The Effect of Cost Stickiness on Peer-Based Valuation Models

Abstract

We investigate how asymmetric cost behavior (also termed cost stickiness) affects peer-based valuation models. Using a sample of U.S. firms for the period 2000–2016, we provide evidence that the higher the degree of a target firm`s level of cost stickiness vis-à-vis its peer group, the greater is the peer-based underestimation of the firm’s market value (downward peer-based valuation bias). Furthermore, we show that the underestimation of firm value is weaker for firms exhibiting potential for agency issues. Overall, our findings suggest the following: First, using information on cost management strategies reduces the downward peer-based valuation bias. Second, investors partially understand and incorporate available information about cost stickiness into their evaluation of firm value, c.p. leading to deviations from the value estimates of the peer-based models. These models generally do not recognize information on cost stickiness as an input factor. Ultimately, the findings support the assumption that investors assess the likelihood that cost stickiness is either driven by economic or non-economic reasons and adapt their value estimate accordingly.

Keywords:

• Cost Stickiness
• Peer-Based Valuation
• Capital Market

JEL:

• M41
• G32

Acknowledgements

We thank Beatriz García Osma (Editor), two anonymous reviewers, Dmitri Byzalov (discussant), the attendees of the 2020 European Accounting Review Annual Conference, Lisa Silge, Friedrich Sommer as well as workshop participants at the University of Gießen for their helpful comments and suggestions. This work was presented at the 2020 European Accounting Review Annual Conference.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Although the occurrence of cost stickiness is well established, its fundamental cause is subject to controversial debate (Ciftci et al., 2016; Silge & Wöhrmann, 2019; Weiss, 2010). For a detailed overview of the discussion see Banker et al. (2018).

2 M. C. Anderson et al. (2007) empirically show that cost stickiness is more pronounced if managers have optimistic expectations (and costs are anti-sticky if managers have pessimistic expectations about the future). Thus, an increase in cost stickiness provides positive information about managers’ expectations. This information could be used by investors when assessing market values.

3 The average market capitalization in our sample is 7.433 billion US$. Therefore a 6.3% (standardized regression coefficient for delta sticky using Model 1c) reduction is equal to approximately 470 million US$.

4 Anti-sticky means that costs increase less with increasing activity than they decrease with decreasing activity.

5 For a detailed overview of the cost stickiness research and its implications see Banker et al. (2018)

6 The authors regard an increase as intended if a firm’s past SG&A ratio was below its industry average, which likely represents efficiency in SG&A cost management. In contrast, an above industry average SG&A ratio is interpreted as deficient cost control.

7 Adjustment costs include monetary and social components (Banker & Byzalov, 2014). The first category includes retrenchment costs (e.g. severance payments, costs for the disposal of capital equipment or penalties for early termination of leases) as well as restoration costs (e.g. employee search and training costs, costs for the installation of capital equipment, facility-opening costs). Social adjustment costs include non-monetary effects, such as the loss of work ethic, the reduction of human capital through the dissolution of existing team structures and the loss of corporate reputation.

8 Banker et al. (2018, p. 197) describe the underlying rational as follows: ‘If the demand for different firms’ output is positively correlated (e.g., all firms face high demand during macroeconomic booms and low demand during recessions), then each firms’ demand in the output market is positively correlated with the aggregate demand in the resource market. Therefore, the firms will tend to face higher resource prices during firm-level demand increases and lower resource prices during firm-level demand decreases.’

9 For an overview the discussion on the determinants of cost stickiness see Banker et al. (2018)

10 The probability of a sales increase, in any given period, is higher than the probability of a sales decrease (M. C. Anderson et al., 2007). Banker and Byzalov (2014) also show that 72.1 percent of firms have an increasing sales trend and only 27.9 percent have a decreasing trend. Consequently, slack that was retained during a past sales decrease is typically absorbed by sales growth in subsequent periods.

11 Among the key drivers of firm value are risk, growth and profitability (An et al., 2010; Damodaran, 2006).

12 Computing peer group multiples using the median increases valuation accuracy compared to other aggregation methods, e.g. arithmetic mean (Baker & Ruback, 1999; Liu et al., 2002; Schreiner & Spremann, 2007; Sommer et al., 2014).

13 $\mathrm{\Delta }=$ target value – peer group value.

14 This line of reasoning is supported by Bhojraj and Lee (2002), who use the fundamental factors growth, risk and profitability in order to construct accurate peer groups.

15 Note, the relative valuation error can take positive and negative values, indicating over- and underestimation respectively. The negative link between $\mathrm{\Delta }SIZE$ and the valuation error for example expresses that ceteris paribus the value of larger firms tend to be underestimated by peer-based valuation models and firm value of smaller firms tend to be overestimated.

16 Note, we expect that these two factors are sufficient to control for differences in risk levels due to the use of industry peer groups. The general operating risk structure of firms in the same peer group should be comparable.

17 This finding differs from the findings of other studies (e.g., Weiss, 2010), that report firms on average exhibit sticky cost behavior. Our untabulated analysis reveals that this difference is driven by the elimination of firms reporting negative values for EBITDA, EBIT or IB. Since assessing the value of firms using negative value indicators seems highly impractical and will likely distort our findings, we do believe that the elimination of these observations is beneficial to the validity of our findings.

18 An alternative explanation might be that our measure of stickiness is derived based on the difference between sales and IB. As a consequence our ΔSTICKY measure might be a weaker proxy for the impact on the EBIT or EBITDA multiple.

19 Research provides mixed results for the impact of firm size on earnings characteristics (e.g. earnings quality). Some studies predict that firm size would be negatively associated with earnings quality due to income-decreasing accounting method choices used by larger firms in response to greater regulatory scrutiny (Jensen & Meckling, 1976; Watts & Zimmerman, 1990). This is a possible explanation for our finding.

20 A possible explanation might be that capital markets adjust value estimates for mean reversion effects (Blume, 1975).

21 See Table 5 for a complete list of the variables.

22 Same approach as in the main analysis.

23 While our results show that the downward valuation bias is reduced if cost stickiness is driven by agency issues, the overall effect of cost stickiness when agency issues are present is less clear. We find that the sum of the coefficient on $\mathrm{\Delta }PG\mathrm{\_}STICK{Y}_{j,t}$ and the coefficient on OPCF_j,t×$\mathrm{\Delta }PG\mathrm{\_}STICK{Y}_{j,t}$ is significantly negative in Model 2a (p = 0.015), 2b (p < 0.01), and 2c (p < 0.01). This implies that cost stickiness in the presence of agency issues may lead to peer-based valuation estimates that exceed market values (or at least reduce the downward valuation bias) probably because investors understand that cost stickiness in this case destroys value. If we conduct the same test for SLACK_j,t,, however, we find insignificant effects for Model 2a (0.874) and 2b (0.565) but a significantly positive effect for Model 2c (p < 0.01). A potential explanation for these different findings is that cash flow variables also used in e.g., Chen et al. (2012) are a better proxy for agency issues.

24 D_STICKY_j,t is a dummy variable that takes a value of 1, if the observations’ costs can be classified as sticky and otherwise 0. This is true for observations with STICKY_i,t values of < 0.

25 In line with our main tests for H1, we find that the sum of the coefficient on $\mathrm{\Delta }PG\mathrm{\_}STICK{Y}_{j,t}$ and the coefficient on D_STICKY_j,t×$\mathrm{\Delta }PG\mathrm{\_}STICK{Y}_{j,t}$ is significantly positive in Model 4a (p = 0.01), Model 4b (p = 0.01), and Model 4c (p < 0.01).