The effect of institutional shareholders on cost stickiness

(1)

The effect of institutional

shareholders on cost stickiness

Name: Fraukje Ligtvoet Student number: 11080914

Thesis supervisor: Mr. M. Schabus MSc

Date: 22 June 2017

Word count: 12,426

MSc Accountancy & Control, specialization Control

(2)

1

Statement of Originality

This document is written by student Fraukje Ligtvoet 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)

2

Abstract

This research aims to contribute to the existing literature by researching whether institutional shareholders affect cost stickiness. This research investigates this relationship using a sample of North America publicly traded companies from 2002 to 2014. It is expected that when a company has more institutional shareholders there will be more cost stickiness than when shareholders are merely interested in the short-term performance of a company. This is based on prior literature, which suggests the ability of institutional shareholders to monitor management could affect managements decisions. Since institutional shareholders have a long-term horizon, they are not only interested in the current value, but also in the future value of the company. This could cause more cost stickiness since cost adjustments could be delayed when a sales decrease occurs, until the period of the sales decrease is known. When the sales decrease is not permanent, it is not desirable to cut the costs because this will hurt the future value of the company. The hypothesis of this research is that the fraction of institutional shareholders is positively related to cost stickiness. The findings of this research suggest that this is not the case. Other literature also suggests the cost stickiness will be higher with more short-term shareholders than with long-term (institutional) shareholders. This research could not find significant results for the hypothesis, but I am convinced that the relationship between cost stickiness and (institutional) shareholders are worth future research. There is not much prior literature on this relationship, and there are different arguments to build on expectations of this relationship. Future research could contribute to a better understanding of the relationship between cost stickiness and (institutional) shareholders and how this relationship will affect companies and their stakeholders.

(4)

3

Abstract ... 2 1. Introduction ... 4 2. Literature review ... 6 2.1 Cost stickiness ... 6 2.2 Institutional shareholders ... 12 2.3 Hypothesis development ... 16 3. Methodology ... 18 3.1 Sample selection ... 18 3.2 Measurement of variables ... 18 3.2.1 Cost stickiness ... 18 3.2.2 Institutional shareholders ... 19 3.2.3 Control variables ... 20 3.3 Empirical model ... 21 4. Results ... 23 4.1. Descriptive Statistics ... 23 4.1.1 Correlation ... 25 4.2 Multivariate Analysis ... 26 5. Conclusion ... 31 Bibliography ... 34

(5)

4

1. Introduction

In the traditional models of cost behaviour costs are divided into fixed and variable costs. With variable costs, there is a linear relationship between costs and volume. Therefore, it was expected that costs are mainly determined by the magnitude of the cost driver. However, the relationship between costs and adjustments in cost drivers is not constant since management decisions determine the level of resources a company needs and hence adjustments. Costs will decrease less with a sales decrease than they increase with an equal increase in sales, which is defined as cost asymmetry or cost stickiness (Bugeja, Lu, & Shan, 2015). Asymmetric frictions in making resources adjustments cause sticky costs, which force limits or slow down the downward adjustment process more than the upward. These adjustments will replace current resources that caused adjustment costs, which can be training cost for newly hired employees or severance pay when employees are fired (Anderson, Banker, & Janakiraman, 2003).

Who owns shares of a company can have a major impact on the incentives of management and the board of directors. Management behaviour is influenced by distribution of ownership (Jensen & Warner, 1988). There are three types of shareholders; each has its own monitoring technology. There are two types of institutional shareholders and the third type consists of individual shareholders. Individual shareholders are not capable of monitoring. Because of their dispersed ownership, monitoring costs are too high for each individual shareholder. On the other hand, institutional shareholder can assess managerial performance. Institutional shareholders are divided into potentially active and passive institutional shareholders. The active ones consist of investment advisers and investment companies. These types have more skilled employees, can collect more information, face less regulation, and have less potential for business relationships with companies. Bank trust departments and insurance companies represent the passive institutional shareholders (Almazan, Hartzell, & Starks, 2005). Several researchers1_{found evidence}

that confirms institutional shareholders can afford to monitor more active than smaller or less informed investors. As suggested by Almazan et al. (2005), monitoring by institutional shareholders may influence the decisions of management. Institutional shareholders are also able to monitor whether management implements their preference. That institutional shareholders could use these technologies, as well as their time horizon, may affect managers. Since institutional shareholders generally hold their shares for a long period, they are expected to affect managers differently than shareholders who have a short-term horizon (Gaspar, Massa, & Matos, 2005).

(6)

5

Based on these arguments, I examine whether cost stickiness is affected by institutional shareholders. That is, I expect these kinds of shareholders to prefer management make cost adjustments that take a relatively longer horizon into account. I think this will have an impact on managerial behavior because institutional investors have the ability to monitor management’s behaviour. I, therefore, expected that cost stickiness will increase with the institutional investor base of a company. I hypothesize that the fraction of institutional shareholders is positively related to cost stickiness.

In this paper, the incentives of a company’s shareholders will be examined instead of the managers’ incentives. Therefore, the relationship between institutional shareholders and cost stickiness will be researched. With this research, the paper will contribute to the existing literature. To my knowledge, there is no research about how a company’s shareholders affect cost stickiness. It is expected institutional shareholders’ interest will affect managers’ behaviour and cost stickiness since institutional shareholders are the most important stakeholder of the company. The only prior literature about the relationship between shareholders and cost stickiness, by Calleja, Steliaros, and Thomas (2006), was not specifically about institutional shareholders. This research was conducted with shareholders in general instead of institutional shareholders compared to short-term shareholders. Calleja et al. (2006) found that in the United Stated the level of cost stickiness for operating costs is lower than in Germany and France. One of the reasons for the lower level of cost stickiness they mentioned is that companies in the United States are more focused on maximization of shareholders value. They suggested companies in the United States are more motivated to reduce costs when revenue decreases because they are under scrutiny and pressure to maximize shareholders value. However, I suppose this finding holds for short-term shareholders since they are merely interested in the current shareholders value. Since institutional shareholders are also interested in the future value of the company, I suppose they will not pressure companies to cut costs immediately because this could hurt the future value.

The remainder of this paper is structured as follows. The next section presents a literature review about cost stickiness and institutional shareholders. The hypothesis development follows after the literature review. In section 3, the methodology used in this research is explained, which consists of the sample, the measurement of the variables, and the empirical model. In section 4, the analyses are presented and the results are shown. Finally, section 5 includes the discussion and the conclusion of this research.

