Basu (1997) introduced the first, and currently the most popular, empirical measure of accounting conservatism, commonly known as the ‘Basu measure’. The resulting measure is described by Basu as the asymmetric timeliness of earnings coefficient
1Acknowledgment: A paper based on this chapter, co-authored with my thesis supervisors, is
currently under the 2nd review atJournal of Accounting & Economics. I thank the editor, Thomas Lys, and the anonymous referee of Journal of Accounting & Economics for their invaluable con- structive criticisms and suggestions. I am especially grateful to Robert M. Bushman for suggesting the option-pricing based method of estimating default risk. I also thank the numerous comments on earlier versions of this chapter by the following people: Michael Bradbury at the 2009 New Zealand Accounting PhD Students Colloquium; Richard Morris at University of New South Wales Research Seminar; Stephen L. Taylor at the AFAANZ 2009 Annual Meeting; and Peter Wells at the University of Technology Sydney 2009 Summer Accounting Consortium. I also with to extend my gratitude to all other people who have provided comments on earlier versions of this chapter in the above conferences and seminars, as well as in the EAA 2009 Annual Meeting, and AAA 2009 Annual Meeting.
(abbr. “AT” measure). Since Basu’s influential paper, a large and still growing liter- ature has emerged applying the Basu measure to examine accounting conservatism from a variety of theoretical perspectives.
However, the validity and characteristics of the Basu measure of accounting conservatism have received limited attention. Only recently, Dietrich et al. (2007), Givoly et al. (2007), Ryan (2006) and others have begun to directly examine the validity of the Basu measure. These recent studies have highlighted a number of weaknesses in the Basu measure. Dietrich et al. (2007), for example, find that the Basu measure is biased upward, because of what they call the sample-variance- ratio bias and the sample-truncation bias. Givoly et al. (2007) empirically test the validity of the Basu measure, and discover that the measure can demonstrate neither the power to distinguish conservative firms from aggressive ones, nor the stability expected in a time-series context.
The analysis in this chapter has two related objectives: First, I extend this recent critical appraisal of the Basu measure by investigating the relationship between the Basu measure and a firm’s default risk. Using Merton’s (1974) call-option pricing model of equity, I argue that the Basu AT measure is a biased measure of the degree of accounting conservatism. The higher the default risk, the higher the bias in the Basu AT measure. In general, default risk means the uncertainty around a firm’s ability to repay its debts when the debts fall due. In this chapter, I use Merton’s (1994)Distance-to-Default concept as the analytical definition of default risk.
Second, I use the insight provided by my analysis of the Basu measure to con- struct an improved version of the ‘Basu’ measure, and I call the new measure the
Default-Adjusted-Basu (“DAB”)measure, because it makes adjustments for the ef- fects of default risk on the Basu measure. I contend that the DAB measure can substantially reduce the bias caused by default risk, and hence is a more robust
measure of accounting conservatism than is the standard Basu measure.
Empirically, I adopt Vassalou and Xing’s (2004) iterative procedure for estimat- ing firms’distance-to-default. I find that firms with higher default risk indeed tend to have a higher Basu measure of conservatism, consistent with my analytical pre- diction. I also test the validity of the DAB measure of conservatism empirically, and the result suggests the the DAB measure likely provides a more robust measure of accounting conservatism than does the original Basu measure.
My analysis of the Basu measure bears a close relationship to the analysis of the earnings response coefficient (ERC) by Dhaliwal et al. (1991). Dhaliwal et al. (1991) show that ERC is negatively correlated with the default risk of the firm. Since the Basu model is essentially a reversed ERC model, the negative associa- tion between ERC and default risk implies that there exists a positive association between Basu regression coefficients and default risk. This positive association is exactly what this chapter attempt to analyze. In addition, this chapter has another similarity with Dhaliwal et al. (1991) — both papers use the classic Merton (1974) model as the analytical foundation.
There are, however, also major differences between this chapter and Dhaliwal et al. (1991). The most significant difference is that Dhaliwal et al. (1991) treat the value of the firm as a function of earnings, but this chapter treats earnings as a function of the changes of the value of the firm. This difference is perhaps the defining characteristic that sets the Basu model apart from the more traditional ERC models.
Before I proceed to the main analysis, I briefly introduce the Basu measure itself. The Basu (1997) measure is based on a cross-sectional, piece-wise regression of accounting earnings on stock returns, as follows:
EPSit
Pi,t−1 =α0+α1DRit+β0Rit+β1RitDRit+εit (2.1)
where
EPSit : Earnings per share for firm i year t
Pi,t−1 : Opening stock market price for firm i year t
Rit : Stock market return for firm i year t
DRit : Dummy variable that is equal to 1 if the stock market return for firm i
year t is negative, and equal to 0 if the stock market return for firm i year t is non-negative.
The regression model above, known as the “Basu model”, regresses accounting earnings (EPS/P) on stock returns (R) separately for ‘good-news’ and ‘bad-news’ firms. A firm-year is deemed a ‘good-news’ one if the return on its stock return is positive or zero. Likewise, a firm-year is deemed a ‘bad-news’ one, if its stock return is negative. By using the dummy variable, DRit, the Basu model allows the slope coefficients to differ between the good-news and bad-news groups (β0
and β0+β1, for good- and bad- news coefficients respectively). The difference
between the bad- and good- news timeliness coefficients,β1, is theBasu asymmetric
timeliness coefficient, which measures the degree of conservatism in the sample of firms.
The rest of this chapter proceeds as follows: Section 2.2 examines analytically how default risk impacts on the Basu measure. Section 2.3 develops the Default- Adjusted-Basu (DAB) measure of accounting conservatism, which, I argue, is more robust to default risk than is the original Basu measure. Section 2.4 discusses the sample selection and the proxies used in the empirical tests. Section 2.5 reports the main empirical results with respect to both the original Basu measure and the DAB
measure. Section 2.6 reports the results of the robustness tests. Finally, I conclude the chapter in Section 2.7.