The Value Relevance of Cost Stickiness

(1)

The Value

Relevance of

Cost Stickiness

Master Thesis, Msc Controlling

Rijksuniversiteit Groningen

Faculteit Economie en Bedrijfskunde March 20, 2015

[

WIETZE-JAN HAISMA

]

Student number: 2441179 Postal code: 9233 KN

Street: Domela Nieuwenhuisstraat 51 Adress: Boelenslaan

E-mail: w.j.haisma@student.rug.nl Phone: 06-2346461

(2)

2

Abstract

This thesis studies the relationship between cost stickiness and share prices, and thereby the influence of cost stickiness on investors’ decisions. Cost stickiness is the phenomenon which describes the reaction of costs on changes in activity, where the reaction of costs is dependent on the change in direction. By employing a price model, this study finds a relationship between cost stickiness and share prices. Furthermore, this study shows that cost stickiness moderates the relationship between book value per share and share price, and the relationship between earnings per share and share price. From these results it can be concluded that investors take cost stickiness into account, and that preparers of annual statements should inform investors of the causes of cost stickiness in their firms to reduce information asymmetry and thereby reduce the impact of cost stickiness.

(3)

3

1. Introduction

Firms commit to long term resources before knowing about demand (Cooper & Kaplan, 1992). Variable costs follow demand and therefore they are considered to respond symmetrically to activity levels. That is, when activity levels change, the response of costs is not dependent on the direction of that change (Banker & Byzalov, 2014).

The symmetrical response, or proportional relationship between costs and activity, has been subject of debate. Noreen & Soderstrom (1997) found that overhead costs are not proportional to activity levels. The implication of this result is that decisions based on the assumption of proportional costs can lead to unexpected results. Anderson, Banker & Janakiraman (2003; henceforth ABJ), build on this research and find that selling, general and administration costs are ‘sticky’. The relatively new concept of cost stickiness has shown that costs do not necessarily have a symmetrical link with activity levels (ABJ).

Sticky costs differ from variable costs in their relationship with activity levels. They are adjustable on the short run, but not without incurring adjustment costs (Banker & Byzalov, 2014). Sticky costs are not mechanically determined, as variable costs are. They are instead the result of managerial discretion (Banker & Byzalov, 2014).

Cost stickiness occurs due to managerial discretion (Banker & Byzalov, 2014). Managers decide to retain costs (capacity) when faced with a decrease in activity, either due to high adjustment costs, uncertainty, or agency problems. Therefore, the presence of sticky costs can be a signal of poor cost structure, poor corporate governance, or optimistic management (ABJ; Chen et al. 2010). As

Baumgarten, Bonenkamp & Homburg (2010) show, sticky costs can also be the result of efficiency improving investments which means it’s not necessarily a negative phenomenon.

While previous research into cost stickiness has explored the multitude of causes of cost stickiness, this paper will look at a possible consequence: decisions made by investors. Banker & Chen (2006) argue that the extent to which investors have rational expectations about the impact of cost stickiness on future earnings remains unclear. As will be discussed in the following sections, cost stickiness can make the link between activity levels and earnings less clear. One could argue that this should have an effect on investment decisions by shareholders, as earnings are an important piece of information on which they base their decisions (Marquardt & Wiedman, 2004). Therefore, the aim of this paper is to examine whether cost stickiness is relevant information for investors. As cost

stickiness can stem from a multitude of causes, with both positive and negative consequences, it is unclear whether investors are able to discern between the various causes of cost stickiness and if yes, how they handle cost stickiness. If cost stickiness is part of the information used by investors to make investment decisions, then management should explore the causes of cost stickiness in their firms and address these in their communication with the investors.

The question remains whether investors are able to discern between the various causes of cost stickiness. Previous studies in the field of value relevance have shown that earnings have a significant influence on share price (Ohlson, 2001). As cost stickiness dilutes the link between activity and earnings (Holzhacker, Krishnan & Mahlendorf, 2014), investors might pay less attention to earnings and use other information instead. The notion of cost stickiness shows that costs do not move in proportion to sales and therefore investors have to look beyond earnings. The amount of extra information required might be influenced by the level of cost stickiness. This paper has some

(5)

5

practical implication to provide some insights about investors’ behaviors on whether they take cost stickiness into account. If so, firms might provide extra information explaining why their costs show sticky behavior.

This study may contribute to the literature in several ways. First, the results of this study might be vital for prepares of annual statements to understand the reactions of shareholders on the

information presented in the statements. As Baumgarten et al. (2010) find, in certain situations cost stickiness is a signal of investments aimed at improving operating efficiency which should lead to higher future earnings. Cost stickiness is not necessarily a negative concept. Retaining capacity when a decrease in activity is temporarily might be more cost effective than continuously adjusting

capacity to demand. However, investors can be unaware of the justification of sticky costs.

Besides showing the possible asymmetric relationship between costs and activity levels, an important finding in itself, the research following ABJ has been informative to other fields as well such as that of corporate governance (Chen, Lu & Sougiannis, 2010). Chen et al. (2010) find a relationship between cost stickiness and agency problems, pointing to increased inefficiency and thus lower earnings. Therefore, this paper may also contribute to the field of corporate governance.

The link between cost stickiness and corporate governance is also apparent in other research. Kama & Weiss (2013) find that when faced with the proper incentives, managers do cut costs more easily when sales volume decreases. This suggests that managers can influence the stickiness of costs, as long as they are motivated to do so. Incentive systems can moderate the relationship between cost stickiness and empire building, as found by Chen et al. (2010). From this finding, one might infer that low cost stickiness is one of the results of a properly working corporate governance system. While corporate governance is not explicitly investigated in this study, cost stickiness might be a factor to be used in judging the quality of performance measurement systems.