(7)

6

2. Literature review

2.1 Cost stickiness

Anderson et al. (2003) called costs that are not strictly proportional to activities, sticky. Costs are sticky when the increase of the cost associated with the increase of activities is higher than the decrease in the cost with an equal decrease of the activities. In their research, Anderson et al. (2003) found that the percentage increase of SG&A (Selling, General and Administrative) costs associated with an increase in sales revenue was higher than the percentage decrease in SG&A costs associated with an equal decrease in sales revenue. According to their research, when the sales revenue increased by 1%, SG&A costs increased by 0.55%, but if the sales revenue decreased by 1%, the SG&A costs only decreased by 0.35%. This reaction can be explained by a different cost behaviour of managers when the future demand is uncertain. Managers will delay the reduction of cost of resources until they have more certainty of the demand. Also, the opposite of cost stickiness was found in prior research. Weiss (2010) called this phenomenon anti-sticky, which appears in a situation when the increase of costs is less when activities increase than the decrease of costs with an equal decrease of activities.

Not only SG&A costs were tested for cost stickiness in prior literature. Calleja et al. (2006) found that also total operating costs are sticky. On average, from the companies in their sample, when revenue increased by 1%, operating costs increased with 0.97%. However, when revenue decreased by 1%, operating costs only decreased by 0.91%. In their research, they compared cost stickiness in companies across the United Kingdom, the United States, Germany, and France. They found the level of cost stickiness was higher in Germany and France, and they explained this by Germany’s and France’s systems of governance and the pressure of the market of corporate control. The United Kingdom and the United States, on the other hand, are under more scrutiny and are more focused on maximization of shareholders’ value, which led to a lower level of cost stickiness. However, I assume their research is based on short-term shareholders since they are merely interested in the current shareholders value. Institutional shareholders, on the other hand, are also interested in the future value of the company. I assume they will not pressure managers to cut the costs immediately since this could hurt the future value.

Furthermore, Calleja et al. (2006) found that companies that have attracted more debt show, on average, no cost stickiness. They suggested this occurs because these companies have to meet their interest payments and could be more monitored by their creditors. For these reasons, these companies’ managers are encouraged to make sure they have a cost structure, which could react flexibly to adjustments in the business. Calleja et al. (2006) also found that companies from

(8)

7

the United States, United Kingdom, and France that have high employee-intensity show more cost stickiness with a sales decrease.

Banker, Byzalov, Ciftci, and Mashruwa (2014) stated that cost stickiness is not driven by a specific part of SG&A costs, but rather cost stickiness occurs in a broarder range of expenses, which also applies to the main elements of operating costs and holds for the quantity of labor input. Furthermore, for all the cost categories (SG&A, advertising, R&D, COGS, and number of employees) they tested, the reaction of costs to a current sales increase was considerably stronger with a previous sales increase than with a previous sales decrease. Regarding the number of employees, when managers are positive, because of a previous sales increase, they are more willing to hold some slack in the current sales decrease because this lets them decrease not only the present adjustment costs, for example compensation for fired employees, but also the adjustment costs that should be made in the future, like hiring costs for new employees, which they would make when the future expected higher demand is realized.

Additionally, when Banker et al. (2014) used another model than that of Anderson et al. (2003), they found other results than prior literature. They found cost stickiness is followed by a previous sales increase and anti-stickiness is followed by a previous sales decrease. These findings are in line with the theory of asymmetric cost behaviour. This theory clearly recognizes managers have influence on cost behaviour and is based on resource adjustment costs and deliberate managerial decisions. In this view, management’s forecasts of future sales on the company’s current resources is one of the influences management has on cost behaviour.

Jiang, Frazier, and Prater (2006) explained in their paper the difference between SG&A and operating costs: operating costs also includes the COGS (Cost Of Goods Sold). SG&A costs are mostly driven by sales activities and, therefore, represent the indirect costs, and the operating costs represent the direct costs.

Subramaniam and Weidenmier (2003) researched both SG&A and operting costs and found that these costs are sticky when revenue changes by more than 10%. If the increases in the revenue is more than 10%, managers should increase capacity by increasing the company’s resources. When the revenue decreases by more than 10%, managers are not able or do not want to change the company’s resources, which causes cost stickiness. Also, Shust and Weiss (2014) researched operating costs and found operating costs are significantly more sticky after depreciation than they are before depreciation. They suggested that prior investment decisions, which lead to depreciation in the current period, caused a higher level of cost stickiness in the current period. They further suggested that when costs are measured by reported costs, the GAAP (Generally Accepted Accounting Principles) requirements to report depreciation caused the higher

(9)

8

level of cost stickiness. Also, Bugeja et al. (2015) found in their research in Australia that after the adoption of IFRS (International Financial Reporting Standards) the extent of cost stickiness increased.

Bugeja et al. (2015), furthermore, resarched whether different industries show different cost stickiness behaviour. For their research, they used the findings of Ely (1991), which show that each industry has its own production, which includes product markets, technology, and regulatory environments. The industries are determined by the SIC (Standard Industrial Classification) codes. Through the difference in production across different industries, also the characteristics of those industries vary, which affects the accounting variables (Ely, 1991). The different accounting variables can affect cost stickiness. Since Bugeja et al. (2015) conducted their research in Australia, they used the ASX (Australian Securities Exchange) industry classifications, based on the GICS (Global Industry Classification Standard) code, instead of the SIC codes. After eliminating the financial and utility industries, they use six industries: manufacturing, retail, resources, services, construction, and other industries. Bugeja et al. (2015) found that manufacturing, services, and other industries showed cost stickiness behaviour. The retail, resources, and construction industries do not show cost stickiness in their sample. The manufacturing industry showed the highest level of cost stickiness. Also, Subramaniam and Weidenmier (2003) found that the manufacturing industry has the most sticky costs. The authors explained this by the high amounts of inventory and fixed assets in this industry. Because of the compititve market, the merchandising industry has the least cost stickiness. They also found cost stickiness in the service and financial industries; in the services industry, stickiness was driven by inventory and employee intensity and, in the financial industry, by interest cost.

Additionally, Bugeja et al. (2015) found that the existence of cost stickiness is less when the CEO has a short-term focus, thus CEOs in their final period. Also, when a company has more nonexecutive directors and boards with non-CEO chairs, cost stickiness will decrease. Their research also showed that, in general, Australian companies have less sticky costs than United States companies. Finally, they found, in Australia, firm-level and economy-wide measures for managerial incentives and resources cost adjustments will determine the extent of cost stickiness. Bugeja et al. (2015) also found when managers have strong incentives to avoid earnings decreases or losses, there is more asymmetric cost behaviour. They found that 16% of the companies in their sample avoided reporting a loss, and 18% avoided reporting decreases in accounting earnings.

(10)

9

According to Balakrishnan, Petersen, and Soderstrom (2004), there are two factors, regarding managers’ reactions, that lead to cost stickiness. The first is magnitude of the change. When change appears, transaction costs, also called adjustments cost, follows and was found to be larger with larger changes in activities than with smaller changes. The second factor is capacity utilization of a company. When there is excess capacity, the reaction to change in activities will be different than the reaction when there is strained capacity.

Banker and Hughes (1994) mentioned in their paper another reason for cost stickiness, other than the manager’s reaction to sales decrease: certain costs are fixed, so if the sales decrease, these costs cannot decrease immediately and the proportion between SG&A costs and sales will increase.

