CSFB/TREMONT避險基金績效與風險的探討

Abstract

投資績效的衡量一直是投資人最關心的課題，投資的盈虧清晰可見，但投資績效的優劣卻不容易判定，尤其是避險基金強調其追求絶對報酬，與股票、債券市場的相關性低，對於一般投資人來說，顯少揭露其投資風險。本研究針對Credit Suisse/Tremont investable index避險基金作常態分配檢定，以報酬率、變異係數、Sharpe指標及Sharpe ratio of VaR指標，分析各種投資策略的績效排名，並以不同的風險值模型估算避險基金的風險值並分析其間之差異。 實證發現在小樣本的研究，避險基金的報酬率分配多為常態分配，然仍有部分的投資策略的避險基金為非常態分配有高峰偏態現象。在平均報酬率的績效排名以新興市場策略的避險基金的績效最好，但考慮到風險的波動性，以變異係數為績效排名時，則以事件導向策略的避險基金為首選。另本研究發現，當避險基金出現超額報酬為負值時，運用Sharpe指標及Sharpe ratio of VaR指標的績效排名會出現偏誤，使得績效較佳的基金有較差的排名，而有修正調整的必要。將修正後的Sharpe指標予以績效排名時，發現Sharpe 指標與平均報酬率的績效排名相較，其排名次序變動不大，係高報酬率的避險基金其風險的波動性也很高，故高的超額報酬有高的風險波動及低的超額報酬有較低的風險波動。 在風險值的計算本研究實證發現，以歷史模擬法計算之風險值，均高於同時期以蒙地卡羅模擬法估算的風險值，並且沿著蒙地卡羅模擬法計算的風險值做小幅度的變動，兩者風險值相當接近，且在其信賴區間95%下，以Delta-Normal、歷史模擬法或蒙地卡羅模擬法，所驗證之回溯測試值都具有可信度。

This research uses the normal distribution test to evaluate the performance of monthly index return for nine hedge funds of Credit Suisse/Tremont investable index. The ranking for various investment strategy of hedge fund is made based on the return rate, the coefficient of variation, the Sharpe ratio, and the Sharpe ratio of VaR. Besides, the risk values of various hedge funds are evaluated in the Delta-Normal approach, the historical simulation approach, and the Monte Carlo simulation approach. The empirical results show that the return rates for parts of hedge funds are not the normal distribution with leptokurtic and skewed phenomenon even though most hedge funds fit the normal distribution in small sample. If ranked in the mean return rate, the performance of the investment strategy for emerging market is best. If considering the risk volatility and ranked in coefficient of variation, the performance of the investment strategy for event driven is best. Besides, when the excess return is negative, the ranking would be incorrect if using the Sharpe ratio and the Sharpe ratio of VaR. It is necessary to modify the evaluation method. Comparing the ranking in the modified Sharpe ratio and the ranking in the mean return rate, there is the minor change only. The result shows that the higher the return is, the higher the volatility is. The risk evaluation of the empirical results shows the risk estimated in the historical simulation approach is higher than the risk estimated in the Monte Carlo simulation approach for the investigated period. Both values are very close. At 95% confidence interval, the trace tests are reliable in Delta-Normal approach, the historical simulation approach, and the Monte Carlo simulation approach.