風險值方法實證研究─以一壽險公司為例

Abstract

風險值(VaR)目前是金融機構計算市場風險最常使用的方法。雖然這個方法這麼頻繁地被使用，它仍然有一些缺陷。近年來，金融機構的投資活動成長相當快速，其投資的商品也越來越多元和複雜，在這樣的情況下，公司內部複雜的結構型模型無法在99%信賴水準下，比簡單的單變數模型有更好的準確性和預測能力。因此，單變數模型對於公司內部的結構性模型至少是一個相當有用的參考和輔助。本篇論文是第一篇使用單變數模型並採用一家台灣壽險公司歷史資料的實證論文，且有比較單變數模型和公司內部多變數結構模型的表現。

Value-at-Risk (VaR), nowadays, is the most widely adopted risk management method for measuring market risk in financial institutions, like banks, securities companies, and insurance companies etc. Although this measure is so widespread, it has some setbacks. In recent year, trading activities in financial institutions have grown substantially and became progressively more diverse and complex. In this situation, the complicate structural models were not able to outperform a simple univariate model in terms of accuracy and forecasting ability in 99th percentile. Univariate models, therefore, are at least a useful complement to large structural models and might even be sufficient for forecasting VaR. This paper is the first article adopts univariate methods with historical data from a life insurance company in Taiwan and provides a comparison of the performance between the univariate one and the models actually in use within firm.