The effect of family firms on cost stickiness
Name: Jorrit Weidema Student number:11416815
Thesis supervisor: Dr. A. Sikalidis Date: June 15th , 2018
Word count: 13,666
MSc Accountancy & Control, specialization Control
Statement of Originality
This document is written by student Jorrit Weidema 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.
Abstract
The purpose of this thesis is to examine the relationship between family ownership and cost stickiness. It aims to figure out whether firms controlled and founded by families show different cost behavior in contrast to non-family firms. Prior academic research has shown that family firms have different characteristics, that on the one hand reduce type I agency problems, while on the other hand type II agency problems are increasing. Furthermore, cost stickiness depends on managerial decision making, whether costs need to be cut, or resources need to be adjusted. Based on this differences between family firms and non-family firms, and the determinants of cost stickiness, I will argue whether family firms do affect cost stickiness. The findings of this thesis, suggest that there is no significant positive association between family firms and cost stickiness. Even with the superior voting rights, proved by the presence of dual-share class structures, there is no association, that such structures affect cost stickiness within family firms. Since there is not much literature on the relationship between cost stickiness and family firms, I suggest future research could be very valuable to the cost accounting and family-firm literature and how this relationship affects management, ownership and other shareholders.
Contents
1. Introduction ... 1
2. Literature review ... 3
2.1 Cost stickiness ... 3
2.2 Family ownership ... 5
2.3 Dual class shares ... 8
2.4 Hypothesis development ... 9
3. Data sample and empirical model ... 11
3.1 Cost stickiness Model: ... 11
3.2 Family ownership variables ... 12
3.3 Control variables ... 12 3.4 Sample selection: ... 13 3.5 Empirical models ... 14 4. Empirical findings ... 16 4.1 Descriptive statistics ... 16 4.2 Correlation matrix: ... 18 4.3 Regression analysis: ... 20
4.5 Regression analysis dual share structure ... 27
4.6 Robustness check ... 30
5. Conclusion: ... 33
1. Introduction
Reaching earnings targets and making profit are some of the most important fundamentals for most organizations in the day-to-day job. Besides revenues and profits, cost management is just as important for organizations. In the accounting literature cost accounting has a significant role. Of course, managers must be able to perform analyses regarding profitability, hence the attention for cost information. Mostly, traditional models of cost accounting assume a linear change of, mostly variable, costs compared to activity. In these symmetrical models, costs change in a linear fashion based on output. Management uses cost information in their decision making processes. Does the level of resources need to be adjusted, and what are the consequences of these changes to on the short and long -term?
However, academic literature in cost accounting has provided a lot of evidence that costs do not always behave symmetrical. According to Anderson et al., (2003), costs appear to be sticky. Which means that if costs rise due to increased sales activity, it is relatively higher compared to sales decrease. In other words, costs appear to decrease less with a decrease in sales activity, compared to when cost increase, if sales activity increases. Costs behave asymmetrical. One of the possible explanations if the level of adjustments caused by managers. They face decisions to adjust the organization in optimal fashion to changes in sales activity. Whether it is production level or cost structure (Wiersma, 2004). It is likely that managers find it a more positive process if sales activity increase instead of a decrease in sales activity. It is easier to hire more employees, then laying off employees and cutting resources.
Because academic literature has stated managerial actions in many cases as driver of cost stickiness, it is interesting to examine to what extend family firms differ in their behavior and if cost stickiness differs between family and non-family firms. Moreover, because, according to Anderson and Reeb, (2003) family firms are likely to have higher firm performance compared to non-family firms. Most of the studies conducted towards family firms name agency problems as distinguishing factor in firm performance. Family firms tend to be able to, on the one hand, reduce type 1 agency problem. According to the theory of socioemotional wealth, family firms tend to put more emphasis on success on long-term success. Families tend to see their organization as an asset of their own, and will do everything to preserve the good family name. Non-family firms are more likely to focus mostly on financial goals, while family firms obtain a more balanced focus, which also included non-financial goals (Ali, Chen & Radhakrishnan, 2007). Therefore, family firms show more risk-aversion, compared to non-family firms. However, on the other hand family firms are subject to increasing type 2 agency problems, due to abuse of voting rights, or power in organizations management. Benefits on behalf of the major shareholders, often come at expense of minority shareholders (Shleifer & Vishny, 1986). Since family firms often hold large amounts of shares, it is more likely that they would make corporate decisions which may not be beneficial for the entire organization, but they are done anyway, since the risk mostly lies with minority shareholders.
Due to the differences that have been pointed out in academic literature I will examine to what extent cost stickiness differs from family firms compared to non-family firms. This led to the following research question:
“Does family firm ownership affects cost stickiness, compared to non-family firms?”.
To my knowledge, there is no literature present that has established such a difference between family and non-family firms on the matter of cost stickiness. There has been established a positive association between managerial incentives and cost stickiness (Kama & Weiss, 2013) and between the agency problems and cost stickiness (Chen, Lu & Sougiannis, 2012). However, to my knowing there are no similar researches conducted that has determined whether there are differences for cost stickiness in combination with family firms. Therefore, I hope this thesis contributes to existing cost accounting literature and family firm literature. I will try to provide new insights on the existing agency theory, about managerial incentives and the level of cost stickiness which is influenced by expected differences.
This thesis is structured in the following manner. Section 2 provides the literature review with all theories and institutional background that are relevant for this thesis, followed by the hypothesis development and analytical framework. Section 3 includes the data sample and empirical models that are going to be used. Next, in section 4 the empirical findings will be clarified. Lastly, section 5 includes the conclusions on this thesis.
2. Literature review
2.1 Cost stickiness
This thesis relates to theory regarding cost stickiness. Cost stickiness is defined by Anderson, Banker and Janakiraman (2003) as: “Costs are sticky if the magnitude of the increase in costs associated with
an increase in volume is greater than the magnitude of the decrease in costs associated with an equivalent decrease in volume”. Cost stickiness is therefore known as asymmetrical cost behavior
within organizations.
The Anderson et al. (2003) research measures cost stickiness through the selling, general and administrative costs (SG&A costs). SG&A costs are considered a logical manner of measuring because of their relationship with activity and revenues. Sales activity- and volume is the main driver of several components of SG&A costs. Anderson et al. (2003) addressed cost stickiness by setting the following example. Following their research cost stickiness occurs when, for example, sales activity and revenues increase with 1%, SG&A costs increase with 0.5%. However, when sales activity and revenues decrease with 1%, SG&A costs ‘only’ decrease with 0.35%. This reaction is addressed as cost stickiness; it seems that 0.15% of SG&A costs does not follow the traditional symmetric cost behavior and therefore remains sticking in the organization.
Calleja, Steliaros and Thomas et al. (2006) proved the presence of cost stickiness by using operating costs instead of SG&A costs. Their study in European countries showed the same asymmetrical behavior in operating costs as in SG&A costs. An average in- and decrease in sales activity and revenues of 1% leads to an in- and decrease of operating costs of respectively 0.97% and 0.91%. However, they found that the companies in their sample share characteristics that show that the degree of cost stickiness decreases over a longer period of time. The ability of its management to gain more information over a longer period of time, gives them the possibility to adjust resources better over time, and to control more for sales activity decrease. However, adjusting resources remains is a trade-off for management, between maintaining excessive resources or obtaining costs for adjusting. Do managers feel the need for deposing resources, because sales activity levels are decreasing on the short term, or do they expect future growth and is disposal perhaps not beneficial on the long term.
