1
The effects of family ownership, founder-CEO presence and
dual-class shares on US firm performance before and during
the financial crisis
Master thesis
University of Amsterdam
MSc Business Economics: Finance track Berend Reeskamp 10431659
Supervisor: Dr. F. López de Silanes July 2016
This study examines the effects of family ownership, founder-CEO presence and dual-class shares on US firm performance before and during the financial crisis. A sample of 1,541 large US industrial firms is used for family ownership and dual-class regressions and a S&P1500 subsample of 1,221 firms is used for founder-CEO regressions. Although previous research suggests that family-owned firms outperform during non-crisis years, this study finds that during both the pre-crisis and crisis years, accounting performance (ROA) and market performance (Tobin’s Q) are not affected by family ownership. During pre-crisis years, however, dual-class firms do underperform in both ROA and Tobin’s Q and a value discount is also found for dual-class family-owned firms. These results are consistent with the literature. During crisis years, inferior accounting performance of dual-class family-owned firms relative to other family-owned firms is found. Additional results also show that founder-CEO family-owned firms outperform other firms during crisis years in terms of accounting performance. Overall, firm performance is not affected by family ownership in general during both pre-crisis and crisis years. However, during the financial crisis, dual-class family-owned firms perform worse and family-owned firms with founder-CEOs perform better. This evidence confirms the notion of previous non-crisis research.
2
Statement of Originality
This document is written by Student Berend Reeskamp 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
Content
1 Introduction ... 5
2 Literature review ... 7
2.1 Performance of family-owned firms ... 7
2.1.1 Theories and evidence on costs of family ownership ... 8
2.1.2 Theories and evidence on benefits of family ownership ... 9
2.1.3 Theories and evidence on endogeneity of family ownership ... 10
2.1.4 Hypothesis 1: Performance of family-owned firms ... 10
2.2 Founder-CEOs and firm performance ... 10
2.2.1 Theories and evidence on founder-CEOs and firm performance ... 11
2.2.2 Hypothesis 2: Performance of firms with founder-CEOs ... 12
2.3 Performance of family firms during the financial crisis ... 12
2.3.1 Theories and evidence on family firm performance during the financial crisis ... 12
2.3.2 Hypothesis 3 & 4: Family-owned and founder-CEO firm performance during the crisis ... 14
2.4 The use of dual-class shares and family firm performance ... 14
2.4.1 Theories and evidence on dual-class shares and family firm performance ... 14
2.4.2 Hypothesis 5: Effect of dual-class shares on firm performance ... 15
2.5 Family firms and performance during crisis-years using a United States sample ... 15
2.6 Research focus ... 15
3 Data and methodology ... 16
3.1 Sample and variables: data collection and construction ... 16
3.1.1 Sample ... 16
3.1.2 Measuring family-ownership, dual-class shares and founder-CEOs ... 16
3.1.3 Dependent variables: ROA and Tobin’s Q ... 17
3.1.4 Control Variables ... 17
3.1.5 Criteria for sample and variable construction ... 19
3.2 Descriptive statistics: summary statistics, correlations and differences of means tests . 19 3.3 Methodology ... 25
3.3.1 Performance regressions: family-owned firms and dual-class shares ... 26
3.3.2 Performance regressions using S&P 1500 subsample: founder-CEO firms ... 27
4 Regression results ... 28
4.1 Results of family-owned and dual-class share regressions using total sample ... 30
4.2 Results of dual-class regressions using family-owned firm sample ... 31
4
4.4 Founder-CEO results using family-owned firms in the S&P 1500 subsample ... 35
5 Robustness checks ... 37
5.1 Robustness checks: alternative specifications ... 37
5.2 Endogeneity of family-ownership and founder-CEO status ... 41
5.3 Tests of multicollinearity ... 43
6 Conclusion and discussion ... 44
References ... 47
5
1 Introduction
Family firms play an important role in most national economies. Even in the United States, where ownership dispersion is at its highest, approximately one-third of the largest companies are family firms (Anderson and Reeb, 2003). These family firms have distinctive characteristics and their presence affects national economies significantly (Astrachan and Shanker, 2013). The prevalence of family firms in the US and the absence of recent papers that study their effect on performance during the financial crisis make it an interesting topic.
