Cost stickiness and the effect of managers’ growth expectations

(1)

Amsterdam Business School

Cost stickiness and the effect of managers’

growth expectations

Name: Ricardo Borhem

Student number: 10895418

Date: 20 June 2016

MSc Accountancy & Control, specialization Accountancy and Control Faculty of Economics and Business, University of Amsterdam

Abstract: Research has shown that cost increase more when activity rises than they decrease when activity falls by an equivalent amount. This phenomenon occurs because managers make asymmetric decision that lead to asymmetric cost behaviour. Some previous studies have already identified which factors lead to these asymmetric decisions. This study investigates the effect of managers’ growth expectations on cost stickiness. I hypothesis that firms with higher growth expectations exhibit more cost stickiness, because managers might be less willing to decrease committed resources if they expect future growth. The findings indicate that the stickiness of cost is higher in firms with higher growth expectations.

(2)

Statement of Originality

This document is written by student Ricardo Borhem who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1 Introduction

1.1 Background

Traditionally, the commonly received model of cost behaviour made the distinction between fixed and variable cost and assumes that variable cost changed symmetrically with changes in their activity divers (Noreen, 1991). It implies that there is a linear relation between operational activities and variable cost (Balakrishnan et al., 2008). However, some cost rise more with an increase in operational activity then they fall with a decrease in operational activity (Noreen and Soderstrom, 1997; Cooper and Kaplan, 1998).

Anderson et al. (2003) argues that selling, general and administrative cost (SG&A costs), which are mainly driven by sales volume, respond differently to an increase or decrease of operational activities. Their statistical indicators estimated that, in a one-year period, SG&A costs increased 0,55% per 1% increase of sales revenues, which is a proxy for operational activities, while there would be a decrease of only 0,35% per 1% decrease of sales revenues. These findings are in contrast with the traditional model of cost behaviour, where variable cost change symmetrically with change in operational activities, and are consistent with the alternative model of cost behaviour. This phenomenon is referred to as ‘cost stickiness’ and is, according to Banker & Byzalov (2014), a global phenomenon.

Managers may choose to decrease their committed resources when there is a permanent decline in demand. However, when there is uncertainty about whether or not the demand will recover, managers might also choose not to decrease committed resources because reducing and restoring committed resources leads to adjustment costs. So managers may purposely delay the reduction of committed resources until there is more certainty about future demand (Anderson et al. 2003). This asymmetry in decision-making courses cost stickiness.

Managers that expect that the firm will grow will probably be less eager to decrease committed resources when demand decreases. Since they expect that demand will eventually increase, so they avoid adjustment cost. While managers who don’t expect that the firm will grow might be more willing to decrease committed resources. Because the cost of operating with unutilized capacity might be higher than the adjustment cost. If this is the case, then cost will be stickier at firms with high growth expectations than at firms with low or no growth expectations.

(5)

1.2 Research question

According to the study of Anderson et al. (2003), cost increase more when activity rises than they decrease when activity falls by an equivalent amount. This occurs because managers deliberately adjust the resources committed to activities. After this study, there have been conducted multiple studies about cost stickiness and why manages make asymmetric managers decisions that lead to this phenomenon. One of the variables that might affect the decisions of managers is the expectations about future growth. One part of future growth is the expectations about future demand. This expectation might affect the choice to either operate with unutilized capacity or to decrease committed resources when demand decreases. Based on this rational, this paper will address the following research question:

“To what extent does the managers’ growth expectation influence the asymmetrical cost behaviour?”

This paper hypothesises that costs exhibit greater stickiness in firms with higher growth expectations. Usually, managers decrease committed resources when there is a decrease in demand, because it leads to lower sales revenue. When managers expect that the firm will grow, they might be less willing to decrease committed resources when demand decreases. Since they expect that demand will eventually increase. This should lead to more asymmetric cost behaviour in firms for which there is a higher growth expectation.

Based on a sample of 55.996 observations of 9.711 firms over a period between 1998 and 2015, there has been found some evidence to support this hypothesis. Estimating results indicate that there is a significant difference in cost stickiness between firms with high growth expectations and firms with low growth expectations. Based on the statistical indicators it is estimated that, in a one-year period, SG&A costs increased 0,65% per 1% increase of sales revenues of firms with a high growth expectations. And that SG&A cost decrease 0,37% per 1% decrease of sales revenues. While in firms with a low growth expectations, SG&A cost increase 0,66% per 1% increase of sales with a decrease of 0,42% per 1% decrease of sales revenue. This difference of 0,04 percentage point per 1% decrease in sales revenue is statistically significant.

(6)

1.3 Motivation and contribution

This paper will contribute to the literature regarding asymmetric cost behaviour. A better understanding of cost behaviour is very relevant to accounting researchers, managers, auditors, financial analysts and professional investors, because these groups rely on cost accounting data for their core activities (Baumgarten, 2012). Weiss (2010) even found evidence that asymmetric cost behaviour influences analysts’ earnings forecast and that firms with stickier cost behaviour have less accurate analysts’ earnings forecasts than firms with less sticky cost behaviour. So contributing to the subject of cost stickiness, which is an asymmetric cost behaviour, is very relevant for people who rely on cost accounting data.

