Problem directors on the audit committee and financial reporting quality

Abstract

The objective of this paper is to examine empirically the consequences for financial reporting quality of having audit committees that include problem directors, that is, directors with prior involvement in corporate bankruptcies, major accounting restatements, or other accounting scandals. An ordinary least squares regression model is used to examine the association between problem directors on the audit committee and financial reporting quality as proxied by accruals and real earnings management. Results reveal that there is a positive association between the presence of problem directors on the audit committee and real earnings management, and this association is more pronounced in cases where those problem directors have been involved in prior instances of accounting restatements and fraudulent reporting practices.

Keywords:

• corporate governance
• problem directors
• discretionary accruals
• real earnings management

Acknowledgements

We thank the Associate Editor and two anonymous reviewers for many constructive comments. This paper also benefited from comments from the participants of the 2014 Accounting and Finance Association of Australia & New Zealand (AFAANZ).

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. For a discussion on how to better model simultaneously real and accruals-based earnings management, see Walker (2013, p. 455).

2. Roychowdhury (2006, p. 341) suggests that price discounts to boost sales, and overproduction, reduce abnormal CFO, while cutting discretionary expenditures increases abnormal CFO. The net effect on abnormal CFO, therefore, is ambiguous.

3. Following the econometrics literature, we test which empirical model: pooling, random effect, or fixed-effect regression; is most suitable for estimating the relationship. We conduct the Lagrangian Multiplier (LM) test of the random-effect model and Pooling OLS (Breusch and Pagan 1980). The null hypothesis for the LM test is that the individual effect, a_i, is 0 for all i. In all of our models, the null hypothesis is rejected (χ² value is 3.40, p-value.07), which suggests that random-effects regression is appropriate for the REM model. Then, we follow the Hausman (1978) test to choose our model between fixed effect and random effect. The null hypothesis is that fixed effect is not correlated with the regressor (or our main independent and control variables). We reject the null hypothesis in the REM model (χ² value is 90.54, p-value.001) suggesting that the fixed-effect model is appropriate.

4. We report White (1980) t-statistics for this model because of the small sample size (Gow et al. 2010). Still results hold if standard errors are computed using a two-way cluster.