Thus, cost stickiness arises when managers are subject to responding to largely demand changes (Subramanian & Weidenmier, 2003). Their research examined if the change in sales activity actually result in resource adjustments and if these factors are considered the primarily factor for cost stickiness. It turns out that costs are not considered sticky, if the sales activity change is considered small. In their research, a small change is any change under 10%. Changes above 10% in sales activity require immediate investments in additional resources which leads to increasing costs. However, following a sales activity decrease of 10% or more, managers lack in making decisions towards resources, which causes cost stickiness to arise. Managers are not able to reduce or dispose resources on such a
short-term. Besides the managerial determinants in their research, they also investigated organizational determinants. They have compared cost stickiness across several industries. It turns out that different industries, tend to have different asymmetrical behavior. The manufacturing industry suffers most from cost stickiness. The authors explain that this is caused by the high amount of fixed assets, mostly manufacturing machines, and their high levels of inventory. Decisions made by managers regarding resources are therefore mostly on the long term and difficult to reduce on short-term notice. The merchandise industry tends to suffer least from cost stickiness. Because of their competitive market, managers tend to adapt quickly to sales activity.
Banker, Byzalov and Plehn-Dujowich (2011) tried to reduce the gap between decision making and industry determinants. Their research showed that the degree of SG&A cost shows differences between industries whom are fast growing and those which are shrinking. In fast growing industries managers tend to be optimistic about their future sales activity. This will cause cost to be particularly sticky, since managers are optimistic about their chances in the future. On the other hand, shrinking industries will cause managers to be more pessimistic, instead of optimistic. Therefore, cost stickiness will decrease, or even not exist. This might be a controversial argument. Cannon (2014) investigated sticky costs in the United States Air Transportation Industry and found that shrinking industries might not cause the disappearance of cost stickiness. He states that decreasing demands will incentivize managers to lower the selling price rather than reducing resources. Adjusting the selling price will also be asymmetrical to the decreasing capacity. Therefore, resources are not disposed or reduced and cost stickiness will still be present in companies when industries shrink and demand decreases. So, decision making also depends on managerial optimism.
Following the Calletja et al. (2006) research, firm characteristics are considered important in determining cost stickiness. Besides adjusting resources, different corporate governances and managerial oversight seem to affect cost stickiness in their study. The gap between the Anderson et al. (2003) and Calleja et al. (2006) research is investigated by Chen et al., (2012). Their research shifts the focus from economic determinants towards agency characteristics that may drive the asymmetric behavior of SG&A costs. Hereby considering the characteristics of corporate governance and managerial oversight. Making sure that managers use discretion on behalf of the shareholders and not only for their own personal wealth. They based this on the empire building theory. Manager using empire building tend to increase the size and scope of their individual power and influence with in organizations. Mostly with negative consequences for companies. Companies may expand above their maximum capacity or leave resources unused, if they are not beneficial for management. This could be seen as aggressive decision making to enable excessive growth. Growth in such a way, that is destroying future firm value on the short term and affects firm performance in long term vision (Hope & Thomas, 2008).
Chen et al. (2012) research reveals the relation within agency factors and cost stickiness. Companies with a weak corporate governance show great possibility to empire building. A weak
corporate governance implies less managerial oversight and provides the opportunity for managers to exercise their discretion in a way, more beneficial for themselves. The academic literature provides plenty of determinants that lead to cost stickiness, however, there is less mentioned about the actual consequences of cost stickiness. Malik (2012) provides consequences of cost stickiness, which will be further used in this thesis.
First of all, studies of Dierynck, Landsman and Renders (2012) emphasizes an association between cost stickiness and earnings managers. Especially because of the already mentioned conflicts between self-interested managers and other stakeholders in the organizations. In their study, they primarily focus on labor costs and earnings management incentives for management. Mostly it focuses on managers whom are being rewarded on zero earnings benchmarks. Thus, their study focuses more on small-profit firms. Their results show that managers, during periods of sales increase, increase labor costs in a smaller matter, than those compared to other companies. Furthermore, if sales decrease during a certain period, managers tend to decrease on a greater extent. Dierynck et al. (2012) describe this as opportunistic behavior, mostly to avoid reporting losses. In essence, their study shows that, mostly small profit companies show more symmetric cost behavior, and therefore less cost stickiness, in order to avoid reporting losses.
Secondly, Malik (2012) states, that the presence of cost stickiness leads to inaccurate earnings forecasts. It turns out, that companies with highly asymmetric cost behavior, show less accurate analysts’ earnings forecasts, compared with firms with more symmetric cost behavior. According to Weiss (2010), this mostly depends on management’s behavior. If companies experience more asymmetric cost behavior, the cost adjustments will not have much effect on actual cost savings if sales activity decreases. Therefore, analysts’ ability of predicting earnings precisely will reduce. Moreover, his findings also imply that the markets response to surprise earnings will be significantly reduced if a firm shows asymmetric cost behavior. Thus, the market response is considered reduced, when firms show cost stickiness. Analysts and investors find difficulties forming beliefs regarding firm value.
To summarize the theory on cost stickiness. It seems that the existence of cost stickiness is proven by a link between increasing (decreasing) SG&A costs and increasing (decreasing) sales activities. This might depend on different organizational and industrial characteristics. Therefore, cost stickiness depends on a lot of different determinants. Managers use a certain decision making to adjust costs (or not), which depends on their own capacity, forecasts and optimism. The overall effect for organizations as a whole depend on how which agency problems occur and in what matter they are managed.
2.2 Family ownership
Many articles in academic literature do not particularly agree on one single definition about family firms. Multiple business characteristics are taken into consideration in defining the core elements of a family
firm. Therefore, there is not a single most important condition that defines family ownership. Family firms show characteristics that are not found within non-family firms, but they differ from each other as well. Research from Villalonga and Amit (2006) provides three fundamental elements in defining family firms; ownership, control and management.
According to Chen, (2014) family firms have some fundamental characteristics which are different to those from non-family firms. Firstly, he states that family owners express more long term incentives than other shareholders. The investment horizon of family owners is therefore longer, because they feel incentivized to pass the assets on to future generations. The long term horizon implies that family owners act less on short-term gain, and rather create or gain value on the long term.
Secondly, Chen (2014) states that on average, founding families are more actively involved in managing their firms. Especially within the S&P 1500. In 98,4% of the cases, founding families appoint at least one or more members of the family to the boards. In 56,4%, these are at least two or more persons are appointed in the board and in 22,9% of the family firms there are at least three or more members of the family appointed to the board. Therefore, Chen (2014) implies that founding families want to keep or even gain considerable involvement in the management of their firm. Families in this case want to make sure that their preferences are reflected in the company.