One of the reasons that cost stickiness exists is because of asymmetric management decisions. The decisions that managers make are based upon a number of variables and one of them is their expectation about future demand. A manager who has a high growth expectation will probably react differently to a decline in demand than a manager who doesn’t have growth expectations. By researching the variables which affect management decisions, one is able to gather some insight in how decisions are affected and what the consequences are on the behaviour of cost. If this paper will provide indications that the future expectations of a managers affect the stickiness of cost, then it will provide new insight on how managers react in different situations.

Managers should learn how they can avoid cost stickiness and how they can permanently adjust cost structures to a permanently changing environment (Guenther, 2013). To avoid cost stickiness, it is necessary to be aware of the variables that lead to sticky cost.

1.4 Structure

This paper will be structured as follows. In section two, there will be a review of the prior literature and hypothesis development. Section three discusses the research methodology including the sample selection, empirical models and the used variables. The empiric results will be provided in section four. This section will consist of descriptive statistics and a multivariable analysis. Finally, in section five, the conclusion will be presented including the limitations of this study.

(7)

2 Literature review and hypothesis

2.1 Cost Stickiness Phenomenon

In the traditional accounting literature, there is a distinction between fixed and variable cost. A fixed cost is seen as a cost that does not vary in the short run and that depends on the amount of a resource that is acquired rather than the amount that is used. A variable cost, on the other hand, is seen as a cost that increases proportionally with the change in its cost driver (Atkinson et al., 2012). This model implies variable cost behave symmetrically and that the magnitude of change in cost depends only on the extent of change in the level of activity, not on the direction of the change (Anderson et al., 2003).

The study of Anderson et al. (2003) provided the first empirical evidence of asymmetric cost behaviour. They investigated the behaviour of selling, general and administrative cost (SG&A cost) in relation to the sales revenue, which is a proxy for operational activities. Their sample showed that, in a one-year period, SG&A costs increased 0,55% per 1% increase of sales revenues while there would be a decrease of only 0,35% per 1% decrease of sales revenues. This phenomenon is referred to as ‘cost stickiness’ and the findings of this study are in contrast with the traditional model of cost behaviour, where variable cost change symmetrically with change in operational activities, and are consistent with the alternative model of cost behaviour.

2.1.1 Adjustment cost

The alternative model makes a distinction between cost that move mechanistically with changes in volume and cost that are determined by resources committed by managers. Cost that move mechanically, such as direct material cost, are expected to move mechanically with changes in volume because they can be adjusted without incurring any adjustment cost (Banker et al. 2011). Cost that are determined by resources committed by managers, on the other hand, are considered to be lumpy. This means that some resources cannot quickly be added or subtracted when there is an increase or decrease in demand. Besides the lumpiness, there might also be a time lag. First, managers need information about sales before they can react on it. And after they receive that information, it also takes some time before a decision of cost reduction is realized.

But cost lumpiness alone should not lead to cost stickiness. Cost stickiness occurs when there are asymmetric frictions in making resource adjustments. These asymmetric frictions might be caused by the fact that there are forces acting to restrain or slowdown the downward adjustment process more than the upward adjustment process (Anderson et al., 2003). These slow downward adjustment processes might be caused by the fact that firms must incur

(8)

adjustment costs to remove committed resources and to replace those resources if demand is restored. Examples of adjustment costs are service pay when employees are dismissed and cost for searching, hiring and training new employees (Balakrishnan et al., 2008). In addition to the adjustment cost, which are directly related to removing and recovering committed resources, there are also organisational cost. Examples are the loss of morale among remaining employees when colleges are terminated, and erosion of human capital when new employees are hired (Anderson et al., 2003). These adjustment and organisational cost make managers purposely delay cost reductions until they are more certain about the future performance of the firm.

When demand increases, managers increase committed resources to the extent necessary to accommodate the additional sales. When demand decreases, however, some committed resources will not decrease unless managers make the deliberate decision to remove them. Since demand is not stable, managers must evaluate whether the decrease in demand is temporary or permanent. If managers expect that the decrease is temporary, they might rather retain unutilized resources than incur adjustment cost (Anderson et al., 2003).

2.1.2 Magnitude of change

The magnitude of change in economic activity has also been viewed as a possible cause for cost stickiness (Venieris et al., 2015). Balakrishnan et al. (2004) researched the effect of the magnitude of the change in operational activity. They found that managers do not significantly change staff hours in response to small (3% or less) changes in activity levels. But there is a significant change in staff hours in response to large (more than 3%) changes in activity levels. However, the difference in response to small versus large changes is not statistically significant. Thus, their evidence does not support the hypothesis that the response to large changes differs from the response to small changes. A different study of Subramaniam & Weidenmier (2003) found evidence that costs are sticky when revenues chance by more than 10%. They suggest that if revenues decrease by more than 10%, managers may not want, or be able, to reduce the capacity of the firm, causing cost stickiness.

2.1.3 Managerial incentives

The decision to maintain unutilized resources may also be caused by personal considerations that relate to agency problems which lead to agency cost. The agency theory implies that mangers are self-interested and make decisions to maximize their own utility but are not optimal from the perspective of the firm’s stockholders (Jensen & Meckling, 1976). In a study about asymmetric labour cost behaviour, Dierynck et al. (2012) found that cost stickiness occurred in labour cost. These results are consistent with the findings of Anderson et al. (2003). But in addition, they also

(9)

found that this was only the case if there were no incentives to manage earnings. If there were incentives to manage earnings, it would lead to symmetrical labour cost behaviour. The supposed reason is that managers, who have incentives to manage earnings, limit the increase in labour costs when there is an increase in operational activities and are more willing to cut labour cost if there is a decrease in operational activities.

