透過您的圖書館登入
IP:3.22.181.81
  • 學位論文

考慮極值分布之行為投資組合最佳化

Applying Extreme Value Theory in Behavioral Portfolio Optimization

指導教授 : 張國華

摘要


摘要 近年來,隨著投資組合的多樣化,各種金融商品不斷的推陳出新。目的就是要讓投資者願意購買此投資商品,但往往在優渥的報酬下,總是潛藏著不為人知的風險。所以近年來財務的風險控管,可說是很重要的議題,而且經過2008年金融海嘯後,銀行,或者是政府機構對於風險的控管可說是越來重視。而學者們近幾年也更專注的去研究風險管理的問題,但往往就是會低估極端事件發生的風險,所以應該以更精確的下端分配來正視風險發生的問題。 而為了能夠更準確的估計到尾端分配,我們應用極值理論(Extreme Value Theory)對尾端資料進行估計,以符合厚尾分配的情況;並且結合Copula相關性結構進行資料的模擬與擴充。而隨著行為財務學(Behavioral Portfolio Theory)的興起,投資者心理的狀態會影響到投資者未來對於股市報酬漲跌的看法不同,所以本研究結合SP/A(Security Potential/Aspiration)理論去改變每組情境模擬的機率分配,而在根據投資者對於每筆錢心理狀態的不同去劃分風險規避(Safety-First)與風險追求(Risk-Seeking)的心理帳戶(Mental Account),而本研究的觀察對象是參考摩根台灣股價指數(MSCI Taiwan Index)中的公司當作研究對象。 本研究的每個心理帳戶中共有三種模型,第一種是計數型的模型,第二種是Chebyshev不等式的推論型模型,第三種是Markov不等式的推論型模型。而從實證結果得到以下結論: 1.應用極值理論產生的模擬資料加上SP/A換機率的方法,結果表現優於使用歷史的原始資料。2應用極值理論產生的模擬資料加上SP/A換機率的方法,表現優於應用極值理論的模擬資料。3.應用極值理論產生的模擬資料加上SP/A換機率的方法在風險追求心理帳戶下表現優於大盤,然而在風險規避心理帳戶表現不如大盤。

並列摘要


Abstract Recently, because of the diversification of the portfolio and financial products risk control is an important issue. Several studies shows that we are likely to underestimate the downside risk of return, so we need to use precise way to estimate stock returns. In order to estimate the downside risk of the return precisely, we used the EVT and copula structure to simulate the stock return. Human behavior also influence the investor’s stock return estimates. There are a lots of evidence that showed human behavior affects investment performance, and people have different risks attitude towards their different mental accounts. In this study, we select the stocks in MSCI Taiwan Index as our investment objective. We used EVT and Copula to generate return scenarios, and based on SP/A theory, we consider the investor perception of fear and hope to assign the probability to each return scenario. In this study we consider two mental accounts: Risk-Seeking mental account and Safety-First mental account. In each account we have three different models Chebyshev Inequality Interference estimation model,Markov Inequality estimation model, and Counting scenario model. Our results showed that applied EVT to generate return scenarios and combined with SP/A changed probability measure performed well than used historical data. Second, applied EVT to generate the return scenarios and combined SP/A changed probability measure performed well than return scenarios. Third, return scenarios had good performance than market in risk seeking mental account.

參考文獻


IV. 李美杏、丁聖祐、關聯結構與最適投資組合-Copula 模型的應用
I. 王政鈞,使用模擬尾端分配資料以考慮單邊風險上界指標之數理規劃求解最佳化投資組合,2012
II. 楊千億行為投資理論建構下、投資組合之研究,2014
III. 林玉媚應用模擬資料之風險控制投資組合最佳化,2010
15. Kuo-Hwa Chang, Hsiang-Ju Chen, Chang-Yuh Liu “A STOCHASTIC PROGRAMMING MODEL FOR PORTFOLIO SELECTION”, Journal of the Chinese Institute of Industrial Engineers, Volume 19, Issue 3, 2002.

延伸閱讀