However, the influence of managers’ cost behaviour shows that SG&A costs are influenced by the company managers’ reactions to certain situations. The decisions managers make in these situations can be influenced by personal considerations, which lead to agency problems (Anderson et al., 2003). According to Jensen and Meckling (1976), agency problems arise when an agent acts on behalf of a principal and the agent has some authority to make decisions. In a relationship like this, if they both want to benefit the most, the agent will not always act in the best interest of the principal. With certain incentives, the goals of the principal and the agent can be aligned. Hence, managers should maximize the shareholders’ value and not their personal interests, otherwise agency problems will occur.

More research about the effect manager’s incentives have on cost stickiness was been conducted. According to Kama and Weiss (2013), incentives to meet earnings targets induce deliberate resource adjustments that reduce cost stickiness. The incentive to meet earnings targets will be higher when sales fall than when sales rise.

Furthermore, Cannon (2014) found that cost stickiness will arise when the demand grows and managers have more costs expanding the capacity than they have when the demand decreases. He also found that cost stickiness occurs when sales are used as a proxy for activities that generate costs. Managers then will adjust the selling prices asymmetric to the capacity adjustments. When the demand decreases, managers would rather set lower selling prices than remove capacity, but when the demand increases, they expand capacity rather than set higer prices. Moreover, his findings show that anti-stickiness occurs when the demand decreases and managers save more costs by removing capacity than they save by removing capacity when the demand is increasing. Capacity costs will became more sticky from capacity and selling price adjustments together with the policy managers have for idle capacity. For the future earnings, it is important to identify the managers’ activities that are related to sticky cost behaviour because cost adjustments related to

(11)

10

different activities could generate different expectations for the future earnings. Calleja et al. (2006) agreed it is important to identify these activities. They, furthermore, emphasized managers should recognize and manage idle resources and capacity. In the short run, cost stickiness could affect the decisions managers make, which are based on standard costs and which do not behave as they are supposed to. Examples of this decisions could be adjustments in the sales mix, selling prices, investment decisions, marketing activities, and outsourcing and leasing.

Furthermore, the results of Weiss (2010) show companies with more sticky costs have less precise earnings forecasts than companies with less sticky costs. This can be explained by manager’s behaviour on decrease of sales. Companies with more sticky costs have a higher decrease in earnings when the activities drop than companies with less sticky costs. With stickier costs, the cost adjustments made when the activities drop will be less, and therefore, there will be less cost saving. This results in a greater decrease in the earnings, which increases the variability of the earnings distribution and less precise earnings forecasts.

Dierynck, Landsman, and Renders (2012) examined in their research whether the incentives of private Belgian companies have influence on achieving or exceeding the zero earnings benchmark on labour cost behaviour. They found that when companies have incentives to achieve or exceed the zero earnings benchmark, this restricts the rise of labour expenses when activities rise, and when activities drop, they are more willing to cut labour expenses. This causes symmetric labour cost behaviour. When companies have no pressure from the earnings benchmark, they show asymmetric labour cost behaviour. These results are in line with prior literature on SG&A expenses of American publicly traded companies. Furthermore, Dierynck et al. (2012) found that when the sales rise by 1%, this leads to a 0.60% rise in labour expenses, and when the sales drops with 1%, the labour expenses will drop by 0.34%. Additionally, companies who made profit respond to activity changes by adjusting the number of hours employees should work because this will limit harm to their reputation within the company and in the labour market. However, companies that have to achieve or exceed the zero earnings benchmark will, with an activity change, fire employees, which causes direct cost savings. Lastly, small profit companies are more likely to fire the least costly employees (Dierynck et al., 2012).

SG&A costs are a large component of companies’ total expenses, which is why they are controlled carefully (Chen, Lu, & Sougiannis, 2012). Chen et al. (2012) especially examined cost stickiness and agency problems in relation to other economic factors. The authors used the agency problem in two directions: the empire building and the downsize literature. They explained the empire building as when a company grows beyond its maximum capacity or leaves resources unutilized caused by managers’ tendencies to increase their personal utility, like power,

(12)

11

compensation, and status. Bugeja et al. (2015) suggested that empire building will lead to a greater degree of cost stickiness. They stated the degree of cost stickiness is, in addition to cost adjustments, determined by empire building and the corporate governance of a company. Bertrand and Mullainathan (2003) suggested that when empire building appears in a weak corporate governance, managers would increase the firm size. The downsize literature focuses on managers’ disincentives to downsize the expenses. There are several reasons for those disincentives. First, managers can have incentives for monetary and non-monetary advantages from managing complex organizations. Furthermore, at first, downsizing will have benefits for the shareholders instead of the managers. Finally, managers can avoid difficult decisions and costly efforts in connection with downsizing (Chen et al., 2012).

Furthermore, Chen et al. (2012) examined whether corporate governance reduces the effects agency problems have on cost stickiness. One of the variables to measure corporate governance is the percentage of institutional ownership. The higher the percentage of institutional ownership, the stronger the corporate governance because, with higher institutional ownership, the monitored activities will increase, which indicates stronger corporate governance. The results of the research of Chen et al. (2012) show that the positive relationship between agency problems and SG&A cost stickiness is stronger with a weak corporate governance. Their results also show that when agency problems occur in a weak corporate governance, managers wait longer before they decrease the SG&A costs when the sales drop. On the other hand, with strong corporate governance, agency problems are not the main cause of SG&A cost stickiness.

The results of Chen et al. (2012) also show that cost stickiness is positively related to managers’ incentives to act as empire building caused by agency problems. They also found agency costs affect cost stickiness more in mature than in growing companies. Finally, the authors found that when SG&A cost generate high future value, decreasing too much of the SG&A costs in a short period will hurt the company in the long-term. In contrast, when SG&A costs generate low future value, waiting too long to decrease the costs cause useless expenses and possibly lower performance in the long-term. This leads to the fact that, when SG&A costs generate higher value, managers should delay decreasing the SG&A costs, which indicates higher cost stickiness, and with lower value, vice versa.

According to prior research, the financial crisis could influence cost stickiness. Campello, Graham, and Harvey (2010) found in their research that companies in the United States that were financially constrained had to heavily reduce their expenses in 2009. On average, companies reduced their expenses in marketing expenditures by 33%, technology by 22%, dividend payments by 14%, employment by 11%, and capital investment by 9%. Some of the expenses Campello et

(13)

12

al. (2010) mentioned are part of SG&A costs. Companies that were not financially constrained had to cut fewer expenses. Furthermore, they found that during the financial crisis, 86% of constrained companies in the United States rejected attractive investments because they could not raise enough external money. With unconstrained companies, this was 44%. The results of Campello et al. (2010) also show that some companies had to sell assets in 2008 to keep the business running. According to Campello et al.’s (2010) research, the financial crisis could also affect cost stickiness since cost stickiness is caused by managers having to cut expenses in times when sales decrease.