As argued above, cost stickiness can deteriorate the relationship between activity levels and earnings. Furthermore, cost stickiness can be a symptom of empire building which has negative consequences for investors. This leads to the formulation of the following research question:

Is cost stickiness value relevant information?

This paper shows that cost stickiness indeed is a value relevant piece of information. In a sample of 24.058 firm-quarter observations, the relationship between cost stickiness and share price is found to be negative. As will be discussed in this paper, these results point to the fact that investors indeed are aware of cost stickiness or at least are influenced by its effects on earnings. With these results, this paper adds to the literature on cost stickiness proving that it is of relevance to management. Cost stickiness has an influence on stock performance, something management should take into consideration in the decision making process.

The next section will discuss the literature on cost stickiness and inform the hypotheses. The third section describes the dataset and methods used in this study. The fourth section presents the results, and the fifth section will provide a sensitivity analysis. In the final section I will describe the

(6)

6

2. Cost asymmetry

In this section, I will first explain the concept of cost stickiness and the possible explanations for the phenomenon brought forward in previous research. Second, I will develop a theoretical argument as to why cost stickiness is value relevant.

2.1 Cost stickiness

Costs are the consequence of the usage of resources (Banker & Byzalov, 2014). Traditionally, one of the common ways of classifying costs is classifying costs as either fixed or variable. Fixed costs are those costs that are independent of activity levels (within a relevant range). Fixed costs cannot be adjusted in the short term. Variable costs are dependent on activity levels. They are adjustable on the short term and closely follow demand. Some resources result in mixed costs, which partially consist of fixed costs, and partially of variable costs.

Recent research has shown however, that costs can be ‘sticky’ (Anderson, Banker & Janakiraman, 2003). Stickiness of costs refers to the phenomenon that costs adjust asymmetrically to sales volume. In the research by Anderson et al. (2003), selling, general and administration (SG&A) costs increased by 0.55% when sales increased by 1% and only decreased by 0.35% when sales dropped by 1%. Cost increases due to increased sales volume seem to ‘stick around’ when sales volume decreases in subsequent periods. Cost asymmetry is related to cost elasticity; due to cost asymmetry, the elasticity of costs in a period of increasing activity is higher than in a period of decreasing activity (Holzhacker, Krishnan & Mahlendorf, 2014).

2.2 What are causes of cost stickiness?

ABJ name managerial discretion as a cause of cost stickiness. For instance, they find that when sales increase in the period after a sales decrease, cost stickiness reverses. This points to manager’s reaction to uncertainty where they prefer to retain capacity when they are uncertain about the permanence of lower demand (Balakrishnan, Petersen & Soderstrom, 2004). The adjustment of capacity involves adjustment costs; therefore it might not be attractive to continually adjust capacity to demand.

Banker, Byzalov & Threinen (2006) find that cost stickiness differs between countries. They find that a country’s legal environment and economic development influences cost stickiness in multiple ways. They argue that a better legal system leads to more willingness of managers to commit resources in long-term projects. Also, economic development is paired with greater employee intensity, resulting in higher adjustment costs (Banker et al., 2006). Banker, Byzalov and Chen (2012) expand this notion by showing that adjustment costs, the costs incurred to adjust capacity to demand, are positively related to cost stickiness. They study the relationship between the strictness of Employee Protection Laws and cost stickiness. Stricter EPLs lead to higher adjustment costs. They find that for countries where Employment Protection Laws are strict, costs are stickier. These results confirm the findings by Calleja, Steliaros & Thomas (2006), who find that cost stickiness is more pronounced for companies with high employee intensity. For companies with high employee intensity, the adjustment costs might be too high for managers to continually adjust capacity to demand and therefore costs turn out to be sticky.

Balakrishnan, Petersen and Soderstrom (2004) find that capacity utilization also influences cost stickiness. When a company has excess capacity, it will use this capacity to meet extra demand. However, when a company with excess capacity experiences a further decrease in demand, it will

(7)

7

severely cut back capacity. The authors refer to this as cost anti-stickiness; the downward reaction of costs is more severe in this situation. The converse of this is a situation where a company has high capacity utilization. If such a company is faced with a decrease in activity, it might consider

introducing some slack instead of getting rid of excess capacity. Cannon (2014) finds evidence that managers adjust the prices downward when faced with a sales volume decrease, and instead of increasing the price they increase the capacity when faced with an increase of sales volume. This in turn increases costs relative to activity. From the aforementioned research, it can be concluded that an apparent stickiness of costs can be a symptom of multiple issues, not limited to managerial discretion.

Agency theory describes the problems resulting from the misalignment of the interests of

shareholders (principals) and management (agents). One of the agency problems is empire building. Empire building is the increasing or maintaining of excess/unutilized capacity by management, in order to gain status and power (Chen, Lu & Sougiannis, 2012). Cost stickiness can be an indicator of empire building as it can indicate the presence of excess capacity. Banker et al. (2006) find a negative association between shareholder protection laws and cost stickiness. Chen et al. (2012) find a strong positive association between empire building incentives and cost stickiness. They also find that a properly functioning governance system mitigates this association, in line with other Agency theory research. When faced with the proper incentives, managers are more willing to cut costs (Kama & Weiss, 2013). Cost stickiness therefore can be the result of deliberate actions taken by management to maintain capacity even if it is inefficient.

Balakrishnan, Labro & Soderstrom (2014) take a different approach to cost stickiness. Using

simulation, they find that cost structure alone explains a significant portion of cost stickiness thereby partially countering the claims by ABJ regarding the role of management discretion in relation to cost stickiness. After correcting for scale economies, they find that the overall sample costs are not sticky. However, the results differ between industries as 25 of 48 industries do show cost stickiness with 14 of those with significant values.