The influence of governance on firm performance is a common subject in corporate finance research (Miller et al. 2007). In this field of research, family ownership and its impact on firm performance has been studied extensively. The expectation of the effect of family ownership on performance is ambiguous. Families often maintain large equity stakes in the firm and they could use their controlling stakes to expropriate private benefits at the expense of performance and minority shareholders (Villalonga and Amit, 2006). By contrast, family firms are assumed to be less myopic due to their long-term nature, making them less attracted to investment strategies that harm performance in the long run (Anderson and Reeb, 2003).
In related US studies, outperformance of family firms is usually found (McConaughy et al., 1998; etc.). These papers frequently include other factors, such as family management and control-enhancing mechanisms, because the performance differences are often sensitive to the way family firms are defined (Miller et al. 2007). For example, theories and evidence suggest that founder-CEOs in family-owned firms may affect performance positively by alleviating agency problems and by providing more firm-specific knowledge (Fahlenbrach, 2009). Conversely, dual-class shares are often associated with worse performance as they enhance the family’s control, thereby increasing the risk of expropriation (Masulis, 2011).
The large liquidity shock during the recent financial crisis affected many firms. As most research is conducted during relatively normal financial times, new findings during the financial crisis on the firm performance of US family-owned and founder-CEO firms relative to other firms can be meaningful. This is because theories and evidence of previous studies on this topic could be altered during a crisis. For instance, family ownership may impair performance during crisis-years because the family’s focus on firm survival may now result in survival-oriented strategies at the expense of profitable investments (Lins et al., 2013) By contrast, family-owned firms could access finance more easily using other firms under their control (Lins et al. 2013). Moreover, firms with founder-CEOs that have more firm-specific knowledge may implement better strategies than other firms during a financial crisis.
6 This research is largely comparable to previous family ownership research (Anderson and Reeb, 2003; Villalonga and Amit, 2006), However, as the crisis evidence on the effect of family ownership on performance is scarce, it contributes by using financial crisis years instead of normal financial times. Furthermore, by also examining the effect of dual-class shares and founder-CEO presence on performance during financial crisis years, this study adds new evidence to previous research that study their effect during non-crisis years.
Another contribution is that, different from the crisis study by Lins et al. (2013), the United States is used as the study location. As minority shareholders are well-protected in the US (La Porta et al., 1998), expropriation risk should be less of a concern. This could influence the effects on performance relative to other countries. Moreover, unlike Lins et al. (2013) the effects on accounting performance are also studied to make a more accurate comparison to non-crisis studies. The research question is: How do family ownership, dual-class shares and founder-CEO presence affect US firm performance before and during the financial crisis?
Using a panel dataset of 1,541 large US industrial firms, fixed effects regressions for a pre-crisis period (2009-2007) and crisis period (2008-2010) show that family-owned firms do not under- or outperform other firms in accounting (ROA) and market performance (Tobin’s Q) during non-crisis and crisis years. This is inconsistent with non-crisis studies that indicate superior performance of family firms (Anderson and Reeb, 2003; Villalonga and Amit, 2006). However, in line with existing evidence (Masulis, 2011), dual-class firms do perform worse before the financial crisis in terms ROA and Tobin’s Q. Moreover, a value discount is also found for dual-class family-owned firms when subsetting on family-owned firms. An insightful new finding is that during the financial crisis, dual-class family-owned firms underperform in terms of ROA, which confirms the financial crisis and dual-class theories.
In addition, a S&P 1500 subsample is created for studying the effect of founder-CEO presence on performance, consisting of 1,221 firms with available data for the founder-CEO variable. In the full S&P 1500 subsample regressions, founder-CEOs in family-owned firms outperform other firms during financial crisis years in terms of accounting performance. When regressions are subset on family-owned firms, family-owned founder-CEO firms also outperform other family-owned firms in ROA during the crisis. The new crisis founder-CEO results support previous non-crisis evidence (Anderson and Reeb, 2003, Fahlenbrach, 2009).
The results of this research are robust to alternative specifications such as different regression models, alternative performance measures, higher levels of winsorizing and a different crisis period. The results are also robust to addressing multicollinearity, survivorship bias and endogeneity concerns and to restricting all regressions to the S&P 1500 subsample.
