應計項目可靠性對股價影響—探討資訊延遲與錯估情形

Abstract

本研究旨在探討應計項目可靠性、盈餘持續性與市場定價間的關聯性，並以TEJ資料庫提供之常續性稅後淨利為基礎，評估其相較於傳統盈餘指標之持續性與預測能力。研究首先驗證，排除一次性項目及特殊項目的常續性盈餘能更有效地反映企業真實經營成果，因此具備較高之盈餘持續性與未來績效的預測力。研究亦發現應計數存在迴轉風險，因此進一步依據Richardson等人（2005）之方法，將應計數依資產負債表性質細分，並依各組成項目之經濟性質主觀判斷其可靠性。透過細分應計項目的多變量分析，本研究發現低可靠性應計項目對次期的常續性盈餘具有顯著的負向影響，且投資人未能即時反映其風險，導致市場錯誤定價，形成應計項目異常(Accrual Anomaly)。

接著，本研究將財報發布日後不同期間（週、月、季、年）的股票報酬為應變數，發現應計項目異常會隨時間推移而加劇，顯示投資人隨資訊揭露程度逐步修正其投資決策。最後，透過納入資訊環境變數與應計數之交乘項，本研究發現資訊環境改善未能顯著降低應計異常，顯示應計異常源於投資人系統性低估低可靠性應計項目的迴轉風險。

This study investigates the relationship between accrual reliability, earnings persistence, and market pricing. Using the “core earnings after tax” provided by the TEJ database as a basis, we evaluate its persistence and predictive ability compared to traditional earnings measures. The study first confirms that core earnings, which exclude one-time and special items, more effectively reflect a firm’s true operating performance and thus exhibit higher earnings persistence and stronger predictive power for future performance.

In addition, the study identifies the reversal risk embedded in accruals. Following the empirical framework of Richardson et al. (2005), accrual components are decomposed based on balance sheet classification and assessed subjectively for their reliability according to economic substance. Through multivariate analysis, the results reveal that low-reliability accruals have a significantly negative impact on next-period core earnings. Investors fail to promptly reflect this risk in stock prices, leading to market mispricing and the formation of an accrual anomaly.

Furthermore, by examining stock returns over various horizons (weekly, monthly, quarterly, and annually) following financial statement release dates, the study finds that the accrual anomaly intensifies over time, suggesting that investors gradually revise their decisions as information is revealed. Finally, by incorporating interaction terms between accruals and proxies for information environment quality, we find no evidence that better information environments significantly mitigate the anomaly. This implies that the anomaly stems from a systematic underestimation of the reversal risk associated with low-reliability accruals.