Thirdly, Cheng (2014) states that family firms, due to more concentrated ownership, hold less diversified portfolios. In the S&P 1500 for example, where family firms on average hold 17% of the shares in the firm. Moreover, 69,5% of the family firms own 5% of the shares and 24,7% hold more than 25% of the shares. So, due to high ownership and less diversification, founding families have strong incentives to increase firm value. They enjoy the benefits of positive corporate investments but at the same time they are exposed to the consequences if these investments fail to generate positive cash-flows. Due to low diversification, the founding families are heavily exposed to this risk.
Most of the studies to family firms in the literature have been established using the agency theory framework. The position in this framework is slightly different for family firms compared to non-family firms. The agency conflict between principal and agent almost not exist due to non-family members whom are owners and are also representing management (Chrisman, Sharma & Tagger, 2007). Besides the overall matter of the principal and agent there are two major agency problems which have been addressed frequently throughout the relevant literature.
The first agency could be addressed as one of the most classic tensions between management and shareholders. The separation of control and ownership could lead to actions conducted by management which are not in the best interest of the shareholders. Also known as type 1 agency problems (Jensen & Meckling, 1976). However, the general impression is that family firms experience reduced effects of type I agency problems. Research done by Cheng (2014), states several reasons which cause the reduction of type I agency problems within organizations.
At first, family owned companies hold a very concentrated and under-diversified ownership. Less diversification causes families to bear the risk that is inherent to the firm and for future cash-flows.
Secondly, as mentioned earlier, family firms tend to have more long-term incentives than other shareholders. They often see the entire firm as a whole asset which has to be passed on to future family generations. Where non-family management may obtain more short-term incentives, family management probably does put more emphasis on long term. Therefore, type I agency is reduced (Cheng, 2014).
Thirdly, families are overall concerned with their family’s reputation. Preserving a family’s reputation is likely to cause long term effects. This has to do with the distinguishing factor of family firms compared to non-family firms. Family firms tend to care more about socioemotional wealth, which is a concept that somehow summarizes a family’s different attitude against gaining and increasing value in favor of reputation (Deephouse & Jaskiewicz, 2013). Because family’s tend to care about their socioemotional wealth, they are likely to show more risk-aversion compared to non-family firms, due to family’s wealth that has been concentrated in the organization (Basu, Dimitrova & Paeglis, 2009). Research that has been conducted in Egyptian firms has shown a positive association between risk aversion and anti-sticky cost behavior. Firms who show risk averse behavior, tend to decrease resources in periods of sales decrease, while risk taking firms show contrasting behavior (Salamah & Abulezz, 2017).
Besides ownership, control is an important essence in the theory of socioemotional wealth, if ownership and control are in the hands of the same person and or family within the organization., there is no misalignment of interest between ownership and control (Cheng, 2014).
The second major agency problem addresses a problem between minority and majority shareholders. Most of the time, families hold a considerable amount of shares, and control within the organization. Families are therefore mostly large shareholders with a small matter of other shareholders. The agency conflict that arises is about the majority of the shareholders whom may seek benefits at expense of the small group of other shareholders (Shleifer & Vishny, 1986). Also known as type II agency problems (Jensen & Meckling, 1976).
Research by Cheng (2014) showed that type II agency problems within family firms primarily are being caused by the concentrated amounts of ownership and substantial control. This provides opportunities for management to enjoy their substantial freedoms and power and to attain personal benefits on the expense of other shareholders. Besides inappropriate wealth expropriation at cost of the minority shareholders, it is also possible their interests may be sacrificed. Instead of paying out dividends, management decides to retain all funds on behalf of future firm growth (Carney & Gedajlovic, 2003).
Moreover, founding owners obtain more control through ‘disproportionate board representation’, voting agreements, pyramid ownership structures and dual-class share structures. These matters provide managing family members with the opportunity to seek benefits for their personal wellbeing. Most of these decisions will be suboptimal or inefficient to the organization as a whole, and therefore reduce value to minority shareholders (Cheng, 2014).
2.3 Dual class shares
As mentioned earlier in this paragraph, control is emphasized as one of the fundamental characteristics for family firms, in which they could be distinguishing (Villaonga & Amit, 2006). One of the control-enhancing attributes that can be used, and often is used in prior literature, is the use of dual-class shares. Dual-class shares are present in organizations if there are mostly two or more so called share classed, which have contrasting factors. Typical dual-class companies have two classes of shares. One ‘superior’ class of shares, which mostly contains multiple votes per share and is likely not publicly traded. The other kind are the inferior class. These are the ‘ordinary’ shares, which have mostly one vote per share and in are in general publicly traded (Masulis, Wang & Xie, 2009). The somewhat disproportional differences in voting- and cash-flow rights are considered to have much weaker alignment of interest between management and shareholders which own inferior shares (Smith, Amoako-Adu & Kalimipalli, 2009)
Thus, with the existence of dual-class shares it is possible to expropriate more voting- or cash-flow rights as a shareholder. Through this mechanism founding families could be able to leverage control which is not corresponding with their nominal equity stake. Villalonga & Amit (2009) provided an example of the use of dual-class shares. Back in 2000, a founder and his son owned only 3,14% of the equity stake in their company, but because of the presence of dual-class shares they owned 85,64% of the voting rights. Mostly, families are able to hold the majority of the voting power, even if all shares are in possession. This is because of a so called wedge between cash-flow rights and voting rights, that has been created by family firms. Moreover, besides the use of dual-class shares in non-family firms, it has been shown that family firms are the parties which most often use this control enhancing mechanisms (King & Santor, 2008).
Based on these literature, the use of dual-class structures is mostly focused on circumstances where additional control is enhanced within organizations. The consequences of such structures is mostly associated with the expropriation of wealth on expense of shareholders. A study conducted by Gompers, Ishii and Metrick (2010) finds that the presence of dual-class shares therefore has a negative impact on firm performance. Findings Masulis et al., (2009) find that insiders who held more voting rights relative to cash-flow rights do actually extract benefits on shareholder’s expense. Managers are more likely to be engaged in the so called empire-building activities. All at costs of the minority shareholders, and mostly inefficient activities which reduce or destroy value on behalf of the other shareholders. On the other hand, do findings of Anderson, Duru and Reeb (2009) suggest that the enhancement of control leads to more powerful monitoring, through which families can exercise control towards managers, which likely would lead to reduction of type II agency problems.
2.4 Hypothesis development
The agency theory predicts that a separation of ownership and management in companies results in conflicts of interest where self-interested managers have incentives to extract their own personal wealth at expense of (minority) shareholders. Research has shown that managers can act on such behavior by either adjusting or not adjusting resources (Chen et al., 2012). Actions that depend on manager’s optimism and decision making. Their discretion to allocate costs and adjust resources are established in the academic literature as the determinants of cost stickiness (Malik, 2012). I will first examine whether cost stickiness is actually present in this sample, therefore the first hypothesis is stated as follows:
H1: costs rise to a larger extend compared to sales activity increase, than they decrease if sales activity decreases.
Based on the determinants of cost stickiness and the consequences that has been established in prior literature the hypothesis is stated. More important is the consideration regarding the agency problems. As Anderson and Reeb (2003) implied, family firms mostly experience agency problems type II, because effective monitoring and enhancing mechanism are not present. Mostly because family firms tend to lack transparency and corporate governance mechanisms, to keep or get family members on the board (Anderson and Reeb, 2003).