7 The remainder of this study is structured as follows. Section 2 illustrates theories and evidence based upon which hypotheses are drafted. Section 3 presents the data collection, the variable construction and summary statistics. Thereafter, it explains the methodology. Section 4 provides the empirical results. Next, section 5 describes the robustness of the results and finally, section 6 concludes and discusses the main findings and implications of this study.
2 Literature review
The core of previous family firm studies usually concerns family ownership and its effect on performance. Because a large ownership stakes puts a family in a unique place to influence and control the firm, family ownership is likely to affect performance (Anderson and Reeb, 2003). Therefore, theories and evidence on the effect of family ownership on performance are discussed first. However, as confirmed by Miller et al. (2007) the way in which families affect firm performance is often sensitive to the way a family firm is defined. For instance, when family members are present in top management, for example as the firm’s CEO, this may affect performance as it amplifies the family’s control. Thus, as this study includes founder-CEO presence, theories and research on its effect on performance are discussed as well.
Furthermore, the main contribution of this research is studying family firm performance during the financial crisis. The financial crisis may influence the effect of family ownership and founder-CEO presence in several ways. Hence, the potential influence of the financial crisis is also illustrated using additional economic theories. Lastly, as mentioned by King and Santor (2008), dual-class share are frequently employed in family-owned firms and because they enhance the control through a separation of cash flow rights and voting rights, they presumably affect firm performance. For that reason, this research also examines the effect of dual-class shares. The theories and evidence on the use of dual-class shares in are discussed in this section, for both crisis and non-crisis years Based on the literature discussion of all the abovementioned features of this study, five main hypotheses are drafted.
2.1 Performance of family-owned firms
Multiple theories try to explain why family-owned firms might perform differently from non-family-owned firms. To clearly illustrate why non-family-owned firms may underperform or outperform relative to other firms, or why there might be no differences in performance, theories and previous evidence on the costs and benefits of family ownership are explained next. Based on these explanations, the first hypothesis of this study is drafted.
8 2.1.1 Theories and evidence on costs of family ownership
The ownership of firms by families bears several potential costs according to theories that describe why family-owned firms may be inferior performers. For example, it is frequently assumed that large ownership stakes could lead to families benefitting themselves at the expense of firm performance, because their strategy preferences differ from those of other shareholders (Anderson and Reeb, 2003). This statement is based on Fama and Jensen (1985), who suggest that shareholders with a high level of cash-flow rights may make different decisions regarding the firm’s investments than small shareholders. In order to maximize the utility of these family-owners, the firm needs to employ suboptimal strategies, which eventually affects the performance of the firm negatively (Burkarkt et al., 1997).
Another potential cost of family ownership is that it could create agency problems. A large family ownership stake can put the family in a controlling position, which they may use to extract firm resources at the expense of minority shareholders (Villalonga and Amit, 2006). Anderson and Reeb (2003) state that this expropriation of wealth could be done through, for example, special dividends. Extracting private benefits is not in the best interest of other shareholders or profitability, suggesting negative effects on performance of family ownership. Furthermore, the presence of concentrated family ownership tends to reduce the biddings by other investors, as it is more difficult to capture control of the firm (Barclay and Holderness, 1989). Moreover, the large ownership stakes enables the families to appoint management positions, which can impose additional difficulties on investors in gaining control of the firm, potentially resulting in lower valuations (Anderson and Reeb, 2003). According to these explanations, large family ownership could be detrimental to firm value.
Finally, family-owners might be too focused on firm survival. This is because families tend to have undiversified holdings, making them more risk averse (King and Santor, 2008). This could lead family-owned firms to refrain from investing in profitable opportunities such as mergers or acquisitions, which could be profitable in the long run. The family’s carefulness in implementing investment strategies can therefore be interpreted as a cost, as it is likely to harm the long-term accounting and market performance of the family-owned firms.
The costs of family ownership have been shown empirically. However, not in the most recent studies in the United States. For example, underperformance of family-owned US firms has been shown in less recent papers, such as Holderness and Sheehan (1988). In Europe and Asia evidence on underperformance of family firms is less scarce (Claessens et al., 2002; Cronqvist and Nilsson, 2003). There appears to be variation in the answer to the question
9 whether family-owned firms in fact underperform relative to non-family firms, largely depending on the location and period of the study. Therefore, the importance of the location and the sample period of this study is discussed later in more detail.
