Test Results
The multivariate results in Table 3 show that the logit regression model for testing the hypotheses has a good fit with a pseudo R2 of 0.288 (based on the more conservative Cox and Snell measures). The model Chi-square of 19.86 is statistically significant at p < 0.01 indicating that the model coefficients are not equal to zero. The overall classification accuracy is 75.8 percent with 80.7 percent on fraud firms and 67.1 percent on no-fraud firms. These results show that three of the eight variables investigated, %State, %Legal and DEBT, are statistically significant and the other variables are not significant. However, %State and %Legal are negatively significantly correlated with the dependent variable, FRAUD, while DEBT is positively significantly associated with FRAUD.
Hypothesis 1 considers the relation between the percentage of state shares and corporate fraud in a null form. The result in Table 3, however, shows a strong negative relation with a coefficient of –2.663 (p < 0.01). This result suggests that the higher (lower) the percentage of state shares in a firm, the lower (higher) the likelihood that the firm would commit management perpetrated financial statement fraud.
Hypothesis 2 states that there is no relation between the percentage of legal person shares and fraud. The result in Table 3 for the %Legal variable shows that corporate fraud is significantly negatively affected by this variable, with a coefficient of –2.208 (p < 0.05). Similar to the %State variable, this result indicates that the higher (lower) the percentage of legal person shares, the lower (higher) the likelihood of fraud. However, as the two variables, %State and %Legal, are significantly negatively correlated, caution should be exercised when interpreting the results of these two variables.
Hypothesis 7 is about the relationship between the level of debt and the likelihood of fraud. The result in Table 3 for the DEBT variable shows a significantly positive relationship with a coefficient of 2.111 (p < 0.05). This result means that as the level of leverage increases, the likelihood that management would commit financial statement fraud increases. This finding is consistent with previous studies such as, Dechow et al. (1996) and Sharma (2001) using the western context.
(Insert Table 3 here)
The remaining variables, that is, the percentage of the BOD shareholdings (%BoDshare), the percentage of directors without remuneration (%NoComD), the chair/CEO duality (DUAL), the percentage of supervisors without remuneration (%NoComS) and the top management remuneration as a percentage to the net profit (%ComPro)11, have no effects on the likelihood of corporate fraud.
Discussion
As an exploratory attempt, the results of this study should be interpreted with caution. One key finding of the study is that state shareholding leads to a low occurrence of corporate fraud. However, in light of the unique features of the ownership structure in China, the high percentage of state ownership should not be interpreted as if it would lead to good corporate governance for the following reasons.
Firstly, the predominance of state ownership and control in the listed firms inherits the problem of ambiguous property rights associated with state ownership in the former SOE system. Under the former SOE system (i.e., absolute state ownership), property rights belong to everyone (the whole people) and to no one in particular. As a result, neither so-called principal (state controlling authorities) nor the agent (firms’ management) has adequate incentives to ensure the most efficient use of assets under their control. Such a state ownership undermines the incentives and disciplinary mechanisms essential to corporate governance. According to Lin (2001), by turning SOEs into JSCs through corporatisation and listing JSCs on the stock market does not solve the principal-agent problem arising from the ambiguity of property rights; neither does it improve the incentives and disciplinary mechanisms. 公司治理与企业舞弊英文文献和翻译(12):http://www.youerw.com/fanyi/lunwen_245.html