Moreover, the long-term vision of family firms is a distinguishing factor for this hypothesis development. Family firms tend to preserve the family-business and they possess more long-term incentives than other shareholders (Cheng, 2014). However, the academic literature provides two ways to this matter combined with cost stickiness. On the one hand, it is stated that family firms tend to own stickier Research & Development costs (Ahn, Hwang & Park, 2015), which emphasizes the long-term vision, since R&D costs represent decision-making without certain short-term results. This would imply stickier costs for family firms compared to non-family firms. But on the other hand, has research to family firms in the Phillipines shown that family firms show less SG&A cost stickiness, compared to non-family firms (Uy, 2014). These research results therefore imply that family-firms are better to mitigate the agency costs in the organization. If SG&A costs are not sticky, it means that the firms are able to adjust resources to changes in activity. Because these cost do not act sticky, the principal agency problem is mitigated. Therefore, the second hypothesis is stated as follows:
H2: Family firms experience less cost stickiness than non-family firms
Besides data-availability on family firms or non-family firms, the data from Anderson and Reeb (2009) also provides information about certain dual-class share structures. Dual-class shares show different characteristics compared to single shares. They mostly differ from ordinary shares regarding
voting rights and dividend rights. These hypotheses are based on the potential agency problems arising within dual-class share structures. Firms with dual-class shares structures tend to be more exposed to agency type II problems. Insiders, holding more information and more voting rights, tend to express more expropriation of wealth on expense of the shareholders, also known as empire building (Masulis, Wang & Xie, 2009). Thus, aligned with the research from Chen et al. (2012), which showed that empire building leads to stickiness of SG&A costs, the following hypothesis is stated as follows:
H3: Family firms, who hold dual-class share structure, experience more cost stickiness than-family firms who do not hold dual-class shares
3. Data sample and empirical model
In this section I will provide information about the empirical models what will be used to examine cost stickiness and family firms. After that, all the variables will be operationalized and lastly, the sample will be described.
3.1 Cost stickiness Model:
Measuring the degree of cost stickiness will be done through the model following the Anderson et al. (2003) research. log $%&' ), + $%&' ), + − 1 =b0 + b1 log $1234 ), + $1234 ), + − 1 + b2 67889:;<=;>?; @,A ∗ log $1234 ), + $1234 ), + − 1 +e), +
This model is derived from the Anderson et al. (2003) research and will be used to show whether cost stickiness is present in a company. The logarithm (SG&A) is the log change in the SG&A costs for a firm. The i and the t represent the firm and the year. SG&A costs do form an important role in the adjusting abilities by management and are therefore important for this thesis (Chen et al., 2012). Furthermore, many drivers of SG&A costs are proven to be in a relationship with sales volume (Anderson et al., 2003; Cooper & Kaplan, 1998). SG&A costs are available and are accessible through the Compustat, making them a proper instrument to examine cost stickiness.
The logarithm (Sales) is the log change in net sales in firm i in year t, relative to firm i in year
t-1. By using this proxy, the method used Anderson et al. (2003) is followed, making sales revenue an
valuable and common way of measuring activity levels of organizations. The sales variables in this thesis will also be derived from the Compustat database and will be equal to the net sales / turnover. The dummy variable for a decrease in sales will be the interaction variable in this theses. The variable equals 1 if the revenue of firm i in year t will be lower than the revenue of firm i in year t-1. Logically, the dummy variable equals 0 if the revenue of firm i in year t will be increased compared to the revenue of firm i in year t-1. Because of the dummy variable the effect between SG&A costs and a decrease in sales revenues can be computed.
Therefore, b1 and b2 measure the percentage in sales decrease which are associated with a 1% decrease in SG&A costs. If the equation above equals a negative value, costs are considered sticky, which implies that the SG&A costs decrease less if sales decrease with an equal amount. Following the Weiss (2010) literature, this implies that managers are considered to reduce costs less when sales activity decreases, compared to cost increase if sales increase.
3.2 Family ownership variables
The database for family firms is based on the database used by Anderson et al. (2009). Their database contains a sample about family firms and the presence of dual class shares. The database shows information updated until may 2012. They have used a dummy variable to mark firms as family firm. If families hold a 5% equity stake or more, they are marked as family firm. The sample is based on 2000 largest firms in North-America from 2001 to 2010. It holds a sample of 16,200 firm year observations. In this thesis, the same family firm criteria are applied. Firms are marked family firm when the family holds 5% or more in equity stake. It is implied that the family data is considered family when the company is either founded or has heir ownership. Anderson et al. (2009) indicated the family firms with the use of a dummy variable. The value of a family firm is one, and zero when otherwise. Furthermore, the dataset of Anderson et al. (2009) contains dummy variables about the presence of dual-class shares. Again, a dummy variable indicates the presence of such shares. The value of presence of dual-class shares in firms is equal to one, zero implies otherwise.
Furthermore, data regarding cost stickiness and control variables will be obtained through the COMPUSTAT database. This database provides fundamental Annual and Industry specific information about the North-American companies. Using the GVKEY, which is provided by Anderson et al. (2009), the family ownership and dual-class shares data will be merged into one dataset.
3.3 Control variables
Completely aligned with prior literature this thesis will contain control variables which control for correlations in ownership and cost stickiness, but which are not the main aim of the research. Therefore, two sets of variables will be controlled for; agency factors and economic factors. First of all, there are agency factors that may factor SG&A cost stickiness.
To control for agency factors, this thesis will control for free cash flow (FCF), which will be calculated using the cash flow from operating activities (in year t) minus capital expenditures (in year t), scaled by current assets. A higher amount of FCF rises potential for corporate management to overspend SG&A costs, especially when activity increases Also the other way around, when activity decreases, and therefore causes a lower FCF, SG&A costs could be postponed. More FCF could therefore lead to cost stickiness (Chen et al., 2012).
Furthermore, this thesis controls for economic factors which may interfere with the effect of ownership on cost stickiness. Firstly, there will be controlled for EMPLOYEE INTENSITY (EMPLOYE EINT). This will be calculated as the ratio of number of employees to total sales. The second economic factor is ASSET INTENSITY (ASSET INT). This will be calculated as the ratio of total assets to total sales. Both asset- and employee intensity show positive associations with cost stickiness. Companies with a high amount of fixed assets are highly vulnerable to cost stickiness, since fixed assets are hard and costly to change, drop or fix and therefore appear to be stickier (Anderson et
al., 2003). Employee intensity shows a positive association since it could be costlier to dismiss or educate employees whom may not be necessary to the company anymore (Anderson et al., 2003).
The fourth control variable is about the fiscal YEAR. The database which will be used from Anderson et al. (2009) contains data from 2001 to 2010. As Campello, Graham and Harvey (2010) pointed out, the financial crisis during this period of team also could have affected cost stickiness, moreover about some unregularly SG&A expenses, which may have caused differences for companies.
The fifth and variable controls for INDUSTRY. Following the Subramanian and Weidenmier (2003) research, the different industries will be specified using SIC-codes. These codes will be transformed into four industry categories. Manufacturing, Merchandising, Service and Financial services firms. Every industry has his own characteristics which will cause different cost stickiness drivers (Subramanian & Weidenmier, 2003). Financial services, as mentioned, will be excluded from the sample.