2.2 Institutional shareholders

“Institutional shareholders have become the dominant shareholders in most of the largest corporations in the United States. There is, however, considerable disagreement over the role they play in corporate governance.” (Clyde, 1997, p. 1)

Although institutional shareholders are important for large companies, they do not watch management’s every step. However, some issues need the board’s attention. Shareholders can influence managers so that manangers’ incentives become more aligned with the shareholders’ incentives and can dissuade actions that are in favour of management instead of the shareholders, in the end, taking over when the CEO repeatedly fails (Black, 1992).

Previous research has been conducted about corporate governance, which is affected by the structure of a company’s ownership. Eng and Mak (2003) examined whether corporate governance is connected to voluntary disclosure. They specifically focused on ownership and board structure. Interestingly, ownership structure determines the level of monitoring. The structure of ownership is measured by the percentage of shares owned by managers and blockholders. The percentage of ordinary shares held by the CEO and executive directors is called managerial ownership. Blockholder ownership appears when 5% or more of the ordinary shares are owned by one shareholder. According to Hartzell and Starks (2003), blockholders are different from institutional shareholders in the sense that they buy the shares with the intention to influence decisions. Both managerial and blockholder ownership are important governance mechanisms. Eng and Mak’s (2003) results shows that ownership structure and board composition affect disclosure. They found that with lower managerial ownership, disclosure increases. On the other hand, blockholder ownership is not associated with disclosure.

Corporate governance is not the only regulation that shareholders have to deal with. According to Clyde (1997), legislation for the two most important kinds of institutions, investment and pension funds, are there to decrease institutions’ ability to take direct actions through the

(14)

13

board. Both investment and pension funds have their own legislations. Several authors2_{argued that}

those legislations restrict active shareholders from playing active roles in companies. The legislations make it expensive to own enough shares to influence management. However, the legislations cannot restrict institutional shareholders to facilitate in a takeover. Takeovers can be used to replace the company’s management (Clyde, 1997). Shleifer and Vishny (1986) explained these takeovers in the following situation. Sometimes monitoring activities, to improve the company’s operating strategy, are not enough to maximize a company’s profit. To replace the current management is then another option institutional shareholders have. Despite current managers wanted to keep their jobs, they have no other resistance options than to limited the rise of the takeover costs for the buyer. When the buyer, despite these costs, takes over the company, he or she will replace the management.

Cremers and Nair (2005) also researched takeovers, which are together with market for corporate control the external governance mechanisms of a company. They examined shareholder activism’s, which is a company’s internal governance mechanisms, relationship with external governance mechanisms. They found that internal and external governance mechanisms are heavily dependent on each other. Both mechanisms are related to long-term abnormal returns and accounting measures of profitability.

In the past few years, shareholder activism by institutional investors has received more attention. Shareholder activism, also called relationship investing, is concerned with the worst performing company in the shareholder’s portfolio. Shareholders will pressure the management of these companies to improve their performance and, with that, their shareholders’ value. The increase in instutional shareholder activism is caused by the agency problems between shareholders and managers. Therefore, shareholders are motivated to participate actively in a company’s strategy direction. Gillan and Starks (2000) stated the reason why only large shareholders have the incentive to monitor or take over other activities to control a company is the free rider problem. All shareholders receive advantages from those activies even if they do not pay for them. The large shareholders pay for the activities since they can cover the costs with the increased profit. Furthermore, Gillan and Starks (2000) argued activism is focused on the long-term and helps management improve the performance for a long-term period. Opponents of institutional activism stated that pension fund managers do not have the expertises to consult management. They also stated that pension funds should focus on their main role, managing money for their clients (Gillan & Starks, 2000).

2_{Jensen (1993) and Roe (1990)}

(15)

14

Currently, shareholders are not only concerned about a company’s financial performance. Ryan and Buchholtz (2001) analysed a model in which they began with the shareholder’s perceptions of the investment. The amount of information shareholders gathered about the company depends on the search skills, interest level, acces to information, and willingness and ability to invest effort and time. Also, the shareholders trust and risk-taking determined their trading behaviour. Decision-making by individual investors was influenced by monetary return, qualitive issues, and macroeconomic expactations. Financial performance investors are also interested in the situation in which the profit is made. Unethical behaviour of a manager will effect a company’s performance, which is why investors are also interested in this kind of information. Furthermore, investors are interested in optimism or pessimism about issues, such as the whole economy (Ryan & Buchholtz, 2001).

Pound (1991) stated that institutional shareholders have incentives to make sure managers run the company efficiently because institutional shareholders have too many shares to just buy and sell the shares. Therefore, institutions cannot receive profit from buying and selling shares; instead, they try to increase the value by changing the management policy.

Results of Bethel and Liebeskind (1993) show that institutional ownership could lead to growth and an increase of investments: “This conclusion is consistent with the argument that institutional owners supported the 'managerial' goal of growth, rather than shareholders' goal of maximizing value” (Bethel & Liebeskind, 1993, p. 29). These findings are not consistent with previous studies, like Pound’s (1991). He stated that institutional shareholders and blockholders could affect managers pursuit of efficient corporate strategies. Bethel and Liebeskind (1993) explained this difference in findings by the fact that institutional shareholders are more aggressively monitoring the managers.

Gompers and Metrick (2001) stated in their paper that institutional shareholders have other requirements for the shares characteristics they like to buy than other investors. Institutional shareholders invest in shares that are larger, more liquid, and have had low returns during the previous year. Furthermore, an adjustment to the Securities and Exchange Act of 1934 in 1978 required institutions with more than $100 million of securities under discretionary management to report their shares to the SEC (Securities and Exchange Commission). This information is reported quarterly by a SEC 13F form. On this form, all common-stock positions of more than 10,000 shares or $200,000 must be reported. When an institution completes the 13F form, the institution is divided into one of the manager’s categories. The five categories are investment company, investment advisor, insurance company, bank, and other. Investment advisor contains large brokerage companies, and the category “other” includes university endowments and pension

(16)

15

funds. Another reason why institutional shareholders differ from individuals is that, except for the category “other”, all institutions are often agents for other investors. When individuals give institutions investment discretion, they can only monitor the institutions’ deficiencies. The incentives of institutions are often different from the incentives of individuals. Thus, when individuals hand over their rights to institutions, they lose their control. The agency costs that arise are costly for individuals, but individuals are willing to pay because of the economies of scale or other investment advantages when they join an institution (Gompers & Metrick, 2001).