2.2 Future expectations

As previously explained, cost stickiness is partly the result of managers’ decisions about resource adjustments. One of the factors that presumably affect the decisions of managers is their expectation about future demand (Venieris et al., 2015). If there is an expectation that the demand will decrease, it means that the managers expect that the firm will shrink form an economic perspective. This will lead to lower production, lower sales revenue and lower profits. When managers expect that demand will increase, it means that he mangers expect that the firm will grow from an economic perspective. This will lead to higher production, higher sales revenue and higher profit.

According to Chen et al. (2013), overconfident managers will overestimate the likelihood of a future sales rebound, which motivates them to keep excess SG&A resources, leading to greater stickiness in SG&A cost. This research confirms the hypothesis that managers’ future expectations have an effect on the stickiness of cost.

2.2.1 Indicators of a firm future growth

When a firm has the ability to become bigger and more profitable, it is called growth potential. It is a conventional wisdom that the market-to-book ratio is a proxy for growth opportunity (Chen & Zhao, 2006). The market-to-book ratio, also known as price-to-book (P/B) ratio, is an indicator for expected return on equity. This indicator is widely used by investment analysts to evaluate the stock price of a firm. Another common measure for growth is the price-earnings (P/E) ratio (Penman, 1996). The P/E ratio is an earnings growth indicator (Malkiel & Cragg, 1982).

2.2.2 Managers’ future growth expectation

Both ratios are based on stock price, and stock price is affected by the expectation of investors and not the expectation of managers. According to the agency theory there is information asymmetry between the agent (manager) and the principal (investor). The agent has an information advantage over the principal. Based on the agency theory, it would not be logical

(10)

that the expectations of investors are equal to the expectations of managers. However, if there would be efficient markets than this would not be a problem. An efficient market is, according to the efficient market hypothesis, a market in which prices always fully reflect available information (Fama, 1970). There are three forms markets according to the effective market hypothesis. Markets with:

Weak efficiency: Prices fully reflect any information contained in past price data. Semi-strong efficiency: Prices fully reflect all available public information.

Strong efficiency: Prices fully reflect all public and private information.

Under the assumption that there is a strong efficiency in the market, market-to-book ratio and price-earnings ratio could be used as proxies for the managers’ expectation of growth. Because, in efficient markets, managers and investors have equal information so they might come to the same conclusion based upon this information. This should be reflected in these ratios.

2.3 Impact of managers’ growth expectations on cost stickiness

The costs stickiness model suggest that costs are asymmetrical, because managers deliberately adjust committed resources to changes in activity levels. This means that cost stickiness is dependent on the willingness of managers to adjust resources. One of the variables that managers presumably based their decision on, is their expectation about future demand. This expectation is, probably, partly based upon the growth potential of the firm.

Firms with managers that expect a future growth in sales revenue are less willing to decrease cost than firms with managers who expect a future decrease in sales revenue. This should lead to more cost stickiness in firms with managers that expect future growth. The reason to expect that managers with growth expectations are less willing to decrease cost, is because they would probably expect that the firm will grow in the long run. And if they will downsize the firm by decreasing costs in the short run, this might hold back future growth. This rational has led to the following assumption:

Hypothesis:

(11)

3 Research methodology

3.1 Sample Selection

The sample is obtained from a the Compustat database, which contains historical accounting and market data from thousands of inactive and active firms. The sample in this study will only contain of firms from the S&P 1500 with firm-year observations from the years 1998 to 2015. Further details about the downloaded variable can be found in appendix 1. The initial number of observations were 174.395 of 21.148 firms. From this initial sample, financial institutions (SIC Codes 6000-6799) and public utilities (SIC Code 4900-4999) were removed, because the structure of their financial statements is incompatible with firms from other industries (Kama & Weiss, 2011). Firms without a number to identify the industry in which they operate were also removed. By removing these firms, 58.121 observations were lost. Observations that had missing values for either sales revenue or SG&A cost were also removed. This led to a loss of 23.388 observations. Subsequently, there was also a loss of 10.463 observations because SG&A cost exceeded sales revenue. Firms without assets or balance sheet information were also removed, which led to a loss of 251 observations. Firms with missing data to calculate the independent variables (market-to-book ratio and price-earnings ratio) were also removed, which resulted in a loss of 14.429 observations. Just as for the independent variables, observations without data to calculate the control variables were also removed which led to a loss of 436 observations. Furthermore, observations of firms with either one observation or non-consecutive years were also removed, for this reason, 1.595 observations were lost. Finally, observations for which no logarithmic change can be calculated were also removed what led to a loss of 9.716 observations. This leads to a final sample of 55.996 observations of 9.711 firms.

Table 1

Sample Selection

Number of observations

Initial sample 174.395

Less: Financial institutions and public utility firms 58.121

Less: Missing sales revenue and SG&A cost 23.388

Less: SG&A cost that exceeded sales revenue 10.463

Less: Firms without assets or balance sheet information 251

Less: Missing data to calculate P/B or P/E ratio 14.429

Less: Missing data to calculate control variables 436

Less: Firms with a single observation 1.595

Less: Observations without data from previous year 9.716

(12)

3.2 Empirical model

The original cost stickiness model, as developed by Anderson et al. (2003), will be used to test for cost stickiness. This model enables measurement of SG&A cost response to changes in sales revenue (Anderson et al., 2003).