2.1.2 Theories and evidence on benefits of family ownership
By contrast, family ownership could also offer potential benefits to the firm and its shareholders, which would suggest that family ownership is an effective organizational structure. A first potential benefit is that large shareholders such as family-owners have significant incentives to maximize firm value by reducing agency problems (Demsetz and Lehn, 1985). This is illustrated further by Villalonga and Amit (2006), who state that agency problems could be less common in family-owned firms, as there are substantial incentives for families to monitor the managers. Due to their large and often undiversified ownership stake, the wealth of families is closely related to the performance of the firm, thereby increasing the families’ incentives to monitor (Anderson and Reeb, 2003). Moreover, families with large ownership stakes or controlling positions are more capable of monitoring managers, even when they are not part of management, as shown by Shleifer and Vishny (1986). Additional monitoring should enhance the performance of family-owned firms relative to other firms.
In addition, families usually maintain ownership stakes for a long period (Anderson and Reeb, 2003). Therefore, families may have longer investment horizons, making them less attracted to myopic strategies focused on short-term earnings at the expense of profitable long-term investments (Stein, 1988). Firm survival is more important for families than wealth consuming, as families view the firm as a life-time valuable asset (Casson, 1999). The long-termism of family-owners could therefore explain outperformance of family-owned firms.
Another theory regarding the benefits of family ownership is concerning the reputation of families. The long-term nature of families results in longstanding relationships with capital providers, who will deal with the same persons for a long period of time (Anderson and Reeb, 2003). These relationships might be a competitive advantage with respect to non-family firms, where owners and directors turn over more frequently. This is because it could result in lower costs of debt for family-owned firms relative to non-family-owned firms, affecting their possible strategies and therefore performance positively (Anderson and Reeb, 2003).
As opposed to evidence regarding the costs of family ownership, evidence on the benefits of family ownership has been found in multiple, more recent US studies (McConaughy et al ,1998; Anderson and Reeb, 2003; Villalonga and Amit, 2006). According
10 to the evidence presented in these papers, family ownership measured by the equity stake of the family is positively related to performance, both in terms of market performance (Tobin’s Q) and accounting performance (ROA). However, as mentioned in Miller et al. (2007), the additional results of these studies show that findings regarding performance are sensitive to the way family firms are defined. Therefore, the effects of other characteristics of family firms, such as the use of dual-class and founder-CEO presence, are also included in this study.
2.1.3 Theories and evidence on endogeneity of family ownership
In contrast with the potential effects described above, it is also possible that family ownership has no significant effect on performance. This is described by Demsetz and Lehn (1985), who suggest that there is endogeneity between structures of ownership and performance. King and Santor (2008) elaborate by suggesting that efficient markets will eventually result in optimal ownership structures given the characteristics of firms and industries. Furthermore, they suggest that firms with inefficient ownership characteristics would not survive in the long run, indicating that there should be no effect of family ownership on performance.
There is evidence regarding endogeneity of concentrated ownership, such as family-ownership. Demsetz and Lehn (1985), do not find a significant effect of ownership on market performance, which is in line with the endogeneity-theory. In addition, Himmelberg et al. (1999) do not find a relationship between ownership and firm performance. They ascribe the deviating results in other studies to the failure of addressing endogeneity.
2.1.4 Hypothesis 1: Performance of family-owned firms
There are opposing theories concerning the costs and benefits of family-owned firms. The theories could explain both outperformance and underperformance of family-owned firms. However, due to the majority of evidence regarding US family firm performance that indicates a positive relationship, the expectation is that family-ownership affects accounting and market performance positively during pre-crisis years. Therefore, the first hypothesis is:
Hypothesis 1: Family-owned firms outperform non-family-owned firms in non-crisis years.
2.2 Founder-CEOs and firm performance
Family firms are often not only characterized by the family’s ownership stake in the firm, as in many cases the founder remains present in the firm as CEO or as other top management (Anderson and Reeb, 2003). Fahlenbrach (2009) explains the frequency of founder-CEO
11 presence by stating that they are less likely to be fired than non-founder-CEOs. Founder-CEO presence is often combined with large equity stakes, which enables the founder and its family to gain additional influence and control on the strategy implementation and decision-making in family firms (Fahlenbrach, 2009). The influence of the founders could potentially benefit or impair firm performance according to economic theories. Therefore, founder-CEO presence is included in this research as a main focus, by studying whether the presence of the founder as CEO affects firm performance. Theories and existing evidence are described next.
