探討以績效控制裁決性應計數作為臺灣市場盈餘管理衡量指標之有效性

The Validity of ROA-Matched Discretionary Accruals as an Earnings Management Measure in the Taiwan Market

裁決性應計數常被用作衡量財務報表品質之代理變數。為使其能有效偵測盈餘管理，其數值的變異不應與企業其他成長特徵或基本經濟面向存在系統性關聯。本研究發現，在台灣市場中，Jones與modified-Jones模型所估計之裁決性應計數，無論是否進行資產報酬率配對調整，其較大數值顯著偏向於擁有極高或極低盈餘股價比與歷史營收成長率的公司樣本。為此，本文提出兩種解決方法，分別為：一、以盈餘股價比為基準進行配對；二、採用因子迴歸進行調整。基於模擬方式的檢定結果顯示，上述兩種方法在無盈餘管理之虛無假說下，均能有效修正傳統模型之過度拒絕問題，並在存在盈餘管理之對立假說下保有足夠之檢定力，顯示其作為盈餘管理偵測具有較高有效性。

Discretionary accruals are often used to measure financial reporting quality. To be a valid proxy for earnings management detection, its large magnitude should not be systematically related to other growth characteristics or economic fundamentals. This study finds that the large values of discretionary accruals estimated from the Jones-type regression model, with and without ROA-matching adjustment, are skewed towards extreme earnings-to- price ratios and past sales growth in the Taiwan market. We consider two alternative approaches: one is based on matching by the earnings-to-price ratio, and the other is based on factor regression adjustment. The simulation-based tests demonstrate that the two proposed methods effectively address the over-rejection issues under the null hypothesis of no earnings management and retain sufficient power to detect earnings management under the alternative hypothesis.