The confounding effect of cost stickiness on conservatism estimates☆
Introduction
Empirical research on conditional conservatism documents a piecewise linear relation between earnings and stock returns and often attributes this pattern to a more timely recognition of bad news than good news (e.g., Basu, 1997). However, we argue that a similar asymmetric relation can arise from a fundamentally different source—cost stickiness. Cost stickiness refers to the asymmetric response of costs to sales increases versus decreases, which is a widely documented finding in the research on cost behavior (e.g., Anderson et al., 2003). We show that cost stickiness can be mistaken for conditional conservatism in standard models, distorting inferences about both the average level of conservatism and the extent of cross-sectional variation in conservatism. To reduce the likelihood of spurious inference, we develop new empirical tests that separate conservatism from the confounding effect of sticky costs.
Conditional conservatism implies that bad news (the usual proxy for which is negative stock returns) is recognized in earnings more quickly and more fully than good news (measured by positive returns). Therefore, the contemporaneous association between earnings and stock returns should be stronger for negative returns than for positive returns (Basu, 1997). Numerous studies find results consistent with this prediction and infer the presence of conditional conservatism (e.g., Pope and Walker, 1999, Ball et al., 2000, Watts, 2003b, Ball et al., 2013a).
The research on sticky costs documents an asymmetric behavior of costs and attributes it to operational decisions by managers who face adjustment costs, such as severance payments to dismissed workers and disposal costs for property, plant, and equipment (e.g., Anderson et al., 2003). To avoid adjustment costs, managers retain some unused resources when sales fall. By contrast, when sales rise, managers have to add sufficient resources to meet the demand. Because of this operational asymmetry, costs are “sticky” on average—they rise more in response to sales increases than they fall for sales decreases.
The cost asymmetry implies that, on average, earnings respond more to sales decreases than to sales increases. Because sales changes and concurrent stock returns are positively correlated, cost stickiness leads to an asymmetric relation between earnings and stock returns, which is stronger for negative returns than for positive returns. Therefore, the standard estimates of asymmetric timeliness are likely biased upwards on average when sticky costs are prevalent. Further, because cost stickiness varies systematically (e.g., Weiss, 2010), it leads to variation in the bias, which is likely to distort inferences about the extent of cross-sectional variation in conservatism. To control for this confounding effect, we incorporate cost stickiness in standard conservatism models.
We use data for publicly traded US firms from 1987 to 2007 to examine the confounding effect of cost stickiness in the Basu (1997), Ball et al. (2013b), and Collins et al. (2014) models.1 Controlling for the asymmetric effect of sales changes reveals an average bias in conservatism estimates of more than 25%. Further tests confirm that the confounding asymmetry for sales changes is consistent with manifestations of cost stickiness and is distinct from conservatism. Thus, the upward bias in the standard conservatism measures arises because the measures mistake cost stickiness for conditional conservatism. We also find that controlling for the variation in cost stickiness has a large impact on estimates of the variation in conservatism. For example, the correlation between industry-level conservatism estimates (at the three-digit SIC level) with and without controls for cost stickiness is just 51.3%, and the difference between the two estimates is statistically significant in 49% of the industries. Thus, the use of standard models leads to mistaken inferences not only about the average degree of conservatism but also about how much conservatism varies cross-sectionally.
Cost stickiness also confounds inferences about the correlates of conservatism. For example, we find that the standard estimates of the association between managerial stock ownership and conditional conservatism contain a large bias. These estimates become statistically insignificant when we mitigate the bias. Controlling for cost stickiness also significantly changes the estimated effects of the standard proxies of the demand for conservatism, such as size, leverage, and the book-to-market ratio. Thus, recognizing the confounding effect of cost stickiness is essential to ensure accurate inferences in future research.
In Section 2, we develop our argument for the correlated omitted variable bias and formulate the hypotheses. In Section 3, we describe the data and the estimation models. Section 4 presents our main empirical results and the additional validation tests. Section 5 further illustrates the confounding effect of cost stickiness. Section 6 concludes.
Section snippets
Hypothesis development
We build on both the research on conservatism in financial accounting (e.g., Basu, 1997, Pope and Walker, 1999, Ball et al., 2000) and the research on sticky costs in cost accounting (e.g., Anderson et al., 2003, Weiss, 2010, Chen et al., 2012). Conservatism and sticky costs are fundamentally different phenomena: conservatism is asymmetry in the information system that translates economic activity into accounting data, whereas cost stickiness is asymmetry in the economic activity itself.
