多重風格投資擇時策略之研究

The Investigation of Multi-style Timing Strategy

陳淑怡(Shu-Yi Chen)

指導教授 ： 胡為善

中原大學/商學院/企業管理研究所/碩士(2007年)

https://doi.org/10.6840/cycu200700465

摘要

本研究運用風格投資的概念，針對1996年10月至2006年10月台灣證券交易所掛牌上市之股票，以公司市值與市價淨值比(P/B比)進行分類，建立單一風格(Single-Style)投資組合(包括大型股、小型股、價值股、成長股等四種)與多重風格(Multi-Style)投資組合(包括大型成長股、小型成長股、大型價值股、小型價值股等四種)，並以等比重加權計算出每一個風格投資組合的報酬率，且分別進行一個月、三個月與六個月的報酬率差異之 t 檢定。同時為了驗證風格輪動策略的價值，本研究進一步以分類與迴歸樹(Classification and Regression Tree，CART)為擇時工具，以1996年11月至2004年10月為學習樣本，而以2004年11月至2006年10月為測試樣本，並以因素分析萃取出金融指標(Finance index，F index)、景氣指標(Prosperity index，O index)及物價指數(Prices index，P index)三個因素構面，作為預測因子，分別建立靜態樹與動態樹模型之預測風格輪動，將其結果與將資金固定投資於單一與多重風格投資組合之報酬率加以比較，試圖找出何種方式能獲得較高的報酬率。本研究實證結果如下：
1. 本研究發現在二十二種不同的組合及三種月份的檢定中，只有大型成長股vs.大型價值股及大型成長股vs.不含金融指數等兩種組合，在三種月份的檢定中，其報酬率均具有顯著的差異。還有大型股vs.不含金融指數及成長股vs.不含金融指數等兩種組合，在單月與三個月的兩項檢定中，其報酬率具有顯著的差異。至於成長股vs. TAIEX及大型成長股vs. TAIEX等兩種組合，則只在單月的檢定中，其報酬率具有顯著差異。此外尚有大型價值股vs.小型價值股的組合，亦只有在六個月的檢定中，其報酬率之差異達顯著水準，其餘十五種組合在三種月份的檢定中皆呈不顯著之差異。
2. 本研究接著運用靜態樹模型進行擇時策略，並將其報酬率與固定投資於單一與多重風格投資組合之報酬率加以比較。實證結果顯示，在學習與測試期間，靜態樹的報酬率皆勝過固定投資風格之投資組合。同時依據靜態樹模型中的自變數與應變數的關聯，找出分割值。當 F 小於分割值時，不論投資單一與多重風格投資組合報酬率皆為負。當 F 大於分割值時，投資小型類投資組合與價值類投資組合，仍有機會獲得正向報酬率。
3. 本研究最後運用進化樹模型進行擇時策略，並將其報酬率與資金固定投資於單一與多重風格投資組合之報酬率加以比較。實證結果顯示，在學習與測試期間，在進化樹的各節點報酬率中，除了大型股的價值報酬差(Value Spread of Large Stocks，VSL)的五個項目中之「大型成長股」類別下的報酬率較低外，其餘各類別的報酬率皆勝過固定投資風格投資組合。當F小於分割值時，投資小型類與價值類仍有機會獲得正報酬。當F大於分割值時，不論投資單一與多重風格投資組合皆可獲得正報酬，顯示進化樹模型的動態作法，能納入及適應市場變化，因而產生較靜態樹佳的預測能力。

關鍵字

分類與迴歸樹 ； 單一風格投資組合 ； 擇時策略 ； 風格輪動 ； 多重風格投資組合

並列摘要

This study attempts to examine the style investing and targets at the stocks traded at Taiwan Stock Exchange from October 1996 through October 2006. First, this work constructs the single-style portfolios (i.e. large stock, small stock, value stock and growth stock) and the multi-style portfolios (i.e. large-growth stock, small-growth stock, large-value stock and small-value stock) based on the market value and P/B ratio, then utilizes the t-test on equally weighted returns of one month, three-month and six-month. This investigation employs the Classification and Regression Tree (CART method) to build a model in order to make the timing strategy of the style rotation. This study divides the data set into two periods: the learning period from November 1996 to October 2004 and the testing period from November 2004 to October 2006. By using the factor analysis, this work extracts three conceptual values: Finance (F) index, Prosperity (O) index and Prices (P) index attributes to construct the static tree and evolving tree model. The conclusions are summarized as follows:
1. The empirical result indicates that, under the 22 varieties of portfolios and three tests of one month, three-month and six-month respectively, there are significant difference between the rate of return on the large-growth stocks vs. large-value stocks, as well as that between the rate of return on the large-growth stocks vs. that on stock index excluding banking industry. The findings also show that, for one month and three month tests only, there are significant difference between the rate of return on large stocks vs. that on stock index excluding banking industry, and significant difference between the rate of return on the growth stocks vs. that on stock index excluding banking industry. The result also demonstrates that, for one month only, there are significant difference between the rate of return on the growth stocks vs. that on TAIEX, and significant difference between the rate of return on the large-growth stocks vs. that on TAIEX. For six month only, there are significant difference between the rate of return on the large-value stock VS. that on small-value stocks. All other 15 groups are insignificant.
2. This study then utilizes the static tree to make the timing strategy and compares the request with the regular investment in single-style portfolios and that in multi-style portfolios. This work finds that, during both learning and testing periods, the returns of static tree model outperforms the regular investment in single-style portfolios and that in multi-style portfolios. This investigation then calculates the split value according to the relation between the independent variable and dependent variable of static tree. The empirical results demonstrate that, under the static tree model, the returns are negative in both the single-style portfolios and multi-style portfolios when F index is less than the split value. However, an investor may acquire positive returns on investing in the small stock portfolio and valued stock portfolio even when F index is greater than the split value.
3. This study also utilizes the evolving tree to make the timing strategy and compares with the regular investment in single-style portfolios and multi-style portfolios. This study finds that, during the learning and testing periods, except the large-growth stocks under Value Spread of large stocks (VSL) has the lower rate of return than regular investments, most of the rates of returns derived by the evolving tree models outperform the regular investment in single-style portfolios and multi-style portfolios. This result indicates that, under the evolving tree model, an investor can obtain positive returns on investing in small stock portfolio and valued stock portfolio when F index is less than the split value. However, the rates of returns are positive in the regular investments in both the single-style portfolios and multi-style portfolios when F index is greater than the split value. This result suggests that the evolving tree model with dynamic nature can provide better predictive results than static tree model does.

並列關鍵字

Style Rotation ； Single-style Portfolios ； Multi-style Portfolios ； CART ； Timing Strategy

參考文獻

李春旺，1989，股價行為與規模效應：台灣股票市場實證研究，國立政治大學企業管理研究所博士論文。

劉秉龍，2002，成長型與價值型投資策略之實證分析—以台灣股票市場為例，靜宜大學企業管理研究所碩士論文。

鄭忠樑，2002，運用分類樹於股價報酬率預測之研究，元智大學資訊管理研究所碩士論文。

Ajayi R. A. and M. Mougoue, 1996, “On the Dynamic Relation between Stock Prices and Exchange Rates”, The Journal of Financial Research,19, 193-207.

Arshanapalli, B., Coggin T. D., and Doukas J., 1998, “Multifactor Asset Pricing Analysis of International Value Investment Strategies”, Journal of Portfolio Management, 24,10-23.