Due to cost stickiness, a firm’s financial performance may suffer disproportionately when faced with an activity decrease (Holzhacker et al., 2014). The effects of an activity decrease on earnings are twofold: (i) Revenues decrease resulting in a lower absolute amount of earnings and (ii) due to cost stickiness costs are higher relative to sales, therefore lowering earnings even more. Future earnings are in part predicted by historical earnings and its relation to costs (Ohlson, 2001). Therefore, a relatively high degree of cost stickiness will lower the predictability of future performance. This leads investors to require a higher risk premium for their investments. The relation between historical performance and future performance is further investigated in the next section.

2.3 Hypothesis development

Fundamental analysis can be informative to the study of value relevance. Fundamental analysis is aimed at creating forecasting models. Value relevance studies the added value of accounting information and the explanatory power of this information in relation to current share prices.

Fundamental analysis interprets financial ratios and assumes these ratios are signals which shed light on future performance (Anderson, Banker, Huang & Janakiraman, 2007). Where fundamental

analysis investigates the usability of certain information in forecasting earnings, value relevance investigates whether this information is actually used by investors.

(8)

8

The traditional interpretation of an increase in the SG&A ratio is that of an increased inefficiency which has a negative impact on future performance (Lev & Thiagarajan, 1993). However, recent research into cost stickiness, as discussed above, has shown that an increase in the SG&A ratio, an indication of cost stickiness, is not necessarily related to inefficiency (Chen et al., 2012; Baumgarten, Bonenkamp & Homburg, 2010). There are several reasons why management deliberately refrains from cost cutting when faced with a sales decrease such as agency problems and expected sales increases in subsequent periods. The question remains how investors interpret a given SG&A ratio. Anderson et al. (2007) find that the effects of an increase in the SG&A ratio on future performance differ between periods with increasing and decreasing sales. An increase of the SG&A ratio in a period of sales increase (decrease) is related to lower (higher) future earnings. Baumgarten, Bonenkamp & Homburg (2010) expand on this study by categorizing SG&A ratio increase into four groups. They find evidence that in some cases, a high SG&A ratio is an indicator of investments which improve operating efficiency, which result in lower costs and thus earnings in future periods. It can be concluded that a high SG&A ratio makes it harder for investors to use current earnings in a forecast.

Weiss (2010) studies the effects of cost stickiness on analyst coverage and accuracy. He finds that the accuracy of analyst forecasts is lower for firms with sticky costs. Furthermore, he also finds analysts are less prone to make earnings forecasts of firms with sticky costs. This means that assuming analysts’ forecasts are an important source of information for investors, investors have to base their decision on less information and the available information is less accurate. Weiss (2010) indeed finds a more pronounced response to earnings surprises for firms with sticky costs.

As argued above, employee costs can be particularly sticky. Termination of contracts comes at the cost of adjustment costs, including severance pay and costs to hire new employees in subsequent growth periods. As Schiemann & Gunther (2013) find, employee expenses persistence (sticky employee costs) are related to increased earnings predictability.

The direction of the relationship between cost stickiness and firm value is not clear. There are

arguments for both a positive and negative influence on firm value. A positive influence on firm value might come from investments aimed at increasing efficiency (Baumgarten et al., 2010). A negative influence on firm value might come from agency problems (Chen et al., 2012). Managers might also resort to increasing capacity instead of readjusting prices to their old levels (Cannon, 2014). Another important antecedent of cost stickiness is employee protection laws, which make it more difficult and costly to lower production capacity (Banker et al., 2012).

The direction of the relationship between cost stickiness and firm value is not entirely clear although the short term effects on earnings are certainly negative. As argued above, investors base their valuation in part on current earnings. Cost stickiness has a distorting effect on the link between volume and earnings (Banker, Basu, Byzalov & Chen, 2013). Therefore, I hypothesize the following:

Hypothesis 1: Cost stickiness is value relevant information.

Firm value can be measured as the present value of expected future dividends. If current performance is a starting point to determine expected performance, then cost stickiness should

(9)

9

influence the relationship between BVPS (EPS) and share price. The theoretical foundation of the price model, used in this study, is the following equation (Ohlson, 1995):

= + +

Where Pt is price, yt is book value, is abnormal earnings and vt is other information. This equation

implies that the market value of a firm is determined by book value adjusted for current profitability. The addition of vt results in a modification of future profitability (Ohlson, 1995). Ota (2003) states

that future abnormal earnings are dependent on the persistence of current earnings and the presence of other information. In this study, cost stickiness takes the role of other information. As hypothesis 1 states, cost stickiness is expected to have an influence on price. However, the question remains whether cost stickiness acts on price through book value, earnings, or both. Cost stickiness arises when demand decreases. If the decrease in demand is permanent, the book value of a company decreases. First, the fair value of assets decreases and companies are required to perform write downs. These write downs affect book value. Second, accounting rules require the creation of provisions, such as those for severance pay (Banker et al., 2013). These provisions further decrease the book value of a company. Due to information asymmetry, it is unclear to investors whether cost stickiness is persistent or not. Cost stickiness can be an early warning of a decreasing future

performance. Therefore, the presence of cost stickiness should decrease the usefulness (value relevance) of book value per share.

Hypothesis 2a: Cost stickiness has a negative moderating effect on the relationship between book value per share and share price.