Model 1: log ( 𝑆𝐺&𝐴𝑖,𝑡 𝑆𝐺&𝐴𝑖,𝑡−1 ) = 𝛽0+ 𝛽1log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1 ) + 𝛽2 𝐷𝑒𝑐𝑟𝑒𝑎𝑠𝑒_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1 ) + 𝜀𝑖,𝑡.

In this model, log(𝑆𝐺&𝐴𝑖,𝑡 / 𝑆𝐺&𝐴𝑖,𝑡−1) represents the logarithm of the percentage of change

in cost for firm i in year t, relative to year t-1. On average, the SG&A cost to total assets ratio is 27% (Banker et al., 2011). Due to the importance of SG&A cost, practitioners pay close attention to controlling SG&A cost spending. Thus understanding the behaviour of SG&A cost and the role of managers in adjusting the cost is important to research (Chen et al., 2012). Cooper & Kaplan (1998) also states that the behaviour of SG&A cost can be meaningfully studied in relation to revenue activity because sales volume drives many components of SG&A cost. SG&A cost data is also widely available for abroad set of firms in the Compustat database (Anderson et al., 2003). All these characteristics make SG&A cost the right instrument to test for cost stickiness.

The logarithmic percentage change in activity level for firm i in year t, relative to year t-1 is represented by log(𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 / 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1). According to prior literature, annual sales revenue

is a common proxy for activity levels (Anderson et al., 2003).

The interaction variable, 𝐷𝑒𝑐𝑟𝑒𝑎𝑠𝑒_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏, is a dummy variable that has value 1 when

sales revenue decrease between period t-1 and t, and value 0 when sales revenue increases in this same period. This dummy enables this model to calculate the effect of decrease in sales revenue on the change in SG&A cost.

The hypothesis of this study is that costs exhibit greater stickiness in firms with higher growth expectations. This hypothesis will be tested with a subsample analyses, that is, by making median splits on the independent variables of interest. In this case, the proxies for growth expectation are market-to-book ratio and price-earnings ratio. The market-to-book ratio is an indicator for expected return on equity and can be calculated by dividing the market value of the firm (market price of stock times total number of shares issued) through the book value of the firm (assets minus liabilities). The price-earnings ratio is an earnings growth indicator and can be calculated by dividing the market value per share through the earnings per share. The ratios will

(13)

determine which observation belongs to which subsample. Because the full sample will be divided in subsamples of firms with high and low ratios. ‘High’ means above median (50% of the firms with the highest ratios), ‘low’ means under median (50% of the firms with the lowest ratios). This will result in four subsamples that can be tested on cost stickiness with the model of Anderson et al. (2003).

Table 2

Subsamples

High Low

(Above Median) (Under Median)

Market-to-Book ratio _{P/B High Sample} _{P/B Low Sample}

(P/B)

Price-Earnings ratio _{P/E High Sample} _{P/E Low Sample}

(P/E)

A z-score is used to test if there is a significant difference between the estimated coefficients. The p-value was determined based on a one-tail hypothesis with a significance level of 5%.

Statistical analyses can be heavily influenced by outliers. There are a few techniques to reduce the influence of this outliers. One of them is trimming. This technique excludes outliers by cutting-off the data below, for example, the 5th_{percentile and above the 95}th_percentile. Another technique is to winsorize the extreme values. By using this technique, the data below, for example, the 5th_{percentile will be set to the 5}th_{percentile, and the data above the 95}th percentile will be set to the 95th_{percentile. In contrast to trimming, where observations are lost,} winsorizing doesn’t eliminate values. Using this technique may also lead to a smaller mean squared error (Kokic & Bell, 1994). In this study the extreme values will be winsorized to the 1st and 99th_percentile.

3.3 Control variables

In model 2, there will be a number of control variables included. The first control variable is the dummy for successive decrease. According to Anderson et al. (2003), costs are less sticky in periods where revenue also declined in the preceding period. The second and third control variable are asset intensity (logarithm of assets divided by sales revenue) and employee intensity (logarithm of employees divided by sales revenue). These control variables are included since adjustment cost tend to be higher when the firm relies more on self-owned assets and employees than on materials and services purchased from external suppliers (Kama & Weiss, 2008). The

(14)

fourth and last control variable is stock performance. According to Chen et al. (2013), the stock performance interaction coefficient is positive, indicating less cost stickiness. The stock performance will be calculated as the logarithm of one plus the percentage of capital gain and paid dividend based on the stock price at the beginning of the fiscal year.

For this study, the original model of Anderson et al. (2003) is extended by including some control variables. In model 2 these control variables are included.

Model 2: log ( 𝑆𝐺&𝐴𝑖,𝑡 𝑆𝐺&𝐴𝑖,𝑡−1 ) = 𝛽0+ 𝛽1log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1 ) + 𝛽2 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1 ) + 𝛽3 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1 ) ∗ 𝑆𝑢𝑐𝑐𝑒𝑠𝑠𝑖𝑣𝑒_𝐷𝑒𝑐𝑟𝑒𝑎𝑠𝑒_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏 + 𝛽4 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1 ) ∗ log ( 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 ) + 𝛽5 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1 ) ∗ log (𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 ) + 𝛽6 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1 ) ∗ log (1 +𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐺𝑎𝑖𝑛𝑖,𝑡+ 𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑𝑖,𝑡 𝑆𝑡𝑜𝑐𝑘 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡−1 ) + 𝜀𝑖,𝑡.

