| dc.contributor | 風險管理與保險學系 | |
| dc.creator (作者) | 蔡政憲 | zh_TW |
| dc.date (日期) | 2014 | |
| dc.date.accessioned | 5-Aug-2015 11:11:48 (UTC+8) | - |
| dc.date.available | 5-Aug-2015 11:11:48 (UTC+8) | - |
| dc.date.issued (上傳時間) | 5-Aug-2015 11:11:48 (UTC+8) | - |
| dc.identifier.uri (URI) | http://nccur.lib.nccu.edu.tw/handle/140.119/77369 | - |
| dc.description.abstract (摘要) | 死亡率風險對壽險公司、社會福利計畫、以及整個社會都很重要。文獻提出三種管 理方法：自然避險，運用與死亡率連結的證券，與建立死亡率的預測模型。建立預測模 型時如果是分別建立，各國的死亡率必然發散。因此我們必須建構跨國的死亡率模型。 建立這類多國死亡預測模型的動機有四個：(1)各國死亡率被觀察到有一致的趨勢； (2)死亡率連結商品常以多國的死亡率為標的；(3)評估運用死亡率連結商品的基差風 險；(4)協助跨國（再）保險公司的負債管理。 在第一年的計劃中，我們綜合主成分分析（PCA）、ARMA、GARCH 等方法來找出 各國的共同因子並為各因子建立時間序列模型，這個方法被稱為O-GARCH。我們將從 Human Mortality Database 取得死亡率資料，再根據社會與經濟狀況將各國分組後建模。 我們也會試試因子間的共整合與共變異關係。 第二年我們將引進更好的估計方法。由於多國的死亡率資料其時間上的維度常比風 險因子的維度小，我們打算採用APCA 來獲得更有效率的估計。我們想採用HFA 來處 理死亡率資料有變異數不齊一的問題。我們也會用更嚴謹的方法來決定因子的個數。 第三年我們將考慮多國死亡率的群組結構，亦即有些共同因子是跨群組的、有些僅 適用於特定群組內。我們將用MGFA 建構性別－國家－地區－國際的群組結構。 | |
| dc.description.abstract (摘要) | Managing mortality risk is important to life insurers, social benefits programs, and the society as a whole. The literature has proposed three major ways to manage mortality risk: implementing natural hedging, utilizing mortality-linked securities, and building mortality projection models. The major problem of modeling the mortality rates of two countries independently is that their mortality rates will deviate almost surely in the long run. We therefore must take into account the co-movement of mortality trends in different countries therefore. This will also enhance the predicting accuracies of those models. This multi-year project intends to build multi-population mortality projection models. The development of such models are driven by (1) the empirical observation on interdependent dynamics among populations, (2) the mortality products linked to multi-population mortality indexes, (3) the basis risk emerging when using the securities that underlie on single-population indexes, and (4) the mortality risk management of multi-national life (re-)insurers. We propose a unified, comprehensive framework for multi-population mortality projections. In the first year, we will integrate the methods of principal component analysis (PCA), ARMA, and GARCH to identify the common factors among the populations of different countries and then model the time-series dynamics of the identified factors. Such an integrated method is called O-GARCH in the literature. We intend to group countries by socio-economic statuses to conduct the O-GARCH analysis. The mortality data will be from Human Mortality Database. We will further compare two alternative approaches that links factor model of individual populations through cointegrations or covariances among factors. Our second year project focuses on improving the estimate of factor model by using more efficient and rigorous techniques. We will incorporate the asymptotic principal component analysis (APCA) and heteroscedastic factor analysis (HFA). The mortality rates that we use in estimating the above factor models often have greater dimensions in age than in period. APCA can mitigate this dimensionality problem. We will further relax the homoscedasticity assumption on the idiosyncratic errors by using HFA. Another improvement in the second year will be adopting more rigorous methods in determining the number of factors. The third-year project introduces the idea of group structures. Imposing the multi-group framework by identifying layers of common factors will probably enhance model performance. We plan to use the multi-group factor analysis (MGFA) to find the consistently estimated factors that span the pervasive subspace at the group level. The group structure that we have in mind has 4-layers: gender-country-region-global. This three-year project will render integrated multi-population dynamic models of mortality rates in which the common factors within a country and across countries are identified. The O-GARCH method is flexible; the estimation will consider the dimensionality problem and heteroscedasticity of mortality data; the choice on the number of factors will be objective; and the multi-group structure will allow us to incorporate the ex-ante demography information. At the end, we will have comprehensive models to deal with the pricing and risk management of mortality risk in the multi-population context. | |
| dc.format.extent | 144 bytes | - |
| dc.format.mimetype | text/html | - |
| dc.relation (關聯) | NSC102-2410-H004-027-MY3 | |
| dc.relation (關聯) | PF10301-0888 | |
| dc.title (題名) | 跨國死亡率模型之建構 | zh_TW |
| dc.title.alternative (其他題名) | Constructing Multi-Population Mortality Rate Models | |
| dc.type (資料類型) | report | en |