2.2.1 Theories and evidence on founder-CEOs and firm performance
Several theories try to explain a negative effect of family management on performance. For example, when family-owners stay on as managers for a long time, even though they are no longer capable of running the firm, this can create significant costs (Shleifer and Vishny, 1997). Morck et al. (1988) describe that this could happen because family managers with enough voting rights due to their large ownership stake can secure their position in the firm. If this occurs, unqualified founder-CEOs with non-value maximizing incentives may remain in their position, which affects family firm performance negatively. Another main concern with family-CEOs and management is that by limiting management positions to family members, the firm does not appoint professional managers that could be more skilled and also more productive (Anderson and Reeb, 2003). Thus, if founder-CEOs are indeed worse performers than hired-CEOs, this would negatively affect performance (Villalonga and Amit, 2006).
By contrast, there are also multiple theories that describe positive effects of founder or family top management on firm performance. Firstly, owner-manager relationships involving families should reduce agency costs (Fama and Jensen, 1983). When the family-owners and the managers are the same persons, there is less need for expensive monitoring as the interest of the owners and the managers are aligned. This reduces agency problems and could therefore increase performance (McConaughy et al., 1998; Jensen and Meckling, 1976). Moreover, founder-CEO present in family firms may bring more firm-specific knowledge and additional innovative and value-increasing abilities to the firm (Morck et al. 1988). Lastly, Anderson and Reeb (2003) suggest that founders often identify with their company and its performance, which creates a substantial incentive for them to improve firm performance.
Existing evidence regarding founder and family management is slightly mixed. Less recent studies such as Yermack et al (1996), find that firms with founder-CEOs suffer from lower valuations, measured by Tobin’s Q. However, Morck et al (1998), show that when a
12 founder or member of the founding family is present as one of the two top officers, firm value increases. Moreover, more recent US papers (Anderson and Reeb, 2003; Fahlenbrach, 2009), also find positive relationships between founder-CEOs and performance. Anderson and Reeb (2003) show higher accounting performance for firms with founder- and family-CEOs, but find that only founder or hired-CEOs add firm value. Villalonga and Amit’s (2006) results from the Fortune 500 are mostly comparable, as they find that family management increases value when the founder is CEO and decreases firm value when the descendant is CEO. Lastly, the evidence of Adams et al. (2009) also implies that the positive relationship between firm value and family firms is caused by founder-CEO presence.
2.2.2 Hypothesis 2: Performance of firms with founder-CEOs
There are again opposing theories, which can explain both a positive and a negative effect of founder-CEO presence on firm performance. However, the majority of US studies concerning firms with founder and family management show a positive effect on firm performance, especially regarding founder-CEOs. This introduces the second hypothesis of this study:
Hypothesis 2: The presence of founder-CEOs positively affects firm performance.
2.3 Performance of family firms during the financial crisis
The theories and evidence illustrated above concern the effects of family ownership and founder-CEOs on performance during relatively normal financial times. However, in times of financial difficulty such as the recent financial crisis, firms suffer from a large liquidity shock, which may amplify both the abovementioned benefits and costs of family-owned and founder-CEO firms (Lins et al, 2013). In addition, several new theories could also explain why expectations regarding possible out- or underperformance of family firms during a financial crisis might differ from those in normal times. The theories are described below.
2.3.1 Theories and evidence on family firm performance during the financial crisis
Firstly, a theory that tries to explain why family firms may perform worse than non-family firms during a crisis is based on the focus on firm survival of family firms (Lins et al, 2013). Families often have undiversified holdings, resulting in strategies based on firm survival at the expense of profitable investments and shareholders, indicating worse performance for family firms during the crisis. Moreover, the risk of expropriation by family firms (Anderson and Reeb, 2003) might be larger, since managers may want to compensate for the plausible
13 decrease in compensation during a crisis. This could be done using the firm’s resources, which also suggests lower performance of family and founder-CEO firms (Lins et al., 2013).
Another theory that suggests underperformance of family firms during the crisis is based on Shleifer and Vishny (1997), who state that unqualified family managers often stay on too long in family firms. During a financial crisis, good management becomes even more important, as there are substantial challenges that firms have to tackle. Family firms with impaired founder-CEOs could therefore result in underperformance during a financial crisis.