In this extended model the 𝑆𝑢𝑐𝑐𝑒𝑠𝑠𝑖𝑣𝑒_𝐷𝑒𝑐𝑟𝑒𝑎𝑠𝑒_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏 represents the successive

decrease. The asset intensity and employee intensity if are represented by log(𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡 /

𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡) and log(𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠𝑖,𝑡 / 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡). And the stock performance is represented by

log (1 + ((𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐺𝑎𝑖𝑛𝑖,𝑡+ 𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑𝑖,𝑡) / 𝑆𝑡𝑜𝑐𝑘 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡−1)) . A description of the created

(15)

4 Results

4.1 Descriptive statistics

Panel A of table 3 provides descriptive statistics on sales revenue and SG&A cost within the sample. The 9.711 firms in the sample have, in the period from 1998 to 2015, a mean sales revenue of $ 2.505,62 million with a median of $ 254,16 million. In this same period, the mean SG&A cost where $ 444,78 million with a median of $ 55,20 million. The SG&A cost as percentage of revenue averaged at 29,88% with a mean of 20,98%. Overall, these statistics are very similar to the statistics of Anderson et al. (2003), with the exception that the values of the sample, in absolute amounts, in the period from 1979 to 1998 are much lower.

Panel B describes the distribution of the independent variables. The average market-to-book ratio and price-earnings ratio are respectively 2,539 and 9,421. Both means are based on the winsorized values. The medians are respectively 1,743 and 11,600.

Panel C describes the distribution of the control variables. These ratios are, just as the independent variables, based on the winsorized values. The average asset intensity ratio and employee intensity ratio are respectively 1,295 and 6,032 with a median of respectively 0,965 and 4,359. The average stock had a return of 22,24% with a median of 1,50%.

Table 3

Descriptive Statistics

Panel A: Distribution of sales revenue and SG&A cost

Mean Deviation Standard Median 25

th

Percentile 75

th

Percentile Sales revenue (Million $) 2.505,62 12.501,55 254,16 56,38 1.122,68 SG&A cost (Million $) 444,78 2.120,48 55,20 14,41 211,24 SG&A cost as percentage

of sales revenue 29,88% 20,98% 25,02% 13,59% 40,83%

Panel B: Distribution of independent variables

th Percentile 75 th Percentile Market-to-Book Ratio 2,539 4,106 1,743 0,957 3,124 Price-Earnings Ratio 9,421 13,278 11,600 -1,894 22,258

Panel C: Distribution of control variables

th

Percentile 75

th

Percentile

Asset Intensity Ratio 1,295 1,103 0,965 0,630 1,531

Employee Intensity Ratio 6,032 6,226 4,359 2,593 6,940

(16)

Table 4 provides a Pearson correlation coefficient between the dependent variables, independent variables and control variables. As expected, there is a strong and significant correlation between the logarithm of change in SG&A cost and the logarithm of change in sales revenue (0,680). This indicates that change in SG&A cost and change in sales revenue move in the same direction, suggesting that the one (sales revenue) might have influence the other (SG&A cost), or that they influence each other, what is probably not the case. However, there is no strong correlation between all other variables. Even the market-to-book ratio and price-earnings ratio, which both functioned as proxies for potential growth, have a weak correlation of only 0,061. Even though they are both proxies for growth potential, it does seems to be wrong to assume that firms with a high market-to-book ratio would also have a high price-earnings ratio.

Table 4

Pearson Correlation Matrix

SG&A Cost Revenue Sales Ratio P/B Ratio P/E Ratio A.I. Ratio E.I. Stock. Perf. SG&A cost 1,000 Sales revenue 0,680 1,000 P/B Ratio 0,065 0,078 1,000 P/E Ratio 0,069 0,097 0,061 1,000 A.I. Ratio 0,070 0,019 -0,015 0,059 1,000 E.I. Ratio -0,034 -0,071 -0,014 -0,023 -0,081 1,000 Stock. Perf. 0,048 0,153 0,128 0,112 -0,001 -0,008 1,000

All correlations are significant at a 5% significance level, besides the one in italics

4.2 Multivariable analyses

4.2.1 Asymmetry of selling, general & administrative cost

In table 5, the regression of the full sample is presented, based on model 2. The full sample contains of 55.996 firm years of 9.711 firms with a period ranging from 1998 to 2015. Based on the t-values and the F-value we can conclude that the results of this test are very significant. This model explains 47,43% of the variance in the dependent variables according to the adjusted R². Without including the control variables, the adjusted R² would be 0,4681 instead of 0,4743. This means that, by including the control variables, the explained variance in the dependent variables increases with 0,62%. This means that including the control variables doesn’t add a lot of explaining power to the model.

(17)

The results of this regression indicates a significant presence of cost stickiness, because β₁=0,655 (p<0,01) and β₂=-0,238 (p<0,01). This means that an increase in sales revenue of 1% leads to an increase in SG&A cost of 0,655%, while a decrease in sales revenue of 1% leads to a decrease of 0,417%. These findings are consistent with the findings of Anderson et al. (2003) and confirm that there is cost stickiness in this sample.