Gaspar et al. (2005) researched how the investment horizon of a company’s institutional shareholders affects the market for corporate control. Their findings show that companies with short-term shareholders are more likely to acquire a takeover bid, but receive less premium. On the other hand, bidder companies with short-term shareholders receive worse abnormal returns when a merger is announced, as well as a long period of underachievement. These results show that companies with short-term shareholders have a worse negotiation position in takeovers. Worse monitoring from short-term shareholder can lead managers to keep reducing the leasehold or to negotiate for personal advantages at the price of the shareholders’ return. Clearly, mergers and takeovers are affected by agency problems between shareholders and managers. Monitoring activities can reduce these problems, but it depends on the shareholders’ cashflow-right and the shareholders’ incentives to monitor managers. How lesser short-term shareholders monitor the managers, how more discretionary power managers get at the bargain. Furthermore, managers might not only negotiate about the price for their shareholders, but also about their position after the merger, the new board structure, or executive compensation (Gaspar et al., 2005).

According to Gaspar et al. (2005), institutional ownership is important for a company and depends who its shareholders are. Short-term shareholders behave differently than long-term shareholders. Long-term shareholders will hold their position unchanged for a long period of time, whereas short-term shareholders will sell and buy their shares more often. Managers should make a trade-off between short-term shareholders, who are not devoted to the company, and long-term shareholders, who can give them a stronger bargaining position (Gaspar et al., 2005).

Furthermore, considerable research has been conducted about the relationship between compensation and institutional shareholders. Almazan et al. (2005) found that the pay-for-performance sensitivity of managerial compensation increases in the concentration of active institutional shareholders, but the authors found no significant relationship with passive institutional shareholders. They explained this result by the fact that passive institutional shareholders face higher cost for monitoring than active institutional shareholders. On the other

(17)

16

hand, the executive compensation decreases in the concentration of both active and passive institutional shareholdership.

Also, Hartzell and Starks (2003) found a positive relationship between pay-for-performance sensitivity of managerial compensation and instititutional ownership. On the other hand, they also found that institutional investor concentration is negatively related to the level of executive compensation. Their results suggest the same as previous studies: institutional investors have a monitoring position. Together with incentive compensation, the monitoring activities of institutional investors can decreases agency problems between managers and shareholders (Hartzell & Starks, 2003). Institutional shareholders have access to a company’s performance and can monitor management, which is why the proportion of institutional shareholders could have an effect on agency costs and, indirectly, also on dividend payments (Han, Lee, & Suk, 1999). The results of the research of Han et al. (1999) show that institutional shareholders would rather have dividend income than capital gains because of the tax exemption of a part of this dividend.

More research about the relationship between institutional shareholders and taxes was conducted. In their research, Khurana and Moser (2012) found that companies with more long-term institutional shareholders showed significant less tax avoidance. The companies that showed these results had weak corporate goverance, following the theory that long-term shareholders are used as another form of corporate goverance, which causes tax avoidance and decreases the potential for managerial opportunism.

2.3 Hypothesis development

Managers’ behaviour will determine the degree of cost stickiness, and managers’ incentives will affect that behaviour. Managers’ incentives should trigger managers to behave to maximize shareholders’ value. Managers behave differently based on their incentives, but shareholders have different incentives. Shareholders can be short- and long-term orientated. As explained in the literature review, institutional shareholders could be more committed to a company than individual shareholders. In this research, I assume that institutional shareholders are more long-term orientated than individual shareholders. I expect that if a company has more institutional shareholders, there will be more cost stickiness than when shareholders are merely interested in a company’s short-term performance because, when sales decrease and future demand is uncertain, companies with long-term orientated shareholders are not only interested in the current firm value, but also in the future value. Therefore, the costs may not directly be reduced, but it will be carefully determined whether this sales decrease is likely to be permanent. Institutional shareholders could influence management to delay cutting expenses with the power they have. They could use this

(18)

17

power to monitor management’s behaviour. On the other hand, if there are more short-term orientated shareholders, they do not have the power institutional shareholders have and they are more interested in the current firm value. Therefore, a sales decrease will lead more quickly to adjustments of costs, which leads to lower cost stickiness. The hypothesis of this paper is, therefore, as follows:

(19)

18

3. Methodology

3.1 Sample selection

Compustat was used to gather the sample of North America publicly traded companies. This database offers, among other things, data about the sales, expenses, and assets of companies. In addition to Compustat, data about the shareholders of the companies is required for this research. Based on the CUSIP numbers of the companies, these two datasets were merged. Before the CUSIP numbers could be used, the CUSIP numbers from Compustat had to be converted to a length of eight, instead of nine, digits.

Most prior studies focus on data from before the financial crisis, so until 2007. With this research, I also want to include the data during the financial crisis, so starting with the most recent data and taking a period of 20 years. Hence from 1996 up to and including 2016, but after cleaning the data, the sample consisted the years 2002 to 2014. According to Anderson et al. (2003), to examine cost stickiness, the data researched should consist of a long period; thus, the initial sample consisted of data from the last 20 years. Anderson et al. (2003) did not define how many years a “long period” was. Since the final sample consisted of more than ten years, I assumed this period was long enough. Also, Calleja et al. (2006) explained in their paper that a longer time period gives managers more certainty about the cause of the reduction of sales and the period the sales will be lower. This will result in less quick cost adjustments when the sales drop.

Before cleaning the Compustat dataset, this dataset consisted of 394,163 observations, after dropping missing and invalid observations, the dataset consisted of 90,366 observations. The dataset of institutional shareholders did not need much cleaning: it went from 97,188 to 97,167 observations. After merging the datasets, 50,623 observations were matched. Furthermore, new variables had to be calculate, the outliers removed using winsorizing, and the data cleaned again, which removed missing and duplicate items. After these steps, the dataset consisted of 30,570 observations. When all the required new variables were calculated, the final sample consisted of 10,347 observations. Following prior research, financial institutions (SIC codes 6000–6999) were excluded from the sample since these companies will react differently than other for-profit companies. The final sample of 10,347 firm-year observations between 2002 and 2014 was used to conduct the research.

3.2 Measurement of variables

3.2.1 Cost stickiness

This research used Weiss’s (2010) direct measure for cost stickiness at the company level. Cost stickiness was tested on company level because the shareholders’ incentives also are based on

(20)

19

company level. In this equation, cost stickiness was measured by SG&A costs and net sales revenue. SG&A costs were used based on the papers of Anderson et al. (2003) and Chen et al. (2012). Furthermore, Jiang et al. (2006) explained that SG&A costs are mostly driven by sales activities and represent a company’s indirect costs. Therefore, in this research, SG&A costs were used instead of operating costs.

The equation used an estimation of the difference between the ratio of decreasing cost and sales and the ratio of increasing cost and sales, both for the recent quarters:

𝑆𝑇𝐼𝐶𝐾𝑌_𝑖,𝑡 = 𝑙og (∆𝐶𝑂𝑆𝑇

∆𝑆𝐴𝐿𝐸)_𝑖,𝜏− 𝑙𝑜𝑔 (

∆𝐶𝑂𝑆𝑇

∆𝑆𝐴𝐿𝐸)_𝑖,𝜏𝜏, 𝜏 ∈ {𝑡, . . , 𝑡 − 3}

Equation 1: Cost stickiness

In this equation, STICKY is defined as follows: in one quarter the sales decrease and in another quarter the sales increase. With the equation, the difference in the slope between the two most recent quarters from quarter t-3 through quarter 1 was examined. Instead of using quarterly data for Weiss’s (2010) equation, in this research, yearly data was used. Yearly data was used because STICKY was otherwise the only variable that was quarter based. To build a proper model, the data all had to be year-based.