The sixth control variable controls for successive years of revenue decrease. This variable is mentioned by Bugeja, Lu and Shan (2015) and is measured through a dummy variable. This dummy variable controls for effects which are caused by a situation of sales decrease of firm i in year 1 and
t-2. It is proven that managers will use their discretion differently. According to Bugeja et al. (2015),
managers tend to cut less resources in year t, which will cause more cost stickiness. However, Chen et al., (2012) made a contrasting contribution. According to them, managers will be less optimistic about the future and do less investments in capacity and try to reduce slack in resources. Thus, their contribution differs from the one from the Bugeja et al., (2015) research. For this research it is likely that the effect of successive years of sales decrease be smaller for family firms compared to non-family firms. The dummy variable equals 1 if the revenue of firm i in year t is smaller than the revenues of firm
i in year t-1 and year t-2. The variable equals 0 if otherwise.
All variables are obtainable through the COMPUSTAT database.
3.4 Sample selection:
First of all, the sample contained 16,230 observations, spread by 2000 companies. Based on the
Anderson & Reeb (2003) research, cost stickiness should be researched containing data
collection from over 20 years. Because the database contains samples of 2001 until 2010 I chose
to only keep observations of companies which had observations spread over 10 years.
Companies which have become inactive during that period and which dit not had observations
for the years 2001 – 2010 in the database are dropped. Based on observations of 10 successive
years, a total of 12,460 observations was held, spread over 1246 companies.
Using the company identification numbers of the 1246 companies I obtained data from
Compustat. After obtaining the data, these were merged with the data coming from the
Anderson & Reeb(2009) database. After merging data, the sample consisted 12,460
observations. First of all were all missing values dropped. These are missing values of all the
variables which will be used in the regression. Therefore 965 observations are excluded.
Furthermore, all observations where the SG&A costs of year t and company i, exceed the sales
were excluded as well. Next, the observations containing negative values on SG&A and sales
were dropped, which were 1109 observations. Furthermore, the financial institutions were
excluded from the sample. Using the SIC-code between 6000-6999 are used to distinguish these
from the others. Financial institutions are being excluded because prior research states that these
kind of organizations act differently than companies in other industries. Therefore, 149
observations are excluded. Finally, there has been a total 333 observations excluded, which
were marked as outlier. The top and bottom one percent of the sample were dropped. The final
sample contained a total of 9,928 observations.
Table 1: Sample selection
Number of observations
Initial sample
12,640
Less: Missing values
965
Less: SG&A > Sales
156
Less: Negative values SG&A, Sales
1109
Less: Financial institutions (6000 - 6999)
149
Less: Top 1% outliers SG&A, Sales
152
Less: Bottom 1% outliers SG&A, Sales
181
Final sample
9,928
3.5 Empirical models
The empirical models test the relationships between the level of cost stickiness, family firms and dual-share structures. To test the first hypothesis, regarding the existence of cost stickiness in the sample the following empirical model is used. This model is derived from the Anderson et al., (2003) research. log $%&' ), + $%&' ), + − 1 =b0 + b1 log $1234 ), + $1234 ), + − 1 + b2 67889:;<=;>?; @,A ∗ log $1234 ), + $1234 ), + − 1 +e), +
As mentioned earlier, the b2 Decrease Dummy variable equals 1 if sales revenues in year t are
lower compared to those from year t-1. If b1 is positive, this means that the dependent variable, SG&A costs will increase a unknown percentage in case of a 1% increase in sales activity. If the sum of b1 and b2 is smaller than b1, cost stickiness exists within organizations. If the sum of b1 and b2 is greater than b1, we speak of anti-stickiness, where costs fall to a larger extend, for a 1% sales decrease. Thus, for the first hypothesis, I expect that the b1 + b2 < b1, which indicates cost stickiness in the entire sample, not taking ownership or share structure in consideration.
To test the second hypothesis regarding family ownership and cost stickiness the following empirical model is used:
Empirical model 1: Cost stickiness within family firms
log $%&' ), + $%&' ), + − 1 =b0 + b1 log $1234 ), + $1234 ), + − 1 + b2 67889:;<=;>?; @,A ∗ log $1234 ), + $1234 ), + − 1
+ b3 67889D>E@FG), ++ b4 67889D>E@FG @,A∗ log $1234 ), + $1234 ), + − 1
+ b5 67889D>E@FG @,A∗ 67889:;<=;>?; @,A∗ log
$1234 ), +
$1234 ), + − 1 +b6
− 12 KLM+NL2 O1N)1P234 ), ++ e), +
The base model is still derived from Anderson et al., (2003), but the model has been expanded using the variables which are needed to run the regression. b3 indicates the dummy variable, which equals 1 if a firm is family firm and equals 0 if otherwise. b4 measures the influence of the family dummy and the logarithm change in sales. Statistically this measures, that for every 1% more family firm, the effect on SG&A costs. b5 measures the same as b2, only with the influence of the family dummy. Based on the hypothesis and theory, it is expected that b5 > b2, which would indicate a positive effect of family firms on cost stickiness, where a positive effect means less cost stickiness because SG&A costs decrease more if b2 of b5 is positive.
Empirical model 2: Cost stickiness in family firms with dual share structure
To test the third hypothesis, regarding family-firms who held dual-share structures, the following empirical model is going to be used.
log $%&' ), + $%&'), + − 1 =b0 + b1 log $1234 ), + $1234 ), + − 1 + b2 67889:;<=;>?;@,A ∗ log $1234 ), + $1234 ), + − 1
+ b3 67889D>E@FG:Q>F), ++ b4 67889D>E@FG:Q>F @,A∗ log $1234 ), + $1234 ), + − 1
+ b5 67889D>E@FG:Q>F @,A∗ 67889:;<=;>?; @,A ∗ log
$1234), +
$1234 ), + − 1 +b6
− 12 KLM+NL2 O1N)1P234 ), ++ e),
Once again, the base model for cost stickiness is derived from the Anderson et al., (2003) research. Compared to the second empirical model, this model has a lot of similarities. However, instead of the family dummy variable, a family-dual dummy is used in this model. The dummy variable equals 1 if an observation contains a family firm and when there is a dual-class share structure. The dummy equals 0 if otherwise.
4. Empirical findings
In this chapter, I will give an overview of the descriptive statistics, the correlation matrix and the results of the multiple regression analyses. I will use the results of the regressions to test whether my hypotheses are supported, and I will present the results.
4.1 Descriptive statistics
Panel A shows results of the descriptive statistics of annual SG&A costs (in millions of dollars) and sales (in millions of dollars). The firms in the sample show an average of 822.50 million in SG&A costs against a mean of 270.94 million. Sales show an average amount of 4427.79 million, against a median of 1423.88 million. Since there are large differences between the mean and median of both variables, a normal distribution would probably not been found. Likely because a small number of firms has abnormal high annual SG&A costs, causing the increase the mean compared to th median. Therefore, I calculated the natural logarithm of both variables to control for heteroscedasticity. As a consequence of the natural logarithms we can see that the means and medians of both SG&A costs and Sales are almost identical. Which implies a normal distribution.
