以股票報酬衡量之盈餘品質指標

Abstract

過去對於資訊風險定價的研究多半採用財務會計數字為基礎，藉由分析盈餘品質指標與股票異常報酬率，推論盈餘品質與市場反應的連結性。Francis et al. (2005, 2006)則建立以報酬率為基礎的指標，衡量市場對於盈餘品質資訊風險的認知敏感度。 本研究參考Francis et al. (2006)的研究方法，以台灣1990年至2004年間上市櫃公司資料進行資本市場效率研究。本文採用應計項目品質作為資訊風險代理變數，將財務會計資料為基礎的應計品質指標(AQ)轉換為以報酬率為基礎的應計品質(AQ factor)，計算系統性資訊風險溢酬因子(e-loading)，並將該指標與過去實證研究中效果顯著的資訊風險代理變數作迴歸分析，測試其作為資訊品質指標的適切性。實證結果發現： 1. e-loading與非裁量性應計品質指標以及Francis et al.(2004)提出的七項盈餘品質指標具有相當程度的關聯性。 2. e-loading與分析師財務預測分散程度呈現顯著正相關，但未與盈餘反應係數以及分析師預期之正確性呈現顯著負相關。 3. e-loading與公司公開發行期間長度呈現負相關，自相關程度亦逐漸增加。 4. 財務報表重編之公司具有較高的e-loading，顯示投資人知覺該類事件樣本的資訊風險較高，反映為較高的資金成本。 本研究之實證結果支持以報酬率為基礎之盈餘品質指標，可有效辨認投資標的報導盈餘內含之資訊風險水準，為一具有辨識性與普及性的資訊風險指標。

The previous researches about pricing information risk of capital market are often based on the financial numbers. Theses literatures analyzed the empirical evidence of earnings quality and abnormal return to confirm the relationship between earnings quality and the efficiency of capital market. Conditioning on the factor-mimicking portfolio approach, Francis et al. (2005, 2006) built a return-based earnings quality measures to demonstrate a risk premium for firms. Referring to the Francis et al.’s (2005) procedures, we select the publicly traded firms between1990 to 2005 as the research sample. We view accruals quality as a proxy of information risk, convert the accounting-based measure (AQ) to a return-based representation (AQ factor). Then, we calculate the slope coefficient from a regression of a firms’ daily excess return in one year on AQ factor to obtain e-loading. To verify the validity of e-loading as a proxy of earnings quality, we regress e-loading on previously significant proxies of earnings quality. Inference made from the empirical results are summarized as follows： 1. e-loadings are highly correlated with the innate determinants and the seven earnings attributes considered by Francis et al. (2004). 2. e-loadings are significantly and positively correlated with the dispersion of analyst forecast, but not significantly negatively associated with earning response coefficient and the accuracy of analyst forecast. 3. The level of e-loading declines as the firm matures and the autocorrelation in e-loadings increases with the firm age. 4. e-loadings are significantly larger for restatement samples than for non-event samples. According to the empirical results, we document that e-loadings are a reliable return-based representation of earnings quality as measured by accruals quality. It can identify the different level of information uncertainty contained in reporting earnings with at least as well as other measures in the contexts that we examine. Thus, our results indicate that e-loading is a valid and popular proxy of earnings quality.