Research Design
This study deploys a “Static-Group Comparison Design” with two groups of firms being investigated. The two groups have been formed based on one given criterion, that is, firms with reported fraud and those without reported fraud and are matched on a one-to-one basis for their industry, size and time period covered in this study. The study uses a multivariate logit regression analysis for that the dependent variable is a dichotomous variable with two levels, fraud vs. no-fraud firms (see other studies using the same method, for example, Stone and Rasp, 1991; Beasley, 1996; Sharma, 2004).
A logit regression model was developed and presented below.
FRAUDi = α+β1%State +β2%Legal +β3%BoDshare +β4%NoComD +β5DUAL + β6%NoComS +β7DEBT +β8%ComPro + εi
Where:
i = firm 1 through 78;
FRAUD = a dummy variable with a value of 1 when a firm is alleged to have committed fraud and a value of 0 otherwise;9
α = constant term;
%State = the percentage of state shares in the firm;
%Legal = the percentage of legal person shares in the firm;
%BoDshare = the percentage of BOD shares in the firm;
%NoComD = the percentage of directors without remuneration from the firm;
DUAL = a dummy variable with a value of 1 when the board chair holds CEO position in the firm and a value of 0 otherwise;
%NoComS = the percentage of supervisors without remuneration from the firm;
DEBT = the percentage of total debt to total assets of the firm;
%ComPro = the percentage of total remuneration of top management (including directors, supervisors and senior executives) to total net profit per annum of the firm;
ε = the error term.
6. Empirical Results and Discussion
Descriptive Statistics
Observations on the nature of the frauds reported reveal that most of them concentrate on the fraudulent acts of management and falsification of financial statements. For example, inside directors and top management were found to have made improper use of their positions for their own benefits, misappropriated firms’ assets, or channelled corporate resources to other entities under their control. All of these frauds resulted in falsification of the financial records (i.e., financial statement fraud, the definition applied in this study).
Table 2 contains descriptive statistics of the independent variables investigated. A brief examination of the statistics (Mann-Whitney z-value) reveals that three out of eight variables show significant differences between the fraud and no-fraud firms, that is, %State (p<0.01), DEBT (p<0.05), and %ComPro (p<0.10). Further observation reveals some typical characteristics of Chinese listed firms. For example, the state shares, on average, amount to about one-third (27.4% for fraud firms and 43.2% for no-fraud firms) and legal person shares nearly one-third (29.2% for fraud firms and 24.5% for no-fraud firms) of the total share issues, which confirms the discussion in Section 2 on the ownership structure of Chinese firms. The BOD shareholdings are extremely low, an average of 0.031% of total share issues for fraud firms and 0.079% for no-fraud firms. This is primarily due to two reasons: the regulatory restriction on shares being issued to firms’ management and the political and official background of BODs. In addition, management remuneration as a percentage of the net profit seems to be low, 1.1% for fraud firms and 2.9% for no-fraud firms. Again, this is another feature of Chinese characteristics. The market for managers has not been established and the appointment of top management is largely a non-economic decision.
(Insert Table 2 here)
Pearson correlations were performed for all variables.10 An examination of the correlation matrix reveals three significant correlations between %State and FRAUD (r = -.247, p<0.05), %State and %Legal (r = -.870, p<0.01), and %NoComD and %NoComS (r = .369, p<0.01). The significant correlation between %State and FRAUD indicates a strong negative relationship between the state shareholdings and the likelihood of corporate fraud. This correlation is not of a problem when data is analysed as FRAUD is the dependent variable. The correlation between %State and %Legal is a strong and negative one, which indicates that a higher (lower) percentage of state ownership is associated with a lower (higher) percentage of legal person shareholdings. This result is consistent with the ownership structure of Chinese listed firms. The significant correlation between %NoComD and %NoComS is not a particular concern as the coefficient (r = .369) is below 0.50, the threshold for collinearity concern as recommended by Stone and Rasp (1991). 公司治理与企业舞弊英文文献和翻译(11):http://www.youerw.com/fanyi/lunwen_245.html