本研究選取1990至2001年上市公司之季資料,依據Ohlson(1995)股價為盈餘或異常盈餘、帳面價值及其他資訊的線性模型,以主要變數之不同操作定義組合成的十二條模式為基礎。主要目的在於測試及比較橫斷面(Cross Sectional)與時間序列(Time Series)模式在價格模式下的有效性,在時間序列方面採用的是Panel Data中的Fixed effect模型;在橫斷面方面採用的亦為Panel Data分析,但解開固定效果模型中(Fixed Effect Model)對常數項之設定。在比較橫斷面與時間序列的過程中,分為解釋力及預測力兩方面,使用正確性及解釋變動能力兩種標準,觀察是否時間序列的分析方式均較橫斷面的分析方式效果為佳。 在正確性方面則兼採MAPE、RMSE、真實差異及絕對差異四個方式來衡量。真實差異高低估的方向雖然不甚一致,但絕對差異之平均數或中位數均是時間序列方式之結果顯著較佳,故大體上來說,時間序列分析方式下所得出估計值的正確性均較橫斷面分析方式下所得出之結果為佳。在解釋變動能力方面,先對產生估計式的Panel Data模式結果進行Vuong Test,比較兩分析方式下相同模式下之判定係數是否有顯著之差異。研究結果發現在時間序列分析方式下,模式一至模式十二之判定係數皆顯著優於橫斷面分析。此研究結果顯示由時間序列分析方式所產生估計模型之整體相關程度顯著較佳。本研究也應用OLS迴歸及Rank迴歸進行單因子迴歸分析及多因子迴歸分析,也發現時間序列分析方式下所得出之估計值對於真實公司價值變動的增額解釋能力顯著的高於橫斷面分析方式所得出結果,時間序列分析方式具有較佳解釋變動之能力。敏感性測試方面則考慮金融業與否及負盈餘的因素,在前述三方面的結果皆與主要測試結果相似,故實證結果尚稱強健。 本研究亦採用部分的樣本期間根據兩種分析方式所產生出之估計式,對後期進行預測,結果發現不僅時間序列分析方式所得出預測值解釋變動之能力優於橫斷面方式,且在預測的正確性方面,也顯著的有著較佳的表現。 由上述結果可得知,使用時間序列的分析方式下,不論是估計值對同時點價值的解釋力,或是對真實價值的預測力,從正確性及解釋變動能力來看,皆具有較好的效果,而根據本研究所提供之證據,希望能對後續研究進行實證的分析方法選擇上有所助益。
In past, most valuation researches about Ohlson Model are cross-sectional studies. Cross-sectional analysis is more usual and emphasized the importance of fundamental value measures that track contemporaneous return and prices. And Time-series research focus on the time-series relation between earnings, book values or other value-relevant variables to explain prices, returns and predict future returns. Lo and Lys(2000) suggested that the Ohlson Model is written as a model for a single firm. And as Francis, Olsson, and Oswald(2000), we compare alternative empirical estimates of intrinsic value using two criteria: accuracy and explainability. This study compares the reliability of value estimates from the cross-sectional analysis and time-series analysis based on valuation model proposed by Ohlson(1995) and Feltham and Ohlson(1995). Using a sample of listed firms in Taiwan over 1990- 2001. And we apply Panel Data to do the time-series and cross-sectional estimates and forward analysis. We find that the time-series estimates are more accurate and explain more of the variation in security prices than do cross-sectional value estimate. On the other hand, the accuracy and explainability of time-series forecasts are also better than cross-sectional forecasts. For the sensitivity test, we consider financial industry and negative earnings. The sensitivity test results are similar to previous outcomes. In summary, this study provides evidences to support that time-series estimate is a better analysis method and can raise the effectiveness of valuation model. So during research process, if we can measure another econometrics methods’ results. The empirical outcome we obtain can be more robust.