Cost stickiness causes a disproportionate drop in costs relative to a drop in sales and therefore affects current earnings (Homburg & Nasev, 2009). In a period with sales decrease earnings are lower than expected when costs are sticky. Ohlson (1995) states that earnings in year t+1 can be predicted using earnings in year t as a starting point. As cost stickiness influences earnings, the prediction of future earnings based on current earnings is less accurate when cost stickiness is high, due to the fact that the relationship between activity levels and earnings is less clear. Banker & Chen (2006) show that an earnings model which takes cost variability and cost stickiness into account outperforms other models which do not contain cost stickiness. Recognizing cost stickiness increases the accuracy of an earnings prediction. If investors take cost stickiness into account, their expectations of future earnings are more accurate. If investors indeed take cost stickiness into account, then the

relationship between current earnings and share price should be lower for firms with high cost stickiness.

Hypothesis 2b: Cost stickiness has a negative moderating effect on the relationship between earnings per share and share price.

3. Methodology

There are two widely accepted measurement of cost stickiness at the firm level: (i) the ABJ method, which refers to a time-series regression model to estimate cost stickiness and (ii) the Weiss (2010) model which uses a direct measure of cost stickiness. The research question addressed in this study requires an estimate of cost stickiness level, instead of its existence. Therefore, in this study the direct measurement of cost stickiness as suggested by Weiss (2010) will be used.

(10)

10

As ABJ focus on SG&A costs, I will use two measures of cost stickiness. The first and main measure determines cost stickiness of SG&A costs, and the second measure determines the cost stickiness of all costs. Both measures are calculated using the formula suggested by Weiss (2010). Banker & Byzalov (2014) support the use of this measure in studies where the aim is to study the effects of cost stickiness.

Cost stickiness of SG&A costs is calculated as follows:

_ , = log _ΔΔ

, − log Δ

Δ ,"

Where τ is the most recent of the last four quarters with a decrease in sales, and T is the most recent of the last four quarters with an increase in sales. Here, ΔSGA is the quarterly change of SG&A costs, and ΔSALES is quarterly increase in net sales excluding extraordinary items.

SG&A cost classification might be influenced by managerial discretion (Anderson & Lanen, 2007). Therefore, I will use a second measure to calculate cost stickiness in a separate model:

# _ , = log _ΔΔ #

, − log Δ # Δ ,"

Where τ is the most recent of the last four quarters with a decrease in sales, and T is the most recent of the last four quarters with an increase in sales. ΔCOST is calculated as follows:

( , − %& & , ) − ( , ( − %& & , ( ). Where EARNINGS is income before extraordinary items and SALES is net sales excluding

extraordinary items in the specified quarter. A negative value of STICKY is an expression of sticky costs. As argued by ABJ and Weiss (2010), sales are an imperfect proxy of activity levels. However, its use is in line with previous research.

3.1 Price model

The basic method used by value relevance researchers is a regression of price on assumed relevant information, and a regression of return on earnings and change in earnings. The theoretical

foundation of these models is the assumption that the market value of a firm is the present value of future dividends. This study uses the price model (Easton, 1999). The price model consists of three variables: Book value, earnings and ‘other information vt’ (Ohlson, 2001). The other information in

the equation is the variable of interest in this study.

t = *0 + *1+, t + *2 t+ *- + .

Where Pt is share price three months after year-end t, BVPSt is book value per share at t, EPSt is

earnings per share at t, vt is other information and ε is random error. The aforementioned extra

information used in this model in this study is cost stickiness. Value relevance is revealed by a significant Beta and to a lesser extent by an increase in adjusted R2 (Ota, 2010).

The return model has certain issues. The market can incorporate events which might have an influence on earnings into the price immediately. In the return model, current return is regressed on current earnings. However, current earnings experience accounting recognition lag due to

(11)

11

conservatism and accounting principles such as reliability of figures. Therefore, the reported (current) earnings experience lag concerning these events, and consist of information that is no longer relevant. Another related issue with the return models is that current earnings consist of both permanent and transitory components. According to Ota (2003), the earnings response coefficients of the two components differ, and therefore the return model will have a lower R2.

The nature of cost stickiness might also influence the results of the return model. Cost stickiness might be temporary, and therefore only a price model can capture its effects. Cost stickiness becomes apparent when accounting figures are reported. The return at that moment has been influenced by cost stickiness, but investors are not yet aware of this. At the reporting date, when cost stickiness becomes apparent, the investors’ response to cost stickiness can only influence the current share price, and this share price is not part of current returns.

Although the use of both models is recommended in some cases, the results might diverge to a point where they produce unclear results (Ota, 2003). Therefore, in this study the price model will be used. In order to test the sensitivity of the results, the sensitivity analysis will be conducted based on the return model.

The regression model is as follows:

= */+ * +, + * + *- ,

Where STICKY is either SGA_STICKY or COST_STICKY.

3.2 Sample selection

In employing this model, I closely follow the data selection procedure used in the Weiss (2010) study. The sample only includes manufacturing firms (SIC divisions 2 and 3). The sample is limited to

manufacturing firms as this ensures that industry effects have no influence on the results. There are differences to be expected between manufacturing firms on the one hand and for instance service firms on the other hand. Where manufacturing firms require long term investments and generally focus on cost strategies, service firms show a higher flexibility. This might influence the level of cost stickiness and therefore might have an influence in the main model. As previous research has shown that cost stickiness differs between countries (Banker et al., 2012), this study will only use data from firms listed on US stock exchanges. The sample consists of observations from 1985 to 2005.

Data used in this study are obtained from the Compustat database. SGA_STICKY and COST_STICKY are calculated for each firm quarter. Observations with missing data are removed from the sample. Observations where the direction of the change in sales differs from that of the direction of change in costs are also omitted in the final sample, in order to comply with the model assumptions (Weiss, 2010). Outliers for all variables are also removed (top and bottom 1%).

