本研究以2010年至2019年因財務違規而受到處罰的中國大陸上市公司為研究對象,篩選出133份涉嫌現金舞弊的年度樣本,設置404份配比樣本,並將樣本按時間分為控制組和測試組。本文使用分層邏輯分析的方法,構建舞弊甄別模型。 首先, 本文根據舞弊三角理論,將舞弊三角之誘因或壓力、機會及態度和行為合理化以代理變數衡量,測試這些代理變數是否與現金舞弊相關聯,並辨認出12個存在顯著差異的變量,據此構建現金舞弊事前預警模型,此為企業現金舞弊之內在原因。 其次,本文根據發生現金舞弊公司的財務報表中各種財務和非財務比率的異常表現,辨認出10個具有顯著差異的現金舞弊風險因子,據此構建現金舞弊事後偵查模型,此為企業現金舞弊之外在表現。 分層迴歸分析中的第一層次模型為事前預警模型和事後偵查模型, 並產生相應的舞弊風險因子。在此基礎上, 本文將事前預警風險因子和事後偵查風險因子結合在一起, 構建出中國大陸上市公司現金舞弊綜合甄別模型,該模型總體判別準確率進一步提高到83.07%。
This study screens 133 observations of suspected cash frauds from listed companies in mainland China that were punished for financial violations over the period 2010 to 2019, and 404 non-fraud observations are also selected as the matched sample. A cascaded logit approach is adopted to construct the fraud-detecting model. First of all, based on the fraud trinagle theory, this study employs proxy variables to measures incentives or pressures, opportunities and rationalization (or justification or attitude) and further tested whether these proxy variables are related to the occurrence of cash frauds. The cash fraud pre-warning model is then built on these implicit factors for cash fraud. Secondly, based on the abnormal performance of various financial ratios found in the financial statements of companies that have committed cash fraud, this study is able to find 10 financial and nonfinancial varibles. The cash fraud post-detection model is then built on these explicit factors for cash fraud. The pre-warning model and the post-detection model represent the first-level model in the cascaded logit analysis, and each of these first-level model generates a risk index. On this basis, this study incorporates these two risk indices into the second-level model to construct an integrative fraud-detecting model. The overall discrimination accuracy is further improved to 83.07%.