台灣加權股價指數的預測與台指選擇權交易策略之研究

Abstract

隨著近期期貨與選擇權在台灣證券市場的興起，民眾多了幾項投資標的可供選擇。選擇權不僅有避險的功能，多數人更將它視為投機工具。本研究希望能為投資人找尋一套有效預測台灣加權股價指數(以下簡稱台指)的方法，以期能在台指選擇權交易中獲利，另一方面基於學術考量，用以檢定台灣證券市場的效率性，看是否能找出一套模式，成功預測指數未來走勢並持續賺取利潤。 本研究蒐集民國91年至95年六月每日台指收盤價歷史資料，及一般認為對台指有顯著影響力的幾個變數：如成交值、美國道瓊工業指數、那斯達克指數與外資買賣超等資訊，利用計量經濟方法中Probit、Logit模型，建立每日漲跌的預測機率，搭配三大權值股GDR前一日股價走勢，設計交易策略，並對民國95年一月三日至六月九日等一百個交易日之台指選擇權進行交易，藉由觀察利潤判斷此方法能否準確地預測台股未來的走勢，以達到獲利的最終目的。 實證結果顯示： 1.本研究所採用的兩種研究模型，預測機率方面Logit模型的機率值較Probit模型的機率值極端，且兩模型對隔日漲跌的預測幾乎完全相同。 2.就預測的準確度與損益情況而言，整體模型無法優於隨機預測，即模型不具適用性。 3.多頭期市場狀況下模型較具適用性，盤整期次之，空頭市場下模型最不具適用性。 4.以三大權值股前一日GDR股價走勢，作為判斷台指漲跌幅的依據，確實能讓我們在選擇權交易中提升獲利。 5.無法利用過去資訊，持續且有效地預測台指未來的變動趨勢，顯示出台灣股市的弱式效率性成立。

As the futures and options markets began to rise, more investment targets are available for people now. We can use options not only to hedge. Most investors even trade them for speculation purposes. We expect this thesis could help investors find an effective method to forecast the value of “Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX)”in order to gain profits from options. On the other hand, for the academic consideration, we will test the efficiency of Taiwanese stock market by this research. See if we can find an effective way to predict the trend of the index value successfully and gain profits continuously. In this thesis we gathered daily historical data of TAIEX and several variables that are considered to be influential to TAIEX from 2002 to 2006. These variables are for example: volume, Dow Jones Industrial Average Index, Nasdaq Composite Index and foreign capital. By using econometrics methods such as “Probit and Logit” models, we can build probabilities of predicting the daily fluctuation of TAIEX. Along with previous day’s GDR prices changes of the three main stocks of TAIEX, we can then buy or sell “TAIEX Options (TXO)” and determine if this method correctly predicts the future trend of TAIEX by observing the profits from TXO. Followings are empirical conclusions available from this thesis: 1.The two models in this thesis bring no significant difference in predicting index value fluctuations. But Logit model has probabilities that are more extreme than Probit. 2.We cannot get a better performance with these econometric models than random predictions. That is, in the long run, they are not applicable. 3.These models will be most useful in the bull market, and least useful in the bear market. 4.Taking previous day’s GDR prices changes of the three main stocks of TAIEX as an accordance showing how fluctuant TAIEX will be, really helps us earn more money in option trading. 5.We cannot use any past information to predict the trend of TAIEX continuously and effectively. That is, the weak form of market efficiency does exist in Taiwan stock market.