By contrast, theories that explain why family firms may outperform during a financial crisis are based on families’ potential financing advantages. During financial crises, firms cannot access finance easily. Family firms often maintain long-standing relationships with banks and could therefore have advantages in accessing finance over non-family firms if they have a good reputation (Anderson and Reeb, 2003). Moreover, Lins et al (2013) state that in a crisis family could possibly access finance more easily via other firms under their control. These notions suggest that family firms outperform non-family firms during a crisis, as their financing advantages makes them less restrained in investing in profitable opportunities.
In addition, the abovementioned theory of Stein (1988) stating that family managers have longer investment horizons and are less attracted to myopic strategies also suggests that family firms would outperform in a crisis. This is because these less myopic founder-CEOs, who tend to have more firm-specific knowledge (Morck et al. 1988), might be able to implement better strategies in difficult times than non-family firms. Furthermore, when a family-owner is also CEO, this aligns the interest interests of the managers and the owners (Anderson and Reeb, 2003). During a crisis, when a firm is doing bad, non-family-owner top-management do not bear the costs when the firm fails. Hence, hired CEOs are more likely than family-owner CEOs to take on investments with a higher risk of default in order to save the firm. This suggests that family-owned firms outperform during a financial crisis.
Research and evidence on family firm performance during the financial is scarce. However, Lins et al. (2013) executed a similar research to this one, namely for 45 countries excluding the United Sates. Their main results show that family-controlled firms underperform on the market relative to other firms. They also find that family-controlled firm cut investments more and that these cuts are associated with larger underperformance. By contrast, Kashmiri and Mahajan (2014) study family-owned firm performance using 275 large listed firms during business contractions and find that family firms outperform non-family firms, which is in accordance with the existing evidence in normal financial times.
14 2.3.2 Hypothesis 3 & 4: Family-owned and founder-CEO firm performance during the crisis The opposing theories could justify both a positive and negative influence of the crisis on family-owned firm performance. In addition, mixed and little evidence during crisis-years is available. However, a substantial amount of theories and evidence suggest outperformance of US family firms in normal times. This, combined with theories suggesting outperformance in a crisis, results in the expectation that family-owned firms outperform other firms during the financial crisis, especially when the founder is CEO. Therefore, the two crisis-hypotheses are:
Hypothesis 3: Family-owned firms outperform during the 2008-2009 financial crisis Hypothesis 4: Founder-CEO presence affects firm performance positively during the crisis
2.4 The use of dual-class shares and family firm performance
In many studies related to family firms, the use of control-enhancing mechanisms, especially dual-class shares, is taken into account (King and Santor, 2008; Anderson et al, 2009; Masulis, 2011). This is because the level of control of families is enhanced by the use of dual-class shares, as they create a clear wedge between the cash-flow rights and voting rights (La Porta et al., 1999). King and Santor (2008) mention that studies often show that control-enhancing mechanisms are used the most in family-owned firms. Because dual-class shares are a clear measure of whether a wedge between voting and cash-flow rights is present, which magnifies a family’s control, its presumable effect on performance is examined in this study.
2.4.1 Theories and evidence on dual-class shares and family firm performance
Theories on whether dual-class shares affect firm performance are mostly based on the fact that the additional control they generate could increase the risk of expropriation by, for example, controlling families (Masulis, 2011). If the additional control is used to extract private benefits, this would suggest a negative relationship between dual-class shares and firm performance (Gompers et al., 2010). Conversely, Anderson et al. (2009) explain that control-enhancing mechanisms such as dual-class shares could also be used for monitoring purposes. The additional control gained through the use of dual-class shares grants the family more power to monitor the managers, which in that case indicates a positive relationship between the control-enhancing dual-class shares and firm performance.
The evidence concerning the use of dual-class shares in firms usually indicate worse performance relative to other firms. For example, Villalonga and Amit (2006) find that family firms create the most value when control-enhancing mechanisms such as dual-class shares are
15 not employed. Moreover, King and Santor (2008) find that family firms with dual-class shares are valued 17% lower relative to other firms, although accounting performance (ROA) is comparable. Masulis (2011) also finds that family firms with wedges between cash-flow and voting rights suffer from worse performance. Lastly, another paper by Villalonga and Amit (2009) shows that dual-class stock negatively affects firm value of large US firms.