(12)

12

Table 1 Descriptive Statistics

Variable Mean Std. Dev Minimum Maximum

Price 17,04013 15,87412 0,187 83,5

BVPS 8,573073 7,567723 -3,2676 45,2429

EPS 0,163831 0,343198 -1,23 1,69

SGA_STICKY -0,01595 0,591705 -1,8522 1,8782

COST_STICKY -0,00532 0,308953 -1,2995 1,342

Table 1 shows the descriptive statistics of the sample containing SGA_STICKY and COST_STICKY. The mean value of SGA_STICKY is negative, meaning that on average the observations show cost

stickiness. The same applies to COST_STICKY. The final dataset contains 24058 firm-quarter observations of 4400 firms.

4. Results

In this section I will present the results of the tests and determine whether the results show support for the hypotheses.

4.1 Price model

The value relevance of information is determined by two statistics: the incremental adjusted R2 (Ota, 2010), and the coefficient of the various variables. Therefore, to determine incremental adjusted R2, I will run several regressions. The first regression of the price model will only consist of book value per share (BVPS). Next, I add earnings per share (EPS). In the next regression, I will add cost stickiness. Models 4 and 5 contain the interaction of the cost stickiness measure with either BVPS or EPS.

4.1.1 Cost stickiness of SG&A cost

To test whether cost stickiness of SG&A costs is value relevant information, I perform a linear regression using panel data. The Hausman test is used to determine whether to use the fixed or random effects model. The results of the Hausman test show that I can reject the null-hypothesis (p < 0.001), and therefore I will use the fixed effects model.

Table 2 Regression on Share Price. ***, **, * Significant at the 1, 5 and 10 percent level. Incr. R2 of model 5 is in relation

to model 3

Variables Model 1 Model 2 Model 3 Model 4 Model 5

BVPS 0,9734*** 0,7920*** 0,7879*** 0,7878*** 0,7880*** EPS 10,9697*** 11,1110*** 11,1226*** 11,1102*** SGA_STICKY -0,6001*** -0,4175** -0,5883*** SGA_STICKYxBVPS -0,0208* SGA_STICKYxEPS -0,0711 Adj. R2 0,15578 0,24294 0,24431 0,24440 0,24430 Incr. R2 0,08716 0,00137 0,00009 -0,00001

Table 2 shows the results of the models using SGA_STICKY. Model 2 shows the main effects of the standard components of the price model. Both are positive and significant. A higher value of BVPS

(13)

13

(EPS) is related to a higher share price. In model 3, the measure of cost stickiness is added. The coefficient of SGA_STICKY is negative and significant in all models, in support of H1. Cost stickiness of SG&A costs does seem to be value relevant information.

Figure 1

H2a states that cost stickiness moderates the relationship between BVPS and share price. H2a is supported; the coefficient of SGA_STICKYxBVPS is negative and significant at the 10% level. Figure 1 shows the simple slope analysis of the interaction between SGA_STICKY and BVPS. The difference between the steepness of the two slopes is minimal. The simple slope is significant. The p-value for the slope when the value of SGA_STICKY falls within the range -1.2 to 1 is less than or equal to 0.05. H2b, which states that that cost stickiness moderates the relationship between EPS and share price, is not supported. The interaction between SGA_STICKY and EPS is not significant.

4.1.2 Cost stickiness of total costs

to model 3

BVPS 0,9734*** 0,7920*** 0,7850*** 0,7838*** 0,7849*** EPS 10,9697*** 11,3007*** 11,3434*** 11,3002*** COST_STICKY -1,3880*** -1,0786*** -1,4080*** COST_STICKYxBVPS -0,0423* COST_STICKYxEPS 0,1911 Adj. R2 0,15578 0,24294 0,24482 0,24491 0,24482 Incr. R2 0,08716 0,00188 0,00009 0,00000

The process is repeated for COST_STICKY; in each model another variable is added. The results of the Hausman test show that the null hypothesis is rejected and therefore I use the fixed effects model. Table 3 shows the results of the 5 models. Again I find support for H1. In the main effects model (3), the coefficient of COST_STICKY is negative and significant. Therefore, COST_STICKY is also value relevant information. 0 2 4 6 8 10 12 14 16 18 Low BVPS High BVPS P ri ce Low SGA_STICKY High SGA_STICKY

(14)

14 Figure 2

The results also show support for H2a, there is an interaction between BVPS and COST_STICKY. Figure 2 shows the simple slope analysis of the interaction between COST_STICKY and BVPS. The slope is significant at the 10% level if the moderator COST_STICKY has a value between -0.5 and 0.5. The slope of high COST_STICKY is less steep, although the difference is minimal. The interaction of COST_STICKY with EPS is not significant, which means H2b is not supported.

5. Sensitivity analysis

In this section, I will perform two additional analyses. First, I will use a different measure for cost stickiness. Second, I will show the results for the return model. As stated in the methodology section, I assume the results for the return model have too many issues to be reliable. However, some authors suggest using both methods, even though this leads to unclear conclusions.

5.1 SG&A Signal

Instead of using the method to calculate cost stickiness suggested by Weiss (2010), I will use the SG&A signal suggested by Baumgarten et al. (2010). This signal is calculated as follows:

& _ = & , , −

& , ( , (

Where SG&A are selling, general and administrative costs, and SALES are total revenue. This method uses yearly data, instead of the quarterly data used in the Weiss (2010) model. The present study deviates from Baumgarten et al. (2010) in that it does not discern between the cost efficiency and cost stickiness explanations of this signal. Therefore, it is an imperfect way to measure cost stickiness. However, one could argue that SG&A_SIG bears the appearance of cost stickiness. Baumgarten et al. (2010) propose differentiating between cost efficiency investments and cost stickiness to better explain the changes in the SG&A ratio. If investors do not make this distinction then any increase in SG&A_SIG might seem to be caused by cost stickiness or lack of cost control in the eyes of investors. Nonetheless, the theoretical basis for using SG&A_SIG as a proxy for cost stickiness is small.