The “i” means that cost stickiness is measured at company level. Furthermore, the “t” is used for time in years. On the other hand, “τ” indicates the most recent year, where “τ” indicates a sales decrease and “τ” a sales increase. Δ COST it was calculated by, (SALE it – EARNINGS it) –

(SALE i, t-1 – EARNINGS i, t-1) and Δ SALE it = SALE it – SALE i, t -1. According to prior literature,

change in sales was used since change in activities is hard to measure.

If the equation results in a negative value, the costs are sticky, which means costs increase more when activities increase than costs decrease when the activities decrease by an equal amount. The smaller the value of STICKY, the stickier the costs. When costs are sticky, managers’ reduce costs less when sales decrease than they increase the cost when sales increase (Weiss, 2010).

3.2.2 Institutional shareholders

Several articles3_{use the data from the CDA/Spectrum to measure institutional}

shareholders. This dataset consisted of the 13F forms completed by required institutions. However I did not have access to this database. Therefore, I used other data to calculate institutional ownership to measure institutional shareholders. The data for institutional shareholders consisted of the proportion of shares held by institutional investors per company per year (= shares held by

(21)

20

institutional investors / all shares outstanding from a given company in each year). An institutional investor is defined as any shareholder who owns more than 5% of a companies’ stock in a year.

To measure the effect of institutional shareholders on cost stickiness, first the fraction of institutional shareholder was used (FRACTION INST). The fraction is the level of ownership; it is the proportion of the company’s shares held by institutional shareholders (Gaspar et al., 2005). The other independent variable measuring institutional ownership was a dummy variable for institutional shareholders, where 1 indicated institutional shareholders and 0 short-term shareholders (DUMMYINS). The cut off that determined whether a company had more institutional shareholders than shareholders merely interested in the short-term company performance was 25%. This means that for each observation when the variable FRACTIONINS was above 25%, the dummy variable was a 1, and otherwise it will be 0. The chosen cut-off was 25% because it seemed a reasonable percentage to divide between long- and short-term shareholders to groups.

3.2.3 Control variables

This research also included control variables to control for differences in timing, market conditions, and company-specific characteristics. Bugeja et al. (2015) found in their research that the degree of cost stickiness will increase with a company’s employee and asset intensity (ASSET INS and EMPLOYEE INS). This means companies that have more employees and use more assets have more sticky costs. Another control variable Bugeja et al. (2015) mentioned is a dummy variable for success (DUMMYSUCC). This variable controls for a situation in which a company has a decrease of sales in the previous and current periods. Managers will react differently in this situation. They will reduce fewer resources in the current period, which will cause more cost stickiness.

As mentioned, the research was performed with a sample from the years 2002 through 2014. The financial crisis started during this period, and the research of Campello et al. (2010) shows the financial crisis could affect cost stickiness. Therefore, the control variable YEAR controlled for these possible effects. YEAR not only controlled for the effects of the financial crisis, but also for other time differences. In the regression models, the fixed effects of YEAR were controlled for. According to Bugeja et al. (2015), the degree of cost stickiness also depends on the industry. The control variable INDUSTRY was used to control for this effect. As explained, financial institutions (SIC codes 6000–6999) were excluded from the sample. INDUSTRY was also used as a fixed effect in the regression analysis. Before industry codes were used in the regression models, the SIC codes were converted to a two-digit SIC code. Another control variable used in

(22)

21

the model was firm size, which was measured by total sales. Since the total sales varied extremely in the sample, the natural logarithm (LOG) of total sales was used (FIRM SIZE).

Furthermore, the model controlled for leverage, based on a ratio of long-term debt to total assets (LEVERAGE). In this way, leverage indicated how much debt was used to finance the assets (Chen et al., 2012; Chen, Chen, Cheng, & Shevlin, 2010). Market-to-Book ratio (MTB) controlled for a company’s growth opportunities (Gunny, 2010). Growth opportunities were controlled for since growing companies may make more investments in assets and this generates timing differences in the recognition of expenses of these investments (Chen et al., 2010). The final control variable was stock performance, which was measured by stock returns in the year before the annual board meeting (STOCKPER). The stock performance compared a company’s performance for the two following years (Chen et al., 2012).

3.3 Empirical model

The empirical model tested the relationship between cost stickiness and shareholders’ incentives. I expected companies that had more institutional shareholders would have more cost stickiness than companies that had more shareholders interested in the company’s short-term performance. The hypothesis tested with these models, therefore, is as follows: the fraction of institutional shareholders is positively related to cost stickiness. The models include the variables for cost stickiness, institutional shareholders, and the control. The models to test the hypothesis are the following:

Model 1: 𝑆𝑇𝐼𝐶𝐾𝑌_𝑖,𝑡

= β0 + β1 FRACTIONINST + β2 YEAR + β3 FIRM SIZE + β4 ASSETINS + β5 EMPLOYEEINS + β6 DUMMYSUCC + β7 INDUSTRY + β8 LEVERAGE + β9 MTB + β10 STOCKPERε i t

Equation 2: Model 1

Model 1 uses the fraction of institutional shareholders, while model 2 uses a dummy variable of institutional shareholders.

Model 2: 𝑆𝑇𝐼𝐶𝐾𝑌_𝑖,𝑡

= β0 + β1 DUMMYINS + β2 YEAR + β3 FIRM SIZE + β4 ASSETINS

+ β5 EMPLOYEEINS + β6 DUMMYSUCC + β7 INDUSTRY + β8 LEVERAGE + β9 MTB + β10 STOCKPERε i t

(23)

22

In Table 1, an explanation of each variable used in the models is given.

Table 1: Explanation variables of the models

Variables Explanation

STICKY Calculated per firm quarter in Weiss’s (2010) equation. As my unit of analysis is the firm year, I used year-based figures.

FRACTIONINST Proportion of shares held by institutional investors per company per year (shares held by institutional investors / outstanding shares).

DUMMYINS 1 is institutional shareholder and 0 otherwise. YEAR Fiscal year.

FIRM SIZE Total sales, Log (total sales).

ASSETINT Ratio of assets to sales (total assets / total sales).

EMPLOYEEINT Ratio of the number of employees to sales (number of employees / total sales). DUMMYSUCC 1 if revenue in t-1 is lower than revenue in t-2, and 0 otherwise.

INDUSTRY Two-digit SIC codes of the companies.