2.4.2 Hypothesis 5: Effect of dual-class shares on firm performance
According to the majority of theories and evidence that suggest a negative relationship between control-enhancing mechanisms and firm performance, the expectation is that using dual-class shares reduces firm performance, in both non-crisis and crisis years. This is represented by the last hypothesis that is tested in this study:
Hypothesis 5: Dual-class shares negatively affect performance in crisis and non-crisis years
2.5 Family firms and performance during crisis-years using a United States sample Although this research is similar to Lins et al. (2013), it contributes, apart from also studying founder-CEO presence and dual-class shares in the crisis, by using United States as the location. There are several reasons why studying performance of US family firms during a financial crisis is of significance and why it may show different results than Lins et al. (2013). The main reason the US is different is that the US regulatory environment is created so that minority shareholders are well-protected (Doidge et al. 2004; La Porta et al., 1998). Thus, expropriation risk and potential private benefits of control are limited in the US, indicating that the market performance of US family firms should be better than those of family firms in other countries. This is illustrated by the European and Asian studies that show mixed or negative effects of family-ownership, while US research usually finds a positive effect. Moreover, Gadhoum et al. (2005), state that the US has the highest level of family-controlled firms, which implies that there might be a difference in the market’s view on family firms between US and other countries, potentially resulting in deviating market performance.
2.6 Research focus
All in all, the main research focus is to examine performance of US family-owned and founder-CEO firms and in what way the crisis influences their performance relative to other firms. It can be summarized using four sub-questions. First, what is the effect on performance
16 of concentrated family-ownership in US firms? Second, what is the effect on performance of founder-CEOs? Third, do family firms and founder-CEO firms outperform or underperform other firms during the financial crisis? Finally, how does the use of dual-class shares in family firms affect performance during both crisis and non-crisis years?
3 Data and methodology
3.1 Sample and variables: data collection and construction
First, this section describes details of the sample of this study, such as the selection, period, location and sample size. Next, an explanation is given of how family-ownership, dual-class shares and presence of founder-CEOs are measured. Thereafter, the data collection and construction of the control variables are illustrated. A detailed description of all the variables included in the multivariate analysis is presented in table A1 of the appendix. Finally, the additional criteria used for the sample and the variables are presented in this section.
3.1.1 Sample
The sample used in this research is based on the one of Anderson et al. (2009), which is updated until 2012. Their dataset contains information on whether firms are family-owned, measured by an equity stake greater than or equal to 5%. It consists of the 2000 largest US industrial firms, based on total assets. In this research, the years 2006 up to and including 2010 are used for pre-crisis and crisis years. After applying additional criteria for the total sample, which are described below, the full sample consists of 1,541 US industrial firms for which ownership and financial data are available, yielding 6,722 firm-years or observations.
3.1.2 Measuring family-ownership, dual-class shares and founder-CEOs
Family ownership is measured in the same way as in Anderson et al. (2009) in this research. Namely, when a family has an equity stake in the company greater than or equal to 5%, the firm is family-owned. This is indicated by a dummy variable, which takes the value of one when the firm is family-owned and zero otherwise, for all firm-years in the sample period. Additionally, the dataset of Anderson et al (2009) contains data on whether a firm uses dual-class shares, which is also indicated by a dummy that is equal to one when dual-dual-class shares are employed in the firm and zero otherwise, for all firm-years in the sample period.
17 Furthermore, data are collected for founder-CEO presence. In order to do this without having to collect data manually, the BoardEx database is used. The BoardEx database covers firms listed on the S&P 1500 firms, starting from 2005. BoardEx contains data on the role of directors of the firms and it therefore indicates whether a director is CEO of the firm. Moreover, it is mentioned in a separate column whether this director is also the founder. Using this data, an indicator variable is coded in Stata, which equals one if the firm’s current CEO is also the founder. Next, this dataset is merged with the full dataset. To be able to match the years during which the CEOs was present with the corresponding sample years in the existing dataset, the “start” and “end” date provided with the BoardEx dataset are used to derive the period the CEO was present. All CEOs with start date after 2010 or end date before 2006 are dropped from the sample. This period is matched with the fiscal years in the full