The sample contains firms from all industries, and spans from 1979 to 2014. The first observation is form the year 1980, as the method to calculate SG&A_SIG for year t requires data from year t-1.

0 2 4 6 8 10 12 14 16 18 Low BVPS High BVPS D ep en d en t var iab le Low COST_STICKY High COST_STICKY

(15)

15

Observations with missing data are removed from the final sample. Outliers (top and bottom 1%) are also removed. This method differs from the method proposed by Weiss (2010) in that a positive value of SG&A_SIG points to cost stickiness. I exclude all negative values of SG&A_SIG to ensure that the measure is a better representation of cost stickiness. The final sample consists of 67159

observations.

Variables Mean Std. Dev Minimum Maximum

P 16,5133 19,7218 0,0000 717,5000

BVPS 9,4328 10,0066 0,0002 99,6739

EPS 0,5691 2,2161 -81,9600 83,1800

SG&A_SIG 0,0433 0,0707 0,0001 1,1926

Table 5 shows the results of the 5 models using SG&A_SIG. In line with the previous results, and in support of H1, the main effect of SG&A_SIG is negative and significant. Cost stickiness of SG&A costs is value relevant information. In model 4, the interaction between SG&A_SIG and BVPS is added. The results show a negative and significant relationship between the interaction and share price. These results are in support of H2a. In model 5, the interaction between SG&A_SIG and EPS is added. Again, the coefficient of the interaction is negative and significant, in support of H2b.

Table 5 Regression on Share Price. ***, **, * Significant at the 1, 5 and 10 percent level level. Incr. R2 of model 5 is in

relation to model 3

BVPS 1,0818**** 0,9005*** 0,9009*** 0,9107*** 0,8846*** EPS 1,4580*** 1,4430*** 1,4352*** 1,7006*** SG&A_SIG -3.9617*** -2,3837* -7,7701*** SG&A_SIGxBVPS -0,4233** SG&A_SIGxEPS -6,1178*** Adj. R2 0,16006 0,19151 0,19172 0,19181 0,19414 Incr. R2 0,0315 0,00021 0,00009 0,00242

(16)

16

Figures 3 and 4 show the simple slope analysis of both moderators. As the regression results show, the interaction between SG&A_SIG and BVPS is significant. The slope representing high SG&A_SIG is less steep than that of low SG&A_SIG, in line with H2a. The effect of higher book value per share on share price is less profound when cost stickiness increases.

Figure 3 Simple slope, interaction between SG&A_SIG and BVPS

Figure 4 lends support to H2b. The slope of high SG&A_SIG is less steep than that of low SG&A_SIG.

Figure 4 Simple slope, interaction between SG&A_SIG and EPS

5.2 Return models

In this subsection, I will present the results for the return model. As stated above, this model can suffer from issues. Furthermore, in the sample used in this study there are issues regarding

multicollinearity. The VIF value is 3.928 for both e and Δe. While there is no consensus on an upper limit for the VIF value, a value of 3.928 is too high for such an important variable in my opinion. The price model does not suffer from this problem. The return model is as follows:

%12 = */+ * 1 + * 31 + *- ,

Where Rett is total return over the last 12 months, et is quarterly earnings per share deflated by Pt-1.

Δet is change in quarterly earnings per share, deflated by Pt-1. STICKY:,; is either SGA_STICKY or

COST_STICKY. 0 5 10 15 20 25 30 Low BVPS High BVPS P ri ce Low SG&A_SIG High SG&A_SIG 0 2 4 6 8 10 12 14 16

Low EPS High EPS

P ri ce Low SG&A_SIG High SG&A_SIG

(17)

17

Variable Mean Std. Dev Minimum Maximum

RETURN -0,30111 13,96458 -42,5 58,2666

dEPS -0,00723 0,593958 -90 13,5747

dDEPS 0,003849 0,450882 -60 13,92

SGA_STICKY -0,01595 0,591705 -1,8522 1,8782

COST_STICKY -0,00532 0,308953 -1,2995 1,342

Table 6 above shows the descriptive statistics for the return model. The sample is the same as that of the return model, except for the substitution of PRICE for RETURN. Again, both cost stickiness

measures are negative, meaning that in the sample there is evidence of cost stickiness.

to model 3

e -0,1689 0,0182 0,0668 0,4305 0,9104* Δe -0,2820 -0,3617 -0,2731 -1,1279* SGA_STICKY 0,6591*** 0,6614*** 0,6510*** SGA_STICKYxe 1,4965 SGA_STICKYxΔe 1,7327* Adj. R2 0,0000516 0,0000664 0,0007215 0,0007817 0,0009811 Incr. R2 0,0000148 0,0006551 0,0000602 0,0002596

Table 7 above presents the results for the return model using SGA_STICKY. Here, the issues of applying the return model to this sample become clear. The main effects of both e (deflated earnings per share) and Δe are not significant in model 2. In model 3, the coefficient of SGA_STICKY is

significant, but the direction is not in compliance with the expectations based on the price model.

to model 3

e -0,1689 0,0182 0,1501 0,6842* 0,3097 Δe -0,2820 -0,4994 -0,4686 -0,6247 COST_STICKY 1,9173*** 1,9628*** 1,8751*** COST_STICKYxe 5,3149* COST_STICKYxΔe 1,1058 Adj. R2 0,00005 0,00007 0,00152 0,00178 0,00158 Incr. R2 0,00001 0,00146 0,00026 0,00006

Table 8 presents the results of the return models containing COST_STICKY. Again, the results are questionable. The coefficients of e and Δe are not significant in any of the models. These findings are inconsistent with previous research using the return model (Ota, 2010). Even though Ota (2010) shows that the adjusted R2 of the return model is lower than that of the price model, they are higher

(18)

18

than the R2 reported in this study. Furthermore, he does report significant and positive coefficients for deflated earnings and deflated change in earnings.

