2.3 Empirical Methodology
2.3.3 Data Selection and Summary Statistics
Information about the annual financial accounting of Chinese listed companies from 1998 to 2011 is extracted from the China Stock Market Accounting Research (CSMAR) database. Although the publication of cash flow statements has been compulsory for all Chinese listed firms since 1998, the regressions in equations (2) to (7) require sales in year t - 2 and assets in year t - 1 to calculate the real and accrual-based earnings management proxies; the variables from cash
flow statements are therefore from 2000. Hovakimian (2009) excludes the financial industry, since investments in the form of capital assets vary significantly between financial and other industries. Therefore, this chapter also restricts the sample to non-financial firms with at least eight observations in each CSRC industrial code grouping per year. In addition, because corporate governance variables are available from the CSMAR database gradually from 2002, the final sample period for the regression analyses ranges from 2002 to 2011. After excluding firms with missing data for control variables and calculating discretionary accruals as well as real earnings proxies, this study obtains a total of 13,840 firm–year observations. There are 12,610 observations for RM_1(3), and 13,602 for RM_2, due to the data requirements to calculate them.
[Insert Table 2.1 around here]
Table 2.1 reports descriptive statistics of the final sample from 2002 to 2011. Panel A reports the sample distribution of firm-years. The number of firms dramatically increased from 2002 to 2011, reflecting the rapid growth of the Chinese stock market. Panel B provides summary measures of various accounting and financial variables in the regressions mentioned above. The median and mean of the majority of variables are quite similar to the results of Firth et al. (2007b), Gul et al. (2010), and Hou et al. (2012), although the sample periods are different in these studies. The mean of ROE is 0.036, with a standard deviation of 0.596. About 7% of firms use one of the international Big 4 auditors or their joint ventures. The mean and median values of the book-to-market (BM) ratio are very
close, 0.377 and 0.388, respectively. The ownership structure of Chinese listed firms is highly concentrated compared to the structure in developed markets; its mean is 38%, which is slightly lower than the results of Ding et al. (2007) and Gul et al. (2010), who find 41.9% and 42.8%, respectively. The difference might be due to the different sample periods and the decrease in the number of NTS held by controlling shareholders after the reform. The mean of STATE is 47.9%, which reveals that the state is the largest shareholder and continues to play an important role in Chinese listed companies. More importantly, Panel B also reports the descriptive statistics of earnings management variables. For the discretionary accrual part, the mean value is 0.001, which is close to the 0.002 reported by Firth et al. (2007b), and its median is 0.003, slightly higher than the mean. Regarding the three individual real earnings management proxies, the average of APROD is negative, with a zero median; ADISX has a negative median and the largest magnitude of the three individual real earnings management proxies. The median of the three aggregate real proxies are all positive and equal to 0.013.
Table 2.2 shows the Pearson correlation matrix for the variables in the main tests during 2002-2011. Consistent with our expectation, DA and RM are highly and positively correlated. Specifically, the correlation figures between DA and RM_2
and between DA and RM_3 are 0.492 and 0.309, respectively. This indicates that firms prefer to use both real and accrual-based earnings strategies to supplement each other. The three aggregate real earnings management proxies are highly correlated, in the range from 0.655 to 0.911. This result suggests that these proxies are indicative of real earnings management and can be substituted for each
other. They are also highly correlated with the three individual real earnings management proxies. For example, the correlation coefficient between RM_3 and the three individual measures is between 0.601 and 0.896, which is much higher than in the US market as reported by Cohen et al. (2008). Of the three individual real proxies, APROD has a positive correlation with DA and the aggregated proxies RM_1 and RM_3, while the correlations between ACFO and DA and between ACFO and RM_2 and RM_3 are all significantly negative. These results suggest that firms attempt to achieve high earnings by manipulating production activities and related cash flows. Consistent with the results of Firth et al. (2006), the compensation of the top three senior managers is significantly positively correlated with DA, but negatively correlated with real earnings management activities. A reverse relation can be found for leverage. The number of TS is positively correlated with real earnings management but insignificantly with accrual-based management. Overall, these preliminary results based on Pearson correlation are in line with previous studies, as well as with the hypotheses that listed firms in China may engage in both accrual-based and real earning strategies, and that their earnings management activities are correlated with firm characteristics and corporate governance features.
[Insert Table 2.2 around here]
To examine whether the incentive alignment of the split share structure reform can improve firm corporate governance in the Chinese market, this chapter tests the significance of differences in the corporate governance variables used in the study between the pre- and post-reform periods. Table 2.3 reports the mean values
of the corporate governance variables and the results of t-tests on the significance of the differences in variables between the two periods. Consistent with previous results and hypotheses, the results in Table 2.3 reveal an improvement in these corporate governance proxies. More specifically, ownership concentration significantly decreased. The largest shareholder’s holding declined by 5%, from 40% to 35%. Similarly, state ownership also significantly decreased, from 54.3% to 40.5%. Regarding board independence and CEO duality, a significant increase in the value of the former, and a decrease in the value of the latter, are observed.
These results indicate that the SSSREF has effectively improved the quality of firms’ corporate governance in China, where the institutional environment, particularly minority shareholder protection, is weaker than in the developed countries. Therefore, the regressions in this chapter include these variables to control for their effect on earnings management. However, the ownership of the largest shareholders in China is still high after the reform. Hence, this chapter perceives that the split share reform may not have sufficiently increased the quality of financial information, because controlling shareholders have an incentive to manipulate earnings to increase their own wealth by inflating the share price after their NTS are converted into TS.