Sample selection and descriptive statistics
Our sample selection procedure is summarized in Panel A of Table 1. We start with the Compustat/CRSP annual sample from 1987 to 2007.3
Estimation results
Table 2 presents the pooled estimates for the Basu model, the sticky cost model, and our full model. The sticky cost model has greater explanatory power than the Basu model (the adjusted R2s are 19.0% and 12.7%, respectively). Column 1 shows a significant positive coefficient on DR×RET, which indicates conservatism, and column 2 shows a significant positive coefficient on DS×ΔS/P, which indicates cost stickiness.
For the full model in column 3, the adjusted R2 improves to 26.8%. Both the Basu
Empirical illustration: the confounding effect of cost stickiness on the association between managerial ownership and conditional conservatism
To further illustrate the importance of controlling for cost stickiness, we examine the negative association between managerial stock ownership and conservatism that is documented by LaFond and Roychowdhury (2008).9
Conclusion
We examine the confounding effect of cost stickiness on conditional conservatism estimates. Because of sticky costs, earnings behave asymmetrically for operational reasons. This operational asymmetry can distort standard asymmetric timeliness estimates. We combine the theory of conservatism from financial accounting with cost stickiness from cost accounting to develop empirical hypotheses. We propose extensions of the standard conservatism models to control for the confounding effect of sticky
References (46)
- et al.
The effect of international institutional factors on properties of accounting earnings
Journal of Accounting and Economics
(2000) - et al.
Employment protection legislation, adjustment costs and cross-country differences in asymmetric cost behavior
Journal of Accounting and Economics
(2013) The conservatism principle and the asymmetric timeliness of earnings
Journal of Accounting and Economics
(1997)Conservatism research: historical development and future prospects
China Journal of Accounting Research
(2009)- et al.
Cash flow asymmetry: causes and consequences for conditional conservatism research
Journal of Accounting and Economics
(2014) - et al.
Acquisition profitability and timely loss recognition
Journal of Accounting and Economics
(2010) - et al.
Estimation and empirical properties of a firm-year measure of accounting conservatism
Journal of Accounting and Economics
(2009) - et al.
Creditor control rights and firm investment policy
Journal of Financial Economics
(2009) Discussion of “Cash flow asymmetry: causes and consequences for conditional conservatism research”
Journal of Accounting and Economics
(2014)The role of information asymmetry and financial reporting quality in debt trading: evidence from the secondary loan market
Journal of Accounting and Economics
(2008)
The contracting benefits of accounting conservatism to lenders and borrowers
Journal of Accounting and Economics
Evidence on the role of accounting conservatism in monitoring managers’ investment decisions
Accounting and Finance
Are selling, general, and administrative costs “sticky”?
Journal of Accounting Research
Econometrics of the Basu asymmetric timeliness coefficient and accounting conservatism
Journal of Accounting Research
On estimating conditional conservatism
The Accounting Review
Asymmetric cost behavior
Journal of Management Accounting Research
The moderating effect of prior sales changes on asymmetric cost behavior
Journal of Management Accounting Research
Predicting earnings using a model based on cost variability and cost stickiness
The Accounting Review
Conservatism and the asymmetric timeliness of earnings
Conditional and unconditional conservatism: concepts and modeling
Review of Accounting Studies
Published corporate accounting data and stock prices
Journal of Accounting Research
Capital allocation and timely accounting recognition of economic losses
Journal of Business Finance and Accounting
Cited by (139)
The misuse of regression-based x-Scores as dependent variables
2024, Journal of Accounting and EconomicsThe effect of corporate reputation on accounting conservatism
2024, Advances in AccountingLabor skill and accounting conservatism
2024, Journal of Accounting and Public PolicyAsymmetric cost behavior: Theory, meta-analysis, and implications
2023, Journal of International Accounting, Auditing and TaxationHow does credit risk affect cost management strategies? Evidence on the initiation of credit default swap and sticky cost behavior
2023, Journal of Corporate Finance
- ☆
We thank Paul Hribar (the reviewer), Eric Press, Anup Srivastava, Lucas Threinen, and seminar participants at Michigan State University, New York University, Seoul National University, Southern Methodist University, Temple University, Indian School of Business Accounting Conference, Temple Accounting Conference, Yale Accounting Conference, AAA annual meeting, FARS midyear meeting, MAS midyear meeting, and the AAA Mid-Atlantic Region midyear meeting for helpful comments and suggestions. We especially thank John Core (the editor) for constructive and detailed advice in developing the paper.