退撫基金證券投資組合之風險值估計

Abstract

近年來，一連串的財務危機，使得財務機構、資產管理機關紛紛要求更健全風險管理。我國之退撫基金是在民國83年才成立，目前屬於成長初期，基金資產將會快速成長，面對這樣大型之基金，若無良好之風險管理方式，勢必影響基金參與人和全體納稅人之權益。也因此，本研究利用風險值系統來評估退撫基金所能承受之風險，透過風險值意義的介紹及參數的挑選，以常態模型、歷史模擬法、拔靴法和蒙地卡羅模擬法來估計風險值，並比較其差異所在。另外，更分析累加風險值資訊供管理者決策之參考。最後，輔以壓力測試，針對可能受傷的情境作模擬。 本研究以退撫基金民國87年底的證券投資組合，搭配民國83年初到87年底的市場資料，實證結果如下： 1. 以指數加權移動平均法來估計其波動性，適合退撫基金這樣大量之投資組合，不僅容易執行，並且不失其準確 性。 2. 常態假設之模型有低估之情形，無法解決寬尾的問題，歷史模擬法、拔靴法、蒙地卡羅模擬法之結果則相差不 多，顯見完全評價法優於常態模型。 3. 累加風險值之資訊，顯露風險來源之所在，可提供管理者作決策。 4. 壓力測試的結果發現，風險值估計仍有預料不到的情境，兩者必須緊密搭配，才能更健全完整。

Recently a series of financial disasters cause financial institutions and asset management firms to request for better sound risk management. Our national pension fund was established in 1994. Today it belongs to growing period and the fund will have a fast growth in the coming days. With so large scale of pension fund, it should affect pension fund participants and all taxpayers if there was no better risk management. Thus, this study uses Value-at-Risk system to evaluate the risk pension fund could take. Through the introduction of Value-at-Risk and the choice of parameters, I use normal model, historical simulation, bootstrap approach and monte carlo simulation to estimate Value-at-Risk and compare the existing difference. In addition, the analysis of Incremental VAR is supplied to managers for decision making. In conclusion, i turn to consider stress testing and attempt to gauge the vulnerability of our portfolio to hypothetical scenarios. The study is based on the pension fund stock portfolio at the end of 1998. The market data are from 1994 to 1998. The result is as follows: 1. The estimate of volatility by exponentially weighted moving average is suitable for large portfolio of pension fund. It is not only easy for implementation but also keeps accurate. 2. The model of normal assumption may understate true VAR and can't solve fat tails. The result for the use of historical simulation, bootstrap approach and monte carlo simulation has no too much difference. It is obvious that full valuation is superior to normal model. 3. Incremental VAR information can tell the source of risk for managers to make decisions. 4. The result of stress testing can highlight weaknesses in VAR procedures. Both are therefore natural complements to each other. So, it can be more complete.