In table 8 COST_STICKY is significant, but as with SGA_STICKY in an unexpected (positive) direction. In the next subsection I will try to address possible issues with the return model, by suggesting a modification.

5.3 Modified return model

In the return model, return is regressed on deflated earnings per share and changes in deflated earnings per share. In this section, I suggest a modified return model, which regresses return on book value per share and earnings per share.

%12 = */+ * +, + * + *- ,

Where Rett is total return over the last 12 months, BVPSt is book value per share and EPSt is earnings

per share. STICKY:,; is either SGA_STICKY or COST_STICKY.

to model 3

BVPS 0,0108 -0,0514* -0,0481* -0,0485* -0,0483* EPS 3,7683*** 3,6520*** 3,6874*** 3,6530*** SGA_STICKY 0,4937** 1,0513*** 0,4788** SGA_STICKYxBVPS -0,0635** SGA_STICKYxEPS 0,0898 Adj. R2 0,0000075 0,0040048 0,0043690 0,0047335 0,0043702 Incr. R2 0,0039973 0,0003642 0,0003645 0,0000012

The results in table 9 above and table 10 below are consistent with the findings of the original return model. The coefficient of both SGA_STICKY and COST_STICKY are positive and significant. However, the coefficient of EPS is also positive and significant. This is in contrast with the results of the original return model, where the coefficient for deflated EPS was not significant in the main effects models.

Table 10 Regression on Share Price. * ***, **, * Significant at the 0.1, 1 and 10 percent level. Incr. R2 of model 5 is in

relation to model 3

BVPS 0,0108 -0,0514* -,0449* -0,0459* -0,0452* EPS 3,7683*** 3,4570*** 3,4902*** 3,4548*** COST_STICKY 1,3052*** 1,5460*** 1,2243*** COST_STICKYxBVPS -0,0329 COST_STICKYxEPS 0,7725 Adj. R2 0,0000075 0,0040048 0,0046558 0,0046798 0,0046893 Incr. R2 0,0039973 0,0006510 0,0000240 0,0000335

(19)

19

6. Discussion and Conclusions

The aim of this paper was to study the value relevance of cost stickiness. This paper has studied value relevance of cost stickiness using a price model which regresses share price on book value per share, earnings per share and two measures of cost stickiness.. In doing so, it has tried to uncover the effects of cost stickiness on decisions made by investors. The results are informative for preparers of annual statements, as they are in the position to reduce the information asymmetry which results from cost stickiness.

The results show that the relationship between cost stickiness and share price is negative and significant. There are multiple implications of these findings. First, these results implicate that when looking at the entire sample, more severe cost stickiness is a signal to shareholders that something is wrong. As discussed in the theory section, there are multiple reasons why cost stickiness has an influence on earnings, and therefore firm value. While some of these reasons should lead to an improved firm performance, the negative causes are stronger. This paper is the first to actually link these causes of cost stickiness to one possible consequence of cost stickiness.

Cost stickiness is a sign of agency problems, such as empire building (Chen, 2010). Research on agency problems has shown that investors are indeed influenced by signs of empire building (Iskandar-Datta & Jia, 2014). This paper adds to this literature by showing a possible new proxy of empire related problems, although it should be used in conjunction with other variables.

The results of this paper show that cost stickiness is indeed value relevant information. The share prices of companies with sticky costs have on average lower share prices. The results also show that for firms with high cost stickiness, the relationship between book value per share and share price is less profound. For firms with high cost stickiness, a high book value per share is related to a lower share price than for firms with low cost stickiness. This moderating effect is also present in the relationship between earnings per share and share price. However, this moderating effect only receives weak support.

This paper suffers from a few limitations. The use of sales as a proxy of activity has its drawbacks, as ABJ argue. As in its most basic form sales is the result of activity multiplied by price, changes in prices can also cause a change in sales even when actual activity remains the same. Cost stickiness is dependent on changes in activity, not prices.

The sample used in this study, with exception to the sample used in part of the sensitivity analysis, only consists of manufacturing firms. ABJ find that cost stickiness is present in all industries. There is a possibility that cost stickiness is more profound in manufacturing firms, due to the cost structure inherent to that industry. Future research might replicate this study in a different industry.

Another possible limitation is the method used to calculate cost stickiness. The method used in this paper differs from the one originally used by ABJ. The method suggested by ABJ is not applicable in this study, therefore another measure was used. However, Banker & Byzalov (2014) state that when the effects of cost stickiness are studied, cost stickiness can be measured using the method

(20)

20

The results of this paper lean heavily on the use of the price model, even though some authors suggest using both the price and return model in conjunction (Ota, 2006). The price model can be influenced by scale effects (Brown, Lo, & Lys, 1999).

As Baumgarten et al. (2012) show, an SG&A signal can be split up into several categories. Not all of these categories are signals of increased inefficiency. The signal can also be the result of investments aimed to increase efficiency. The results of the sensitivity analysis using the SG&A signal do not discern between these categories, therefore both ‘positive and negative’ cost stickiness is treated the same.

The findings might not be generalizable to other countries, due to the fact that the sample consists of only US based firms. Future research might expand on this study. Harris, Lang & Möller (1994) show that the R2 of the price model of comparable firms in Germany is lower than that of firms in the US. It might be interesting to determine the effects of cost stickiness in this setting.