LEVERAGE Ratio of long-term debt to total assets (long-term debt / total assets). MTB Controls for growth opportunities of the companies

(Market value of equity / (book value per share * shares issued)). STOCKPER Stock returns in the year before the annual board meeting

(24)

23

4. Results

4.1. Descriptive Statistics

Table 2 shows the descriptive statistics of cost stickiness, institutional shareholders, and control variables used in the models.

Table 2 : Descriptive Statistics

N St. Dev. 25% Mean Median 75% Min Max

STICKY 10,347 1.086 -0.429 0.037 0.042 0.511 -8.122 7.216 FRACTIONINST 10,347 0.251 0.285 0.480 0.418 0.658 0.013 1 DUMMYINS 10,347 0.388 1 0.816 1 1 0 1 FIRMSIZE 10,347 1.815 5.653 6.859 6.892 8.029 0.648 13.089 ASSETINT 10,347 0.972 0.646 1.191 0.948 1.394 0.081 23.911 EMPLOYEEINT 10,347 5.795 2.540 5.462 4.076 6.197 0.052 118.415 DUMMYSUCC 10,347 0.444 0 0.271 0 1 0 1 LEVERAGE 10,347 0.230 0.086 0.228 0.192 0.311 0 6.309 MTB 10,347 44.677 1.152 2.092 1.867 2.986 -4021.212 1147.528 STOCKPER 10,347 0.748 0.846 1.174 1.088 1.351 0.013 29.095

For all variables missing observations and values smaller than zero were dropped. Furthermore, duplicates were dropped and outliers beyond the 1st and 99th percentiles were removed using winsorizing. For the change variables used to calculate STICKY, observations where SG&A costs were larger than sales were dropped.

STICKY FRACTIONINST DUMMYINS YEAR FIRM SIZE ASSETINT EMPLOYEEINT DUMMYSUCC INDUSTRY LEVERAGE MTB STOCKPER Year-based figures

Shares held by institutional investors / outstanding shares 1 is institutional shareholder and 0 otherwise

Fiscal year Log (total sales) Total assets / total sales

Number of employees / total sales

1 if revenue in t-1 is lower than revenue in t-2, and 0 otherwise Two-digit SIC codes of the companies

Long-term debt / total assets

Market value of equity / (book value per share * shares issued) Stock price t / stock price t-1

The mean of STICKY is slightly positive (0.037), which is not consistent with Weiss’s (2010) prior research since the value of Weiss’s (2010) equation gives a negative outcome when the costs are sticky. In this research, the mean of STICKY indicates that the SG&A costs are not sticky, but rather the opposite, called anti-sticky. Since it is only a slightly positive value and the minimum

(25)

24

is negative, it could be that there are some companies where heavy anti-stickiness occurs and companies where cost stickiness occurs. Alternatively, the value is consistent with the research of Calleja et al. (2006), which found that the level of cost stickiness is lower in the United States and the United Kingdom than in Germany and France. They suggested this difference was caused by the fact that companies in the United States and the United Kingdom are more focused on maximization of shareholders’ value and are under more scrutiny. Their research was based on operating costs instead of SG&A costs. It also should be mentioned that their research was conducted with shareholders in general and not a comparison of institutional shareholders and shareholders who are merely interested in a company’s short-term performance. Therefore, I assume their findings only hold for short-term shareholders.

The positive value of STICKY could also be caused by Weiss’s (2010) equation, used in this research. That the costs are not sticky could be because, in this research, yearly figures were used instead of quarterly, which Weiss (2010) used in his research. Other factors that could cause the difference in STICKY are the difference between Weiss’s (2010) sample and this research’s sample. Weiss’s (2010) sample consisted of industrial firms (SIC codes 2000-3999) from 1986 to 2005. Therefore, both the kinds of companies and the years included in the samples are different since this research’s sample consists of all companies, except for financial institutions (SIC codes 6000-6999) and the research is conducted with observations between the years 2002 and 2014. Both these effects were controlled for in the regression models (YEAR and INDUSTRY).

Regarding the descriptive statistics, FIRMSIZE shows less difference among the companies since the log of total sales was used, instead of the absolute values of total sales. For the control variables, ASSETINT, EMPLOYEEINT, and DUMMYSUCC, the mean, median, and standard deviation are in line with the research of Chen et al. (2012). The descriptive statistics of EMPLOYEEINT and DUMMYSUCC are also similar to the research of Bugeja et al. (2015). However, STOCKPER is not consistent with the research of Chen et al. (2012), who have a much lower standard deviation, mean, and median. ASSTETINT and EMPLOYEEINT show that, on average, the companies in the sample used $1.191 million (median: 0.948) of assets and 5.462 (median: 4.076) employees to make each sales revenue in millions of dollars. The mean and median of DUMMYSUCC are, respectively, 0.271 and 0. Most companies in the sample had, in the past two years, not had two consecutive years of sales decreases. Furthermore, the mean and median of LEVERAGE are similar to the values in the research of Chen et al. (2010).

(26)

25

4.1.1 Correlation

In Table 3, the correlation matrix of the variables used in the model is given. For each correlation, stars indicate whether the correlation is significant. In this research, I chose to show the significance on three levels. One star indicates a significance level of 0.01, where the p-value is below 0.01. For two stars, the p-value is below 0.05, and three stars gives a significance level of 0.10.

Before the regression analyses were conducted, all the independent variables in the models were also checked for multicollinearity. There were no multicollinearity issues because all the values of the VIF-test are below five. Furthermore, the Tolerance levels (1/VIF) are all above 0.2. The dependent variable, STICKY, also seems to be normally distributed.

Table 3: Correlation matrix

STICKY FRACTION INST DUMMY INS FIRM SIZE ASSET INT EMPLOY INT DUMMY SUCC LEVER-AGE MTB STOCK PER STICKY 1 FRACTIONINST 0.0019 1 DUMMYINS -0.0162 0.5673*** 1 FIRMSIZE 0.0089 -0.7248*** -0.3781*** 1 ASSETINT -0.0310*** -0.1121*** -0.0399*** -0.0391*** 1 EMPLOYEEINT -0.0296*** 0.1628*** 0.0785*** -0.1723*** -0.1094*** 1 DUMMYSUCC 0.0179* 0.1844*** 0.1121*** -0.1280*** 0.0117 0.0214** 1 LEVERAGE -0.0028 0.029** 0.0520*** 0.0379*** 0.1615*** 0.0146 0.0405*** 1 MTB -0.0055 -0.0168* -0.0253** 0.0146 0.0073 -0.0118 -0.0186* -0.0177* 1 STOCKPER 0.0661*** 0.0531*** 0.0337*** -0.0461*** -0.0130 0.0103 0.0784*** 0.0189* -0.0171* 1

Significance (two-tailed) at 0.10 level (*), 0.05 level (**), and 0.01 level (***).