Table 5

Regression results - Full sample

Variables Prediction Coef. Std. Err. t-value p-value

β₀: Intercept 0,016 0,001 14,81 0,000 β₁: log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) + 0,655 0,004 180,12 0,000 β₂: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) - -0,238 0,011 -21,78 0,000 β₃: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ 𝑆𝑢𝑐𝑐𝑒𝑠𝑠𝑖𝑣𝑒_𝐷𝑒𝑐𝑟𝑒𝑎𝑠𝑒_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏 + 0,187 0,010 18,12 0,000 β₄: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log ( 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡) - -0,060 0,004 -14,02 0,000 β₅: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log (𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 ) - 0,022 0,003 7,72 0,000 β₆: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log (1 +𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑔𝑎𝑖𝑛𝑖,𝑡 𝑆𝑡𝑜𝑐𝑘 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡−1) + 0,024 0,004 5,77 0,000 Number of observations 55.996 F-value 8.421,62 R² 0,4744 p-value 0,0000 Adjusted R² 0,4743

4.2.2 Cost asymmetry and growth expectations

The hypothesis of this study is that costs exhibit greater stickiness in firms with higher growth expectations. To test this hypothesis, the sample is spitted up in subsamples which enables us to compare the coefficients of firms with a high growth expectation to firms with a low growth expectation. In this study, subsamples were created by making median splits on the independent variables. Each subsample was independently tested on cost stickiness by means of a regression analyses which was based on model 2. After the analyses, the regression coefficients of the subsamples, that where created based on the same independent variable, could be compared. A z-score is used to test if the difference in coefficients between the subsamples is significant. The results of the regression analyses and z-scores can be found in table 6 and 7.

(18)

4.2.2.1 Difference in cost stickiness – based on market-to-book ratio

Table 6 presents the regression of the P/B High and P/B Low subsample, including the z-score and its p-values. Cost stickiness can be found in both samples, however, there seem to be no significant difference between the coefficients of P/B High and P/B Low, except the estimated coefficients of the intercept (β₀) and stock performance (β₆) which are beyond the scope of this study. This indicates that there is no correlation between the market-to-book ratio and the amount of cost stickiness. Based on this test there seems to be no support for the hypothesis.

Table 6

Regression results - P/B Subsamples

P/B High P/B Low Difference Variables Prediction Coef. Std. Err. Coef. Std. Err. z-score p-value

β₀: Intercept 0,024 0,001 0,008 0,002 7,16 0,000 β₁: log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) + 0,650 0,005 0,656 0,006 0,77 0,221 β₂: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) - -0,252 0,020 -0,244 0,014 0,33 0,371 β₃: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ 𝑆𝑢𝑐𝑐𝑒𝑠𝑠𝑖𝑣𝑒_𝐷𝑒𝑐𝑟𝑒𝑎𝑠𝑒_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏 + 0,187 0,019 0,182 0,013 0,22 0,413 β₄: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log ( 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡) - -0,055 0,009 -0,060 0,005 0,49 0,312 β₅: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log ( 𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 ) - 0,018 0,006 0,024 0,004 0,83 0,203 β₆: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log (1 + 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑔𝑎𝑖𝑛𝑖,𝑡 𝑆𝑡𝑜𝑐𝑘 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡−1) + 0,063 0,009 0,020 0,005 4,18 0,000 Number of observations 27.998 27.998 Adjusted R² 0,4802 0,4531 F-value 4.311,72 3.866,57 p-value 0,000 0,000

4.2.2.2 Difference in cost stickiness – based on price-earnings ratio

Table 7 presents the regression of the P/E High and P/E Low subsample, including the z-score and its p-values. Similar to the P/B High and P/B Low sample, cost stickiness is found in both samples. However, between these subsamples there seems to be a significant difference between the β₂ coefficients (cost stickiness coefficients), according to the value of the z-score. This indicates that there might be a correlation between the price-earnings ratio and the amount of cost stickiness. In contradiction to the previous subsample analyses, there seems to be support for the hypothesis if we consider the P/E ratio as the right measure for growth expectations.

(19)

According to this subsample analyses it is estimated that, in a one-year period, SG&A costs increased 0,65% per 1% increase of sales revenues for firms with a high price-earnings ratio. And that SG&A cost decrease 0,37% per 1% decrease of sales revenues. While SG&A cost in firms with a low price-earnings ratio increase 0,66% per 1% increase of sales, and they decrease 0,42% per 1% decrease of sales revenue. This difference of 0,04 percentage point per 1% decrease in sales revenue is statistically significant and is in line with the hypothesis.

There also seems to be a significant difference between the estimated coefficients of β₃, which measures the effect of the successive decrease in sales revenue. These coefficients indicate that the degree of stickiness is lower if the sales revenue declining period was preceded by a sales revenue declining period. According to the z-score, firms with a high P/E ratio will experience relatively less cost stickiness in the second year of decline in sales revenue. A possible explanation might be that managers compensate the decrease in SG&A cost in periods where sales revenue declining was preceded by a sales revenue declining period.

The difference in the estimated coefficients of the intercept (β₀) and other control variables (β₄, β₅ and β₆) is also significant. However, explaining why this happens and why β₅ coefficient is positive instead of negative, which was predicted, is beyond the scope of this study.

