亡率改善之總體因子，透過向量自我相關 (Vector Autoregression；VAR)方法對LC模型的預測方法修正，並以台灣人口死亡率以及總體因子來配適，透過Improvement ratio驗證模型之改進有助於提升預測之準確度。最後，應用於近期較熱門之附保證投資型商品─保證終生最低提領給付商品 (Guaranteed Life-time Withdrawal Benefits; GLWB)，並透過敏感度分析參數與商品之影響程度。;With the advancement of technology, the population have better medical resources. Life is also gradually extended. Every country have the same trend. Through good dynamic model to predict mortality helps reduce the probability of mispricing. Milidonis and Efthymiou (2015) found that mortality risk improvements flow from more developed to less developed countries. And there exists a short-term predictability. In this research, I use the Lee-Carter model (Lee and Carter, 1992) which is one of the most popular model in the related research. This study extends the consideration of other economic factors associated with mortality improvement. Through vector autoregression (VAR) method to modify the LC model. We show that the modified models help to improvement the accuracy of predictability by using the data of Taiwan. Finally, we apply a popular product recently which is the Guaranteed Life-time Withdrawal Benefits (GLWB). And use sensitivity analysis to see impact between parameters and product.