偵測財務報表不當認列─以條件式邏輯迴歸來評估

Abstract

本研究的目的是希望藉由不同層面的偵測指標來分析財報不當認列之特性，以期建立出偵測財報不當認列之模型。為了觀察財報不當認列公司之特性，本研究是以美國證券交易委員會在2007年至2013年3月所發布的會計及審計處分函作為主要研究資料。在整理及分析美國證券交易委員會之處分函後，發現常見的不當認列項目包含:收入及應收帳款的不當認列及費用的不當認列。本研究並進一步選擇應計項目、財務績效、非財務性指標、資產負債表外指標以及與市場相關之誘因來測試財報不當認列公司與無財報不當認列公司兩者間之差異。最後再利用不同層面偵測指標與條件式邏輯迴歸來建立偵測不當認列之模型，以期提升偵測財報不當認列公司之準確度及改善統計上之錯誤。

The purpose of this study is to analyze the significant difference between misstating firms and no misstating firms and develop models to detect misstatements. The sample of misstating firms which this paper chooses was subject to the SEC’s AAERs between 2007 and 2013. After analyzing the features of misstating firms, this paper finds that revenue misstatements and expense misstatements are common misstatements. Moreover, this study concentrates on various aspects, such as accruals quality related variables, performance variables, nonfinancial variables, off-balance-sheet variables, and market-related incentives, to analyze the significant difference between misstating firms and no misstating firms. Finally, this paper uses conditional logistic regression to develop detective models, in order to improve the capability of detecting misstatements.