本研究探討長短期的市場反應是否會因不同形式的未預期盈餘所改變。我們使用時間序列以及分析師預期形成的八種模型去計算未預期盈餘。此八種模型之間的相關性顯示，雖然模型間彼此有相同的資訊內容，但每一種模型都仍有其他模型沒有、僅自己獨有的特別資訊。實證結果發現使用分析師預期的未預期盈餘模型反應出的短期市場反應會高於使用時間序列的未預期盈餘模型，尤其是使用分析師預測的中位數作為預期盈餘。在長期的市場反應方面，我們發現大部分的盈餘宣告後股價延遲反應發生在贏家投組中，而使用時間序列的未預期盈餘模型更為顯著，該發現表示當使用分析師預期所形成的未預期盈餘模型反應不足相對較小。除此之外，雖然使用分析師預測的模型解釋較多，但時間序列和分析師預期的模型都仍有大量能力去預測未來股價報酬。市場反應的延後會因不同定義的未預測盈餘模型而有所改變，但較少證據顯示盈餘宣告後股價的延遲反映是由於使用某一種特定模型所產生的。 This study examines whether the short-term and long-term market reactions to earnings surprises (SUE) can be altered by different styles of SUE. Eight SUE models, including time-series based SUE models and analyst forecast based SUE models, are implemented into SUE calculation. Correlations among eight models show that, though sharing some common information contents, each model has a certain amount of information contents which are not related to others. We find that the short-term market reactions using analyst forecast based SUE models, especially the model using analyst mean as expected earnings, are higher than those using time-series based SUE models. In terms of long-term market reaction, most of post-earnings-announcement drifts are from the winner portfolio and more significant using time-series based SUE models, indicating that the underreaction to SUE is relatively smaller when using analyst forecast based SUE model. In addition, although analyst forecast based SUE explains more, time-series based SUE and analyst forecast based SUE both have substantial abilities to anticipate future stock returns. Overall, the delayed market reactions are altered by different definitions of SUE but less evidence shows that post-earnings-announcement is an illusion from a specific SUE model.
