具有外生變數之動態相關-以台灣類股指數為例

Dynamic Conditional Correlation with Exogenous Variables - Case Study of Taiwan Stocks Indices

陳姿廷(Tzu-Ting Chen)

指導教授 ： 李孟峰

國立臺北大學/商學院/統計學系/碩士(2011年)

摘要

許多研究均證實金融市場間具有市場聯結的特性，早期對於金融商品報酬的研究，多採用固定的相關係數探討金融商品與不同市場間的相關程度，但實際上商品或是市場間的相關性經常隨著時間發生變動。
　　因此本文採用Engle and Sheppard (2001) 所提出的DCC模型進行時間數列分析，探討總體經濟的外生變數對台灣股市的影響，以及台灣股市中不同類股間股票報酬率的動態相關。透過資產間的相關係數、變異數及共變異數，了解資產報酬的波動性、風險性及彼此的關聯性。若能及早預知總經指數的走勢將有助於對價格震盪找出因應對策，希望研究結果有助於投資者運用在風險規避的策略上。
　　本研究蒐集1998年12月至2010年12月之月資料共計145筆，以台灣四大類股報酬為研究標的，物價指數作為外生變數，探討電子、金融、營建及食品類股市報酬的動態相關。由實證研究歸納四點結果，（１）金融與營建類股的關聯性最高，最小的則是食品與電子類股，食品類股與其它三種類股的關聯普遍偏低，從金融與營建及電子的高度關聯顯示，營建與電子都是需要高度金融的產業，此外投資人若要分散投資組合應避免同時選擇金融與營建或電子類股。（２）由模型比較得知，CRB指數比CPI指數更能較好的配適模型，以其落後1至2期有良好的配適而言，CRB指數似乎有領先台股走勢的趨勢；所以若模型配適外生變數，將可除去由外生變數對類股造成的干擾，之後可以較準確的估計類股間的相關係數。（３）從動態條件相關趨勢圖研判，美國的次級房貸危機和雷曼兄弟銀行倒閉等負面事件，使台灣股票市場的波動劇烈，以次貸事件的影響較大，大部分類股間的動態關聯皆急遽降低，長達2至3個月不等。若是輸入性通膨或量化寬鬆等政策，則會使類股間的相關程度提高；此外，不同事件影響類股間的時間長度亦不同。（４）此模型可以較真實的反應類股報酬的動態相關，有助於投資人分散風險、配置投資標的之依據。
　　此外，台灣亦是開放的經濟體，因此投資人應將國際情勢做為投資決策的考量，以做最有利的投資。

關鍵字

消費者物價指數 ； DCC模型

並列摘要

Many researches found out that financial markets are connected with each other. In early researches into financial merchandise’s relations between different markets, that were mostly evaluated by constant correlation coefficient model. But actually either financial merchandise or relationships between markets were time varying.
　　As a result, this research combines Vector Autoregressive (VAR) model and Dynamic Conditional Correlation (DCC) model proposed by Engle and Sheppard (2001) to investigate how is Taiwan equity market influenced by macroeconomic factors, and also discusses the dynamic correlation in different fields of stocks. Furthermore, the changes of correlations resulting from the major financial events are statistically significant during the past 10 years. Through correlations, variances and covariance structural, we figure out the dynamic correlations in different fields of Taiwan stocks.
　　This thesis is focus on the dynamic interdependence among four fields of Taiwan equity markets by taking price indexes as exogenous variables in consideration, in order to eliminate the effect of price indexes to stocks. The four kinds of equities are financial, construction, electronics and food industry. Empirical results show that the relationship between financial and construction is the highest, and the lowest one is food and electronics. That means construction and electronics are highly financial-orientated industries. Besides, the most important finding is that correlation coefficients of the four stocks are significantly changing over time. In addition, comparing CRB price index with CPI price index, find out that the model with CRB as exogenous variable is better fitting than the CPI model, which indicates that if the model contains price indexes as exogenous variables, the model will be fitted better than without exogenous variables. Because of adding these variables can eliminate its disturbance to stocks, and then coefficients among stocks can be estimated accurately.
　　 Moreover, from picturing the dynamic conditional correlations between each stocks, such negative events like subprime mortgages crisis and the bankruptcy of Lehman Brothers effect Taiwan most, which leads to a big drop in stocks ‘correlation about three months long. However, if it happens to the events as imported inflation and quantitative easing monetary policy, which raises the correlations between them. Besides, the effecting periods of time vary from event to event. Since the comovement among Taiwan stock market becomes significantly affected by financial crisis, and the model with exogenous variable is more accurate between theory and real world. As a result, investors should take important financial events into account on spreading their risks and asset allocation.

並列關鍵字

Price Index ； DCC ； Stock

參考文獻

12.李欣儒 (2009)，以ADCC模型探討衝擊事件對亞洲主要股市間動態相關的影響，國立台北大學統計研究所碩士論文。

13.黃坤銘 (2009)，次級房貸危機及金融海嘯下股債市聯動性之研究-台美股債市案例與DCC模型之應用，國立台北大學國際企業研究所碩士論文。

15.曾菀縈 (2008)，台灣上市股票類股報酬率動態相關之探討，國立台北大學統計研究所碩士論文。

21.林大超 (2003)，台灣與美國兩地景氣循環指標關聯性之研究，國立成功大學企業管理學系碩士論文。

1.Akaike, H. (1973). Information Theory and an Extension of the Maximum Likelihood principle. 2nd Inter Symposium on Information Theory, 267-281, Budapest.