金融海嘯前後外匯市場風險值與報酬率-以高收益貨幣匯率為例
自1990 年以來,發生美國加州橘郡事件、英國霸菱事件、德國金屬工業公司、日本大和銀行及美國長期資本管理基金等重大金融事件乃至近期在美國發生次貸風暴衍生金融海嘯,影響全球國家如冰島、西歐三國及國際金融機構倒閉如雷曼及政府金援如花旗、高盛、美林等,而在此際中國大陸快速崛起,在全球經濟體佔有舉足輕重的地位,全球歷經金融海嘯後在逐漸復甦之際,又再發生歐債危機,全球外匯市場更是波濤洶湧,故適當的量化風險是管理風險的重要工作之一。 近代的財務金融文獻,學者Morgan (1976)、Engle (1982)、Bollerslev (1986)、Engle and Manganelli (2000)普遍同...
Main Authors: | , , , |
---|---|
Other Authors: | |
Format: | Thesis |
Language: | Chinese |
Published: |
2011
|
Subjects: | |
Online Access: | http://hdl.handle.net/11536/43905 http://140.113.39.130/cdrfb3/record/nctu/#GT079672526 |
Summary: | 自1990 年以來,發生美國加州橘郡事件、英國霸菱事件、德國金屬工業公司、日本大和銀行及美國長期資本管理基金等重大金融事件乃至近期在美國發生次貸風暴衍生金融海嘯,影響全球國家如冰島、西歐三國及國際金融機構倒閉如雷曼及政府金援如花旗、高盛、美林等,而在此際中國大陸快速崛起,在全球經濟體佔有舉足輕重的地位,全球歷經金融海嘯後在逐漸復甦之際,又再發生歐債危機,全球外匯市場更是波濤洶湧,故適當的量化風險是管理風險的重要工作之一。 近代的財務金融文獻,學者Morgan (1976)、Engle (1982)、Bollerslev (1986)、Engle and Manganelli (2000)普遍同意波動性變化具有因時而異且有叢聚的特性,因此本研究採用最能描繪自我相關條件異質變異的GARCH族群模型包括AR(1)-GARCH模型、AR(1)-GARCH -M模型與AR(1)-EGARCH模型進行匯率報酬率之風險值估計。至於樣本之選擇,因本研究是利用滾動程序(rolling)方法來估計風險值,觀察金融海嘯發生前後匯率波動情形,針對高收益貨幣-(澳幣、紐幣、南非幣) 、主要交易貨幣-(歐元、英磅、日圓)及新台幣、人民幣、新加坡幣匯率報酬率樣本觀察值,藉以觀察觀察金融海嘯發生前後對於風險值模型的績效結果影響,最後再以交叉方式檢驗各模型何者較具準確性、保守性或效率性。 本研究利用滾動程序(rolling)方法來預測風險值,結果發現以金融海嘯前後的視窗長度,並無法讓其移動視窗所估計出之係數皆符合參數約束條件,而使其中部份匯率報酬率計算出的風險值失去有效性,且在相同的風險值模型之保守性或效率性方面,金融海嘯後較金融海嘯前具保守性,而在效率性上金融海嘯前後在效率性差異不明顯。 另外,本研究亦發現使用GARCH模型估計風險值時,在均數方程式引入「風險貼水」項,或變異數方程式採用自然對數形式,無法改進風險值模型之績效。 Since 1990 years, as United States California Orange County (Orange County) event, and United Kingdom Baring Bank (Baring Bank) event, and Germany metal industrial company (Metallgesellschaft), and Japan Daiwa Bank (Daiwa Bank) event and the United States long-term capital management Fund (Long Term Capital Management Fund), major financial event and recently of financial tsunami, effect global national as Iceland, and Western Europe three nationals and the international financial institutions collapse as Thunder Mann and the Government aid as Citi, and Goldman, and Merrill Lynch, At this point the Chinese mainland's rapid rise, occupies a pivotal position in the global economy, global has been in gradual recovery after the financial tsunami, European debt crisis, global foreign exchange markets is rough, so appropriate of quantitative risk is management risk of important work. Modern of financial financial literature, scholars Morgan (1976), and Engle (1982), and Bollerslev (1986), and Engle and Manganelli (2000) General agree fluctuations sexual changes has due to every now and then different and has clustering of characteristics, so this research using most can depicting self related conditions different mass variation of GARCH communities model including AR (1)-GARCH model, and AR (1)-GARCH-M model and AR (1)-EGARCH model for stock reward rate of risk value estimated. As regards choice of the sample, because this research is the use of scroll programs (rolling) method to estimate a value-at-risk, observation on the financial fluctuations in exchange rates before and after the tsunami, against higher-yielding currencies -( AUD, NZD, ZAR) , the main trading currencies -( EUR, GBP, JPY ) And the ASIAN currencies -( EUR, GBP, JPY ) New Taiwan dollar, Renminbi, and Singapore currency exchange rate Sample observations, so as to observe Observation before and after the financial tsunami Performance results for value-at-risk models , And finally to cross examination of the model which is more accurate, conservative or efficiency. Use scroll programs (rolling) method to predict the value-at-risk, and found Research and utilization of this rolling program (rolling) method to predict the value-at-risk, results found for Before and after the financial crisis window of length, , and can't get it to move the window to the estimated coefficients are in compliance with the parameter constraints, and the part of currency rates calculation of value-at-risk losing effectiveness, and value-at-risk model in the same terms of conservative or efficiency, After the financial crisis before the financial tsunami is conservative, and on the efficiency of the financial tsunami in efficiency before and after difference is not obvious. In addition, the study also found that when using the GARCH models estimated value-at-risk, the mean side program introduced a "risk premium", or variance procedures take the natural logarithm of form, are unable to improve the performance of value-at-risk models. |
---|