Table 7

Regression results - P/E Subsamples

P/E High P/E Low Difference Variables Prediction Coef. Std. Err. Coef. Std. Err. z-score p-value

β₀: Intercept 0,021 0,001 0,012 0,002 4,02 0,000 β₁: log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) + 0,651 0,005 0,656 0,005 0,71 0,239 β₂: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) - -0,280 0,019 -0,240 0,015 1,65 0,049 β₃: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ 𝑆𝑢𝑐𝑐𝑒𝑠𝑠𝑖𝑣𝑒_𝐷𝑒𝑐𝑟𝑒𝑎𝑠𝑒_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏 + 0,233 0,022 0,181 0,013 2,03 0,021 β₄: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log ( 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡) - -0,039 0,009 -0,061 0,006 2,03 0,021 β₅: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log ( 𝐸𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 ) - 0,050 0,005 0,018 0,004 5,00 0,000 β₆: 𝐷𝑒𝑐𝑟𝑒𝑎𝑠_𝐷𝑢𝑚𝑚𝑦𝜄,𝜏∗ log (𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1) ∗ log (1 + 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑔𝑎𝑖𝑛𝑖,𝑡 𝑆𝑡𝑜𝑐𝑘 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡−1) + 0,073 0,013 0,019 0,005 3,88 0,000 Number of observations 27.999 27.997 Adjusted R² 0,4869 0,4533 F-value 4.428,65 4.028,12 p-value 0,000 0,000

(20)

5 Conclusion

This paper examines whether asymmetric cost behaviour is associated with the potential growth expectations of firms. According to previous literature, managers deliberately decrease committed resources when there is a decline in sales revenue. However, when there is uncertainty about whether or not the demand will recover, managers might also choose to delay the decrease of committed resources, since restoring committed resources leads to adjustment costs. This is one of the factors that contribute to cost stickiness. Cost stickiness is partly the result of managers’ decisions about resource adjustments. One of the factors that presumably affect the decisions of managers, is their expectation about future demand. Managers that expect that future demand will increase will, probably, be more likely not to decrease committed resources to avoid adjustment cost. While managers who don’t expect a future increase might be more willing to decrease committed resources. Hence, cost might be stickier in firm where managers have high growth expectations.

Using a sample of 55.996 observation in the period from 1998 to 2015, there has been found evidence that firms with a higher price-earnings ratio have a lower downward adjustment of SG&A cost when there is a decrease in sales revenue. This indicates that there is a relation between the price-earnings ratio, which is an earnings growth indicator, and the amount of cost stickiness. In contrast to these findings, no indication is found that market-to-book ratio, which is an indicator for expected return on equity, has any relation with the amount of cost stickiness.

Results of this study contribute to the literature by providing extra insight in how the decision of managers are effected by the growth expectation of firms. These new insides are relevant for people who rely on cost accounting data, because it gives them better understanding of the behaviour of SG&A cost, which make up 30% of sales revenue in this sample.

This study is subject to a number of limitations. The first limitation is that the sample consists only of S&P 1500 firms. By this means, it is not possible to generalize this results for non-S&P 1500 firms. These findings can also not be generalized form financial institutions and private utility firms, since these where excluded in this sample. Further research is needed to find out if these findings are consistent in other settings. The second limitation is that the used proxies for managers’ growth expectations, which where market-to-book ratio and price-earnings ratio, might not be the correct proxies. There seems to be no correlation between the proxies, even though they are used as a measure for the same thing. This might have led to the fact that the results of the P/B subsample are different from the results of the P/E subsample.

(21)

References

Anderson, M., Banker, R.D., & Janakiraman, S.N. (2003). Are selling, general and administrative costs “sticky”? Journal of Accounting Research, 41(1), 47–64.

Atkinson, A.A., Kapan, R.S., Matsumura, E.M., Young, S.M. (2012). Management Accounting.

Information for decision making and strategy execution. Essex: Pearson Education

Balakrishnan, R., & Gruca, T.S. (2008). Cost Stickiness and Core Competency: A Note.

Contemporary Accounting Research, 25(4), 993–1006. doi:10.1506/car.25.4.2

Balakrishnan, R., Peterson, M.J., & Soderstrom, N.S. (2004). Does Capacity Utilization Affect the “Stickiness” of Cost?. Journal of Accounting, Auditing & Finance, 19,(3), 283–300. doi:10.1177/0148558X0401900303

Banker, R.D., & Byzalov, D. (2014). Asymmetric Cost Behavior. Journal of Management Accounting

Research, 26(2), 43–79. doi:10.2308/jmar-50826

Banker, R.D., Huang, R., & Natarajan, R. (2011). Equity Incentives and Long-Term Value Created by SG&A Expenditure. Contemporary Accounting Research, 28(3), 794–830. doi:10.1111/j.1911-3846.2011.01066.x

Baumgarten, D. (2012). The Cost Stickiness Phenomenon. Wiesbaden: Gabler Verlag doi:10.1007/978-3-8349-4131-2

Chen, C.X., Gores, T. & Nasev, J. (2013). Managerial Overconfidence and Cost Stickiness. doi:10.2139/ssrn.2208622

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. doi:101111/j.1911-3846.2011.01094.x

Chen, L. & Zhao, X. (2006). On the relation between the market-to-book ratio, growth opportunity, and leverage ratio. Finance Research Letters, 3(4), 253–266. doi:10.1016/j.frl.2006.06.003

Cooper, R.. & Kaplan, R. (1998). The design of cost management systems: text, cases, and readings. Upper Saddle River , NJ: Prentice-Hall.

Dierynck, B., Landsman, W.R., & Renders, A. (2012). Do managerial incentives drive cost behaviour? Evidence about the role of the zero earnings benchmark for labor cost behaviour in private Belgian firms. The Accounting Review, 87, 1219–1246. dio:10.2308/accr-50153

Fama, E.F. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work. The

(22)

Guenther, T.W., Riehl, A., & Rößler, R. (2014). Cost stickiness: state of the art of research and implications. Journal of Management Control, 24(4), 301–318.

doi:10.1007/s00187-013-0176-0

Jensen, M. C., & Meckling, W. H. (1976). Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics , 3(4), 305–360.

Kama, I., & Weiss, D. (2011). Do managers' deliberate decisions induce sticky costs? Working paper.

Kokic, P.N., & Bell, P.A. (1994). Optimal Winsorizing Cutoffs for a Stratified Finite Population Estimator. Journal of Official Statistics, 10(4), 419–435.

Malkiel, B.G., & Cragg, J.G. (1970). Expectations and the Structure of Share Prices. The American

Economic Review, 60(4), 601–617.

Noreen, E. (1991). Conditions under which activity-based cost systems provide relevant costs.

Journal of Management Accounting Research, 91(3), 159–168.

Noreen, E. & Soderstrom, N. (1997). The Accuracy of Proportional Cost Models: Evidence from Hospital Service Departments. Review of Accounting Studies, 2(1), 89–114. dio:10.1023/A:1018325711417

Penman, S.H. (1996). The Articulation of Price-Earnings Ratios and Market-to-Book Ratios and the Evaluation of Growth. Journal of Accounting Research, 34(2), 235–259. doi:10.2307/2491501

Subramaniam, C. & Weidenmier, M.L. (2003). Additional Evidence on the Sticky Behaviour of Costs. TCU Working Paper. doi:10.2139/ssrn.369941

Venieris, G., Naoum, V.C., & Vlismas, O. (2015). Organisation capital and sticky behaviour of selling, general and administrative expenses. Management Accounting Research, 26, 54–82. doi:10.1016/j.mar.2014.10.003

Weiss, D. (2010). Cost Behavior and Analysts’ Earnings Forecasts. The Accounting Review, 85(4), 1441–1471. doi:10.2308/accr.2010.85.4.1441

(23)

Appendix

Appendix 1: Downloaded database variables

Compustat – Capital IQ > Compustat Monthly Update > North America > Fundamentals Annual

General variables

Date Variable: Fiscal Year

Date Range: 1998-01 to 2015-12

Identifying Information: Company Name

Identifying Information, cont.: SIC – Standard Industry Classification Code

Cost stickiness variables

Income Statement Items: SALE – Sales/Turnover (Net)

XSGA – Selling, General and Administrative Expense

Independent variables

Balance Sheet Items: AT – Assets - Total

LT – Liabilities - Total

Income Statement Items: EPSPX – Earnings Per Share (Basic) Excluding Extraordinary Items Supplemental Data Items: MKVALT – Market Value - Total - Fiscal

PRCC_F – Price Close - Annual - Fiscal

Control variables

Miscellaneous Items: EMP – Employees

(24)

Appendix 2: Created variables

Cost stickiness variables

SGA – Change in SG&A cost log ( 𝑆𝐺&𝐴𝑖,𝑡

𝑆𝐺&𝐴_{𝑖,𝑡−1})

REV – Change in sales revenue log ( 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡

𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1)

DEC – Decrease Dummy: Value 1 when sales revenue decreases between periods t-1 and t, and value 0 otherwise.

So value 1 if 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝜄,𝜏< 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1

And value 0 if 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝜄,𝜏≥ 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1

Independent variables

PB – Market-to-Book Ratio: log ( MKVALT – Market Value – Total – Fiscal

(AT – Assets− Total)−(LT – Liabilities − Total))

PE – Price-Earnings Ratio: log ( PRCC_F – Price Close – Annual – Fiscal

EPSPX – Earnings Per Share (Basic)Excluding Extraordinary Items)

Control variables

SUC – Successive Decrease Dummy: Value 1 when sales revenue decreases between periods t-2 and t-1 and between t-1 and t, and value 0 otherwise.

So value 1 if 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝜄,𝜏< 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1< 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−2

And value 0 if 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝜄,𝜏≥ 𝑅𝑒𝑣𝑒𝑛𝑢𝑒𝑖,𝑡−1

AI – Asset Intensity Ratio log ( AT−Assets−Total

SALE – Sales/Turnover (Net))

EI – Employee Intensity Ratio log ( EMP−Employees

SALE – Sales/Turnover (Net))

SP – Stock Performance log (1 +PRCC_F−DVPSP_Fi,t−1+DVPSP_Fi,t

DVPSP_Fi,t−1 )

Subsampling

P/B Sample: Value HIGH when the P/B ratio is above the median, and value LOW is the P/B ratio is higher than or equal to the median.

So value HIGH if P/B ratio > P/B ratio median And value LOW if P/B ratio ≤ P/B ratio median

P/E Sample: Value HIGH when the P/E ratio is above the median, and value LOW is the P/E ratio is higher than or equal to the median.

So value HIGH if P/E ratio > P/E ratio median And value LOW if P/E ratio ≤ P/E ratio median

Cost stickiness and the effect of managers’ growth expectations