Overall, the results of this study show that cost stickiness influences decisions made by investors, and therefore should be considered by management.

(21)

21

7. Literature

Anderson, M., Banker, R., Huang, R., & Janakiraman, S. (2007). Cost behavior and fundamental analysis of SG&A costs. Journal of Accounting, Auditing & Finance, 22(1), 1-28.

Anderson, M. C., Banker, R. D., & Janakiraman, S. N. (2003). Are selling, general, and administrative costs “sticky”? Journal of Accounting Research,41(1), 47-63.

Anderson, S. W., & Lanen, W. N. (2007). Understanding Cost Management: What Can We Learn from the Evidence on 'Sticky Costs'? Available at SSRN 975135.

Banker, R. D., Byzalov, D., Ciftci, M., & Mashruwala, R. (2014). The moderating effect of prior sales changes on asymmetric cost behavior. Journal of Management Accounting Research, 26(2), 221-242.

Banker, R. D., Byzalov, D., & Chen, L. T. (2013). Employment protection legislation, adjustment costs and cross-country differences in cost behavior. Journal of Accounting and Economics, 55(1), 111-127.

Banker, R. D., Byzalov, D., & Threinen, L. (2013). Determinants of international differences in asymmetric cost behavior. Fox School of Business, Temple University.

Banker, R. D., & Chen, L. (2006). Predicting earnings using a model based on cost variability and cost stickiness. The Accounting Review, 81(2), 285-307.

Balakrishnan, R., Labro, E., & Soderstrom, N. S. (2014). Cost structure and sticky costs. Journal of

Management Accounting Research, 26(2), 91-116.

Balakrishnan, R., Petersen, M. J., & Soderstrom, N. S. (2004). Does capacity utilization affect the “stickiness” of cost? Journal of Accounting, Auditing & Finance, 19(3), 283-300.

Baumgarten, D., Bonenkamp, U., & Homburg, C. (2010). The information content of the SG&A ratio. Journal of Management Accounting Research,22(1), 1-22.

Brown, S., Lo, K., & Lys, T. (1999). Use of R 2 in accounting research: measuring changes in value relevance over the last four decades. Journal of Accounting and Economics, 28(2), 83-115. Calleja, K., Steliaros, M., & Thomas, D. C. (2006). A note on cost stickiness: Some international comparisons. Management Accounting Research, 17(2), 127-140.

Cannon, J. N. (2014). Determinants of ''Sticky Costs'': An Analysis of Cost Behavior using United States Air Transportation Industry Data. Accounting Review, 89(5), 1645-1672.

Chen, C. X., Lu, H., & Sougiannis, T. (2012). The Agency Problem, Corporate Governance, and the Asymmetrical Behavior of Selling, General, and Administrative Costs. Contemporary Accounting

Research, 29(1), 252-282.

Cooper, R., & Kaplan, R. S. (1992). Activity-based systems: Measuring the costs of resource usage. Accounting Horizons, 6(3), 1-13.

(22)

22

Easton, P. D. (1999). Security Returns and the Value Relevance of Accounting Data. Accounting

Horizons, 13(4), 399-412.

Feltham, G. A., & Ohlson, J. A. (1995). Valuation and clean surplus accounting for operating and financial activities. Contemporary accounting research, 11(2), 689-731.

Harris, T. S., Lang, M., & Peter Möller, H. (1994). The Value Relevance of German Accounting Measures: An Empirical Analysis. Journal Of Accounting Research, 32(2), 187-209.

Holzhacker, M., Krishnan, R., & Mahlendorf, M. D. (2014). The impact of changes in regulation on cost behavior. Contemporary Accounting Research. In press.

Homburg, C., & Nasev, J. (2009). How timely are earnings when costs are sticky? Implications for the link between conditional conservatism and cost stickiness. AAA.

Iskandar-Datta, M., & Jia, Y. (2014). Investor protection and corporate cash holdings around the world: new evidence. Review Of Quantitative Finance & Accounting, 43(2), 245-273.

Kama, I., & Weiss, D. (2013). Do earnings targets and managerial incentives affect sticky costs? Journal of Accounting Research, 51(1), 201-224.

Marquardt, C. A., & Wiedman, C. I. (2004). How Are Earnings Managed? An Examination of Specific Accruals. Contemporary Accounting Research, 21(2), 461-491.

Noreen, E., & Soderstrom, N. (1997). The accuracy of proportional cost models: Evidence from hospital service departments. Review of Accounting Studies, 2(1), 89-114.

Ohlson, J. A. (2001). Earnings, Book Values, and Dividends in Equity Valuation: An Empirical Perspective. Contemporary Accounting Research, 18(1), 107-120.

Ohlson, J. A. (1995). Earnings, book values, and dividends in equity valuation. Contemporary

accounting research, 11(2), 661-687.

Ota, K. (2001). The impact of valuation models on value-relevance studies in accounting: a review of theory and evidence. Available at SSRN 280873.

OTA, K. (2010). The Value Relevance of Management Forecasts and Their Impact on Analysts' Forecasts: Empirical Evidence From Japan. Abacus, 46(1), 28-59.

Schiemann, F., & Guenther, T. (2013). Earnings Predictability, Value Relevance, and Employee Expenses. The International Journal of Accounting, 48(2), 149-172.

Weiss, D. (2010). Cost Behavior and Analysts' Earnings Forecasts. Accounting Review, 85(4), 1441-1471.

The Value Relevance of Cost Stickiness

The Value

Relevance of

Cost Stickiness

[

WIETZE-JAN HAISMA

]

Abstract

Table of Contents

1. Introduction

2.

Cost asymmetry

3.

Methodology

4.

Results

5.

Sensitivity analysis

6.

Discussion and Conclusions

7.

Literature