Do Internal Governance Mechanisms Impact on Firm Performance? Empirical Evidence from the Financial Sector in China

Abstract

The financial sector plays an important intermediary role in the Chinese economy. However, there has been very limited research concerning improvement in corporate governance within this sector. Using an unbalanced data set of 139 firm-year observations covering 1999 to 2009, this study examines the impact of internal governance mechanisms on the performance of Chinese listed financial institutions. Findings suggest that state ownership, legal person ownership, board size, and supervisory board meetings are negatively related to the profitability of these institutions, whereas factors including ownership concentration, foreign ownership, independent directors, board meetings, and supervisory board size have no impacts.

Keywords:

• financial institution
• ownership structure
• board of directors
• supervisory board
• China

ACKNOWLEDGMENT

The authors would like to thank Bryan Howieson and Chris Graves of University of Adelaide Business School for the valuable comments on an earlier draft of this article. They also gratefully acknowledge the anonymous reviewers for the constructive comments and suggestions, which have significantly improved the contents of this article.

Notes

1. According to Gunasekarage, Hess, and Hu (2007) and Ma, Naughton, and Tian (2010) ownership concentration in this study is measured as the proportion of shares held by the top-10 shareholders. This measurement is appropriate because most Chinese listed companies only show the top-10 shareholders in their annual reports.

2. The critical value of the VIF to test for multicollinearity is 10. Gujarati (2003) suggested that there is no evidence of multicollinearity unless the VIF of a variable exceeds 10. All values used in this study were well below this critical level.

3. Heteroskedasticity is very common in panel data (Baltagi, 2005). In this study we use the Breusch-Pagan Lagrange Multiplier (LM) test to detect heteroskedasticity. The results indicate that both models, measured by ROE and Tobin's Q, encounter an unknown nature of heteroskedasticity. Thus, heteroskedasticity robust standard errors are computed so that the t and F statistics remain valid (Wooldridge, 2006). The standard errors in the regression analyses of this study are corrected using the White cross-section covariance method.

4. In binary logistic model we coded as 1 if TOBINSQ/ROE is greater than the median value of the sample, otherwise coded as 0. Removing extreme variables may reduce the representational faithfulness of the sample, thereby undermining the validity of the results. Hence, we utilize binary logistic regression analysis as an alternative technique which is robust when using non-normal data (Hair, Anderson, Tatham, & Black, 1998).

5. The association between TOBINSQ and LEGALPERSON is weakly inverse in Model (3) of binary logistic regression, which is indicated by z = −1.661 and p < 0.10.

6. The Company Law, 1995, 2004 & 2005 Amendments, Article 65-71, The People's Congress of China.

7. CBRC 2010 No.7 Decree, Provisional Rules over Assessing Directors' Performance at Commercial Banks, December 10, 2010.