(27)

26

STICKY is positively correlated to FRACTIONINST, which was expected since it was expected that the fraction of institutional shareholders was positively related to cost stickiness. The correlation of 0.0019 is not significant with a p-value of 0.8498. Additionally, the relationship between STICKY and DUMMYINS is negative (-0.0162), but with a p-value of 0.1003, it is also not significant. Furthermore, STICKY is significantly correlated to ASSETINT, EMPLOYEEINT, DUMMYSUCC, and STOCKPER.

Many correlations are significant, especially FRACTIONINST and DUMMYINS have significant correlations. Since there are no high correlations between two independent variables, it can be confirmed that there are no multicollinearity issues. A negative correlation between FRACTIONINS and FIRMSIZE is, however, unexpected. It was expected that when a company grows, the sales grow and the fraction of institutional shareholders, or at least the number of shareholders, also increases. This correlation suggests that when a company grows, the fraction of institutional shareholders will move in the opposite direction, and so decrease. An explanation for this could be that shareholders interested in a company’s short-term performance are also interested in a company when it grows. For FIRMSIZE, the explanation for almost all significant negative relationship could be that when a company grows, the proportion of the other independent variables will change instead of the real amounts of ASSETINT, EMPLOYEEINT, DUMMYSUCC, and STOCKPER.

4.2 Multivariate Analysis

Both models were tested in a regression model. In the regression models, in addition to the control variables, the fixed effects of year and industry were controlled for. The industry SIC codes were converted to two-digit SIC codes before they were used in the regressions. The regression models were checked for robustness, performing regression analyses with the robust function in STATA. Again, the stars indicate the significance levels of the results. With one star, the p-value is below 0.1, two stars indicate a p-value below 0.05, and with three stars, the p-value is below 0.01.

In Table 4 the regression model for model 1 is shown; the test was based on the fraction of institutional shareholders.

(28)

27

Table 4: Regression Model 1- Cost stickiness and institutional shareholders (based on FRACTIONINST)

STICKY Coef. t FRACTIONINST 0.0783 1.20 FIRMSIZE 0.0160 * 1.72 ASSETINT -0.0439 ** -1.98 EMPLOYEEINT -0.0080 *** -3.32 DUMMYSUCC 0.0260 0.91 LEVERAGE 0.0121 0.20 MTB -0.0001 -0.48 STOCKPER 0.1315 *** 5.42 Constant 0.0877 0.37

OLS regression with robust standard errors. Significance (two-tailed) at 0.10 level (*), 0.05 level (**), and 0.01 level (***). STICKY FRACTIONINST DUMMYINS YEAR FIRM SIZE ASSETINT EMPLOYEEINT DUMMYSUCC INDUSTRY LEVERAGE MTB STOCKPER Year-based figures

The p-value of the F-test in Table 4, above, is significant. However, only 3.41% of the variance of cost stickiness is explained by the independent variables. The coefficient of FRACTIONINST is 0.0783, which indicates that with each increase of FRACTIONINST, STICKY will increase by 0.0783. The result is in line with the expected result: that FRACTIONINST and STICKY were positively related. It was expected that companies that have more institutional shareholders will have more sticky costs than companies with more shareholders merely interested in the short-term performance of the company. However, this result is not significant since the p-value is 0.232, which is not below the p-values of 0.10, 0.05, and 0.01.

Furthermore, the coefficient of DUMMYSUCC is 0.0260, which indicates managers would be less motivated to reduce current resources when the sales increased in the year before. This caused more asymmetric cost behaviour. The result, with a p-value of 0.365, is not significant, so this conclusion cannot be used. On the other hand, ASSETINT and EMPLOYEEINT are both significant with p-values of 0.047and 0.001, respectively. Their coefficients are negative

(29)

28

(respectively, -0.0439 and -0.0080), which indicates that companies that have more assets and employees to make a certain revenue will have stickier costs since they should adjust the costs more. These results are in line with the expectations and prior literature.

Also, the result of FIRMSIZE is significant with a p-value of 0.085. The coefficient of 0.0160 indicates FIRMSIZE and STICKY will change in the same direction, which is reasonable since if a company grows, there will be more cost to adjust when the sales decrease. Also, the assets and employees could increase, which is in line with the variables ASSETINT and EMPLOYEEINT.

However, a significant positive coefficient of STOCKPER (0.1351) is not in line with the research of Chen et al. (2012), which, with a negative coefficient, suggests the extent of cost stickiness is higher in companies with a strong stock performance. In this research, the results suggest that cost stickiness is lower in companies with a strong stock performance.

On the next page, in Table 5, the regression model for model 2 is given, which tests the relationship between STICKY and a dummy variable of institutional shareholders.

(30)

29

Table 5: Regression Model 2 - Cost stickiness and institutional shareholders (based on DUMMYINS)

STICKY Coef. t DUMMYINS -0.0292 -1.11 FIRMSIZE 0.0053 0.71 ASSETINT -0.0475 ** -2.16 EMPLOYEEINT -0.0078 * -3.24 DUMMYSUCC 0.0317 1.11 LEVERAGE 0.0185 0.31 MTB -0.0001 -0.52 STOCKPER 0.1319 *** 5.44 Constant _0.2245 0.97

OLS regression with robust standard errors. Significance (two tailed) at 0.10 level (*), 0.05 level (**), and 0.01 level (***).

Again, the p-value of the F-test is significant. The R2_{is equal to the R}2_{of model 1 (0.0341)}

since only one variable is changed between the two models. The most striking difference between the regressions of the two models is that the relationship between STICKY and DUMMYINS is negative (-0.0292). This is not consistent with the expected positive relationship between this two variables. However, with a p-value of 0.265, this result is not significant. Since the result is not significant, it is not reliable, but the coefficient of the dummy variable for institutional shareholders has another meaning than FRACTIONINS. Since DUMMYINS is coded 0 or 1 (0 = short, 1 = long), the interpretation is not that with each increase of DUMMYINS, STICKY will decrease by

0.0292. Instead, for long-term shareholders, the predicted cost stickiness score would be decrease by 0.0292 more than for short-term shareholders.

(31)

30

Furthermore, ASSETINT, EMPLOYEEINT, and STOCKPER have significant results again. In this regression model, as opposed to the regression of model 1, only FIRMSIZE is not significant. The coefficient of the variables also has the same meaning as in model 1 since ASSETINT and EMPLOYEEINT are negative, and FIRMSIZE, DUMMYSUCC, and STOCKPER are positive again.

In summary, the results of the regression models for both models show that institutional shareholders are not positively related to cost stickiness, which was expected in H1 (the fraction of institutional shareholders is positively related to cost stickiness). The results of model 1 show a positive relationship between STICKY and FRACTIONINS, but this result is not significant. For model 2, the result is also not significant, but this relationship was also not what was expected since the results show a negative relationship between STICKY and DUMMYINS. Therefore, the results of both models are not in line with the hypothesis of this research, and H1 is rejected.

The effect of institutional shareholders on cost stickiness