Panel B provides descriptive statistics of the interaction variables. All interaction in this thesis are created dummy’s. Firstly, the sample consist approximately of 9.5% of observations which includes family-firm and dual-share structures. The median is 0, which implies that more than 50% of all observations (n=3,352) does not have a combination of dual-shares and family firm. Secondly, over 33,8% of all observations includes family firms. Since the median equals 0.00, more than 50% of all observations is not a family-firm. Thirdly, approximately 23.4% of the observations contains decreases in sales activity, whilst the median is 0.00 once again.
Panel C provides descriptive statistics about the control variables. Firstly, the mean FCF of 0.064 (median = 0.059) implies that firms, on average had a FCF of 6.4% (median = 5.9%) relative to total assets. Secondly, the employee intensity has a mean of 0.006 (median = 0.004). This means that to generate 1 million dollars of sales revenue, on average and approximately 60 (median = 40) employees are needed. Thirdly, asset intensity has a mean of 1.265 (median = 1.011), which means that to generate every 1 million dollars of sales revenue, approximately 1.265 (median = 1.011) million dollars’ worth of assets are needed. Lastly, successive decrease has a mean of 0.076 (median = 0.00). This means that approximately 7.6% of all observations experienced successive sales decreases. Because the median is 0.00, over 50% does not have successive decreases in the period 2001 – 2010.
Table 2: Descriptive Statistics
Panel A: Descriptive statistics SG&A costs & Sales
Variables Mean Median St.dev Minimum p25 p75 Maximum
SG&A ($millions) 822.50 270.94 1635.43 11.662 110.363 747.66 14563.00
Sales ($millions) 4427.79 1423.88 8605.10 79,92 601.751 3999.94 77953.88
∆ Log(SGA i, t / SGA i, t-1) 0.062 0.061 0.183 -3.176 -0.012 0.135 2.702
∆ Log(Sales i,t / Sales i, t-1) 0.063 0.067 0.193 -1.724 -0.010 0.146 1.783
SG&A costs as % of Sales 0.234 0.207 0.151 0.003 0.117 0.317 0.929
Panel B: Descriptive statistics interaction variables
Variables Mean Median St.dev Minimum p25 p75 Maximum
Family firms & Dual Share structure (Dummy) 0.095 0.00 0.293 0.00 0.00 0.00 1.00
Family firms (dummy) 0.338 0.00 0.473 0.00 0.00 1.00 1.00
Sales Decrease (dummy) 0.234 0.00 0.424 0.00 0.00 0.00 1.00
Panel C: Descriptive statistics control variables
Variables Mean Median St.dev Minimum p25 p75 Maximum
Free Cash Flows (FCF) 0.064 0.059 0.081 -0.387 0.022 0.100 3.800
Employee intensity 0.006 0.004 0.008 0.000 0.003 0.006 0.164
Asset intensity 1.265 1.011 0.970 0.063 0.677 1.516 23.574
4.2 Correlation matrix:
I have conducted a correlation test for all the main variables of this thesis. Correlations indicate relationships between two variables. Values between 0 and 1 are seen as positive values, which indicate a positive correlation between two different variables. Values between 0 and -1 indicate negative correlation between two different variables. To account for multicollinearity, VIF indicators are computed. None of the computed VIF has a value above 10, so there is no concern about multicollinearity problems. Moreover, the Pearson correlation matrix does not show any interactions above 0.7 for the independent variables, which indicates that there are no direct correlations.
The correlation table above shows many significant correlations between the different variables. First of all, the results show that the ∆LOG_SALES and ∆LOG_SG&A are significantly positively correlated. This implies that these variables are moving in the same direction, but still show different patterns. Secondly, Family firms show a significant positive association with family firms who held dual-share structures. The results therefore imply that these variables move differently of each other, but in the same direction. This is a necessary contrast for this research, because it separates family firms who do not
held dual shares from the family firms who do held dual share structures. Furthermore, family firms show positive significance correlation with employee intensity, which implies that family firms tend to operate with more employees
Table 3: Correlation Matrix – Pearson / Spearman
Notes: This table provides pair-wise correlations for firm-year observations from fiscal years 2001-2010 for the main variables for the initial sample (n=9919). Panel A shows the correlations for family ownership and the family firms who held dual-class share structures. Top right shows Spearman and bottom left Pearson Correlations. The fixed effects of fiscal years and industries are excluded from the correlation matrix.
*** indicate significance at the 1% level ** indicate significance at the 5% level * indicate significance at the 10% level
Panel A: Initial sample containing family firms & dual-class share structures
Variable Δ LOG_SG&A Δ LOG_SALES Sales Decrease Family firms Family_
dual Free Cash Flow Employee intensity Asset intensity
Successive decrease Δ LOG_SG&A 1 0.731*** -0.563*** -0.011 -0.042*** 0.009 -0.047*** 0.060*** -0.319*** Δ LOG_SALES 0.660*** 1 -0.773*** -0.019 -0.054*** 0.022* -0.100*** 0.036*** -0.397*** Sales Decrease -0.452*** -0.659*** 1 0.006 0.020 -0.032** 0.025 0.035** 0.0508** Family firms -0.013 -0.010 -0.003 1 0.457*** -0.029** 0.121*** -0.076*** 0.038*** Family_dual -0.027* -0.029** 0.016 0.453*** 1 -0.041*** 0.050*** 0.014 0.038***
Free Cash Flow -0.087*** -0.042*** 0.003 -0.021* 0.035*** 1 -0.031** -0.128*** -0.031
Employee intensity -0.033** -0.044*** 0.015 0.057*** 0.039*** -0.010 1 -0.136*** 0.040***
Asset intensity -0.067*** 0.011 0.037*** -0.005 0.075*** -0.183*** -0.097*** 1 -0.001
4.3 Regression analysis:
In the first hypothesis it is predicted that SG&A costs increase to a larger extend when sales activity increases, than they decrease when sales activity decreases. It is an overall stated hypothesis which does not take ownership structure in consideration. Thus, the first hypothesis tests the existence of cost stickiness in this sample. As explained in the empirical model paragraph, the Anderson et al., (2003) basic model for cost stickiness will be used to determine if SG&A costs behave sticky, in the periods t and t-1 in the initial sample.
In table 4 the b1 has an estimated value of 0.700, this is significant with a t-value of 58.05. This value implies that if sales activity increases with 1%, the SG&A costs approximately would increase with 0.70%. The value of b2 is an estimated – 0.164 and is also significantly. This value provides evidence for the existence of cost stickiness in the initial sample. If sales activity would decrease with 1%, this would mean that the SG&A costs ‘only’ decrease with 0,536%. This supports the first hypothesis, since the SG&A costs behave asymmetric to changes in sales activity. The results support the findings of Anderson et al., (2003). However, the results are slightly different. Anderson et al., (2003) found that SG&A increase 0.55% on average at a 1% sales activity increase. If sales activity decreased 1%, the SG&A costs decreased with only 0.35%. Possible explanations could have something to do with the difference in cost structure of companies or strategy’s. Furthermore, the sample used by Anderson et al., (2003) contained over 20 years of observations. The timeframe the initial sample used in this thesis is shorter. This could have caused some of the differences in the values that have been found. The adjusted R2 of this sample is 0.4394, which indicates that 43.94% of the changes in SG&A costs is explained by the first model. Furthermore, the F value is 3311.26 with a p-value of 0.000, which indicates this model is significant.
Table 4: Cost stickiness – intitial sample
Regression analysis Cost Stickiness initial and full sample, family and non-family firms
Dependent variable Δ LOG_SG&A
Explanatory variables CF b1 ΔLOG_SALES 0.700*** (58.05) b2 SALES_DEC * ∆ LOG_SALES - 0.164*** (-7.86) Intercept 0.012*** (5.81) Industry/year dummies NO N 8,447 Adjusted R2 0.4394 Prob>X2 0.000 F 3311.26
Notes: This table reports the results of the OLS regression, to examine the existence of Cost Stickiness within this sample. The sample consists 8,447 firm-year observations from fiscal years 2001 – 2010, where the dependent variable is ∆ LOG_SG&A. The regression OLS regression shows results without the main effects, and there is not controlled for interaction effects or/and fixed effects derived from fiscal year or industry
*** indicate significance at the 1% level ** indicate significance at the 5% level * indicates significance at the 10% level.
Hypothesis 1a predicts that the extent in which family firms suffer from cost stickiness is smaller than for non-family firms. First of all, cost stickiness within family-firms is assessed using the base model from Anderson et al., (2003). This model is elaborated with just the main effects from family firms. So, other control variables and interaction effects are not taken into consideration in table 5. The results of the regression analysis are shown in table 5 below.
As table 5 shows the b1 is significantly positive, with a coefficient of 0.676 and a t-value of 46.37. This implies that for every 1% of sales activity increase, the SG&A costs increase with 0.676%. The b2 supports, once again, that cost stickiness is indeed proven in this model. The sum of b1 and b2 shows that for every 1% sales activity decrease, the SG&A costs decline with 0.49%. Thus, cost stickiness is supported. Every value that b2 takes, and is more negative, will only increase the degree of cost stickiness. Following hypothesis 1a it is expected that the value of FAMILY * SALES_DEC *
LOG_SALES will be positive, since it is expected that family firms experience less cost stickiness
compared to non-family firms. If the coefficient of this variable is positive, less cost stickiness is implied. The coefficient has a value of 0.068, with a t-value of 5.19 which shows that family firms have an insignificant positive main effect on cost stickiness. The results show that the coefficient of b4 is greater than b2, which indicates the different behavior from family firms when sales decrease compared to SG&A costs. The FAMILY variable shows a coefficient of -0.004, with a t-value of -0.88. The
coefficient of -0.004 tells us that there is an insignificant negative effect of family firms on changes in the SG&A costs.
Overall, the adjusted R2 of this model is 0.4423, which means that 44.23% of the changes in SG&A costs are explained by the explanatory variables of this model. Furthermore, the model is significant with a F-value of 1340.41. Both coefficients are significantly at the 1% level, therefore hypothesis 1 is supported.
Table 5: Cost stickiness – family firms
Regression analysis Cost Stickiness family firms initial sample – without control variables
Dependent variable ΔLOG_SG&A
Explanatory variables CF b1 Δ LOG_SALES 0.676*** (46.37) b2 SALES_DEC * LOG_SALES -0.186*** (-7.86) b3 FAMILY -0.004 (-0.88)
b4 FAMILY * SALES_DEC * ΔLOG_SALES 0.068
(5.19) Intercept 0.013*** (5.19) Industry/year dummies NO N 8,447 Adjusted R2 0.4423 Prob>X2 0.000 F 1340.41
Notes: This table reports results of the OLS regression using ∆LOG_SALES as dependent variable for 8,447 firm-year observations from 2001-2010. The regression OLS regression shows results without the main effects, and there is not controlled for interaction effects or/and fixed effects derived from fiscal year or industry.
*** indicate significance at the 1% level ** indicate significance at the 5% level * indicate significance at the 10% level.
In the following regression analysis, the entire empirical model as described in the theory section. The first column shows the regression analysis including all control variables, for the main effects. The second column provides results for the regression analysis including interaction effects of all control variables.
First of all, both of the models explain cost stickiness. The coefficients of b1 are the same for the both columns, while the coefficients are both significant. The coefficients tell us that for every 1% increase in sales activity, the SG&A costs will increase with 0.651%. The coefficients for b2 are slightly different. For the first column, the negative value of -0.171 is significant at a 1% level and implies in case of a 1% sales decrease a SG&A costs decrease of 0.48%. The b2 in the second column has a
negative value of -0.180 and is significant at a 10% level. If sales decrease with 1%, SG&A costs will decrease with 0,471%. Both b1+b2 are smaller than b1, so cost stickiness is supported for both regression models. Furthermore, the coefficients of b3 show that family firms with sales decrease in year t, compared to t-1, have a significant positive effect on changes in SG&A costs. In line with the results in table 6, b5 shows that family firms indeed have a positive influence on cost stickiness. However, the coefficient is insignificant.
The results on ASSET_INT and SUC_DEC in the first column, show contrary results compared to expectations derived from the literature. ASSET_INT Shows a significant positive coefficient of 0.012, which suggest less cost stickiness for firms with higher asset intensity. For SUC_DEC, both of the columns show contrary results. In the first column, the coefficient of 0.027 SUC_DEC is significant positive. The coefficient tells us that family firms who experienced successive sales decreases, show less cost stickiness on SG&A costs. This outcome supports the outcome of Chen et al., (2012), who stated that firms with successive sales decrease experience less cost stickiness due to less optimistic behavior by management. However, the model which included interactions (column 2), suggests otherwise. The coefficient of – 0.010 implies more cost stickiness for family firms with successive sales decrease. These outcomes support the Bugeja et al., (2015) outcome, where firms experience more cost stickiness is firms suffer successive sales decreases. However, results from the interaction model are not significant for SUC_DEC.
The interaction term b10 was expected positive, following the Chen et al., (2012) research. The significant positive coefficient of 0.122 supports the Chen et al., (2012) research. Managers seem to be less optimistic when sales activity decrease successively, and this is seen more permanent. Therefore, it is likely that costs are reduced and cost stickiness is reduced. The interaction term b11 was expected to be negatively for family firms. Based on the type II agency problems, it is likely that a higher free cash flow opens up the opportunity for managers to make inappropriate expenses, or inefficient investments. However, the coefficient of 0.214 significantly positive. Therefore, it is implied in this sample, that free cash flows tend to decrease cost stickiness, when sales decrease. For b12, a negative coefficient was expected. Since it mostly difficult to ‘get rid’ of personnel quickly, while this is beneficial for the level of cost stickiness. However, in this model there has not been an effect measured, thus it is difficult to say something about the interaction effect om employee intensity in this sample. The interaction term of b13 was also expected to have a negative coefficient. Higher asset intensity would probably make it difficult to adjust these costs on the short term and reduce cost stickiness. The significant coefficient of – 0.070 supports this expectation. Negative coefficients reduce the adjustment on SG&A costs and therefore increase cost stickiness. The R2 adjusted of both columns are respectively 0.4567 and 0.4659, meaning that in the first column 45.67% of the changes in SG&A costs are explained by the explanatory variables. In the second model is 46.59% of the changes in SG&A costs are explained by the explanatory variables and the interaction effect.
For neither column the family-firm’ coefficient is significant. Therefore, it is not possible to statistically say something about to what extent family firms experience less cost stickiness than non-family firms, hypothesis xxx is not supported by the data.
Table 6: Cost stickiness – family firms including control variables, interaction effects
Regression analysis Cost Stickiness family firms including Control Variables
Dependent variable Δ LOG_SG&A (1) Δ LOG_SG&A (2)
Explanatory variables CF CF b1 Δ LOG_SALES 0.651*** 0.651*** (43.33) (43.37) b2 SALES_DEC * LOG_SALES -0.171*** -0.180* (-6.85) (-2.21) b3 FAMILY -0.005 -0.009 (-1.17) (-2.01) b4 FAMILY * SALES_DEC 0.079** 0.098*** (3.07) (3.82)
b5 FAMILY * SALES_DEC * ΔLOG_SALES 0.032 -0.061
(5.20) (-1.34) b6 FCF -0.101*** 0.009 (-4.92) (0.39) b7 EMPLOYEE_INT 0.010 0.031 (0.05) (0.16) b8 ASSET_INT 0.012*** 0.010*** (6.21) (5.38) b9 SUC_DEC -0.027*** -0.010 (-4.82) (-1.41) Interaction effects
b10 SUC_DEC * SALES_DEC * LOG_SALES 0.122***
(3.45)
b11 FCF * SALES_DEC * LOG_SALES 0.214***
(7.68)
b12 EMPLOYEE_INT * SALES_DEC * LOG_SALES 0.000
(-0.02)
b13 ASSET_INT * SALES_DEC * LOG_SALES -0.070***
Table 6 : Continued
Dependent variable Δ LOG_SG&A (1) Δ LOG_SG&A (2)
Intercept -0.015 -0.005
(-0.67) (-3.52)
Industry/year dummies YES YES
N 8,447 8,447
Adjusted R2 0.4567 0,4659
Prob>X2 0.000 0.000
F 284.95 255.09
Notes: This table reports the results of the OLS regressions using ∆ LOG_SALES as the dependent variable for firm-year observations from fiscal years 2001-2010. The sample in the first column consists of 8,447 firm year observations. The first column shows the results of the OLS-regression which includes main effects, and where is controlled for fiscal years and industry effects. The second column shows results of the OLS regression, which also include interaction effects between the explanatory variables and ABJ’ (2003) model of stickiness. All numbers are rounded up to third decimal place.
*** indicate significance at the 1% level. ** indicate significance at the 5% level. * indicate significance at the 10% level.
4.5 Regression analysis dual share structure
Because hypothesis 1 has been supported in above mentioned results, we can say that the sample consists of cost stickiness. That is something that not has to be tested anymore for the hypothesis regarding dual share structures. First of all, the first hypothesis is going to be tested, whether dual structures cause companies to experience more cost stickiness, regarding ownership structure. Thus, this sample consists all observations.
First of all, the significant coefficient of b1 shows us that for every 1% increase in sales activity, SG&A costs rise with 0.691%. Again in this model, cost stickiness is supported. If sales activity decrease with 1%, SG&A costs decrease with 0.531% (0.691 – 0.160). The coefficient is significant. Furthermore, the coefficient of b3 shows that a dual-share structure has a significant, but small negative influence on SG&A costs. Most important for the testing of the hypothesis is b5. The coefficient of b5 shows an insignificant negative coefficient. This implies a negative influence on cost stickiness. However, this difference is insignificant. The R2 adjusted is 0.442, which means that 44.02% of the changes in SG&A
costs are explained by the explanatory variables. However, because the coefficient of b5 is insignificant, there is no support found for hypothesis xxx.
Table 7: Cost stickiness dual-share structure
Dependent variable ∆ LOG_SG&A
Explanatory variables CF b1 Δ LOG_SALES 0.691*** (55.31) b2 SALES_DEC * ∆LOG_SALES -0.160*** (-7.45) b3 DUAL_SHARES -0.010*** (-1.38) b4 DUAL_SHARES * SALES_DEC 0.136** (2.74) b5 DUAL_SHARES*SALES_DEC*ΔLOG_SALES -0.022 (-0.25) Intercept 0.0132*** (5.92) Industry/year dummies NO N 8447 Adjusted R2 0.4402 Prob>X2 0.000 F 1329.54
Notes: This table reports the results of the OLS regression using ∆ LOG_SG&A as dependent variable for firm-year observations from fiscal years 2001-2010. The sample consists of 8,447 firm-year observations. The results are shown using an elaboration of ABJ’s (2003) model. *** indicate significance at the 1% level, ** indicate significance at the 5% level, * indicate significance at the 10% level.
Based on table 8, we can say that there is no statistically support found for hypothesis xxx. The final hypothesis will be tested using, the base model of Anderson et al., (2003), the model including explanatory variables. Due to multicollinearity there is no regression which includes interactive effects. For this regression model the sample has been reduced to 2,791 observations. All non-family firms have been dropped.
Firstly, in both columns, the results show the existence of cost stickiness. For the first column, every 1% in sales activity decrease, SG&A costs decrease with 0,633% (0.745 - 0.112). Compared to 1% sales activity increase, the SG&A costs increase with 0.745%. The second column shows similar results. Every 1% increase in sales activity causes SG&A costs to increase with 0.736%. For every 1% decrease in sales activity, SG&A costs drop with 0.565%. So, cost stickiness has been found.
In both columns, FAMILY_DUAL has a slight positive, yet insignificant, influence on the SG&A costs. The coefficient of the effect between family firms with dual-shares, sales activity decrease and the logarithm change in sales, is expected to be negative. The basic model of Anderson et al., (2003), in the first model shows a negative coefficient, implying less changes in SG&A costs if sales activity decreases. Therefore, a negative association for cost stickiness has been found between the existence of a dual-class share structure and family firms. However, the coefficient in column 1, is insignificant.
For the control variables, the coefficient regarding free cash flows, is expected to be negative, due to manager’s behavior, when there are more free cash flows to spend. As expected, the coefficient of free cash flow is -0.199. This significant coefficient shows that there is a negative association with free cash flows and costs stickiness. The existence of large free cash flows, seems to encourage cost stickiness, because it subtracts the logarithmic change in SG&A costs. This result supports the findings of Chen et al., (2012), who stated that large free cash flows, cause SG&A costs to drift away from their desired level. Furthermore, asset intensity has a slightly positive significant impact on the changes in SG&A costs, which is slightly contrary against expectations. The coefficient for successive decreases is insignificantly negative. This suggest that successive sales decreases tend to be no reason for managers to aggressively adjust resources or cut slack out of their resources. Probably, due to the extensive voting rights, they do not feel the need to temper their optimism.
After running all regressions, it is fair to say that there is no support for hypothesis xxx. In this thesis there is not enough statistical evidence that suggest that family firms with dual class shares experience more cost stickiness compared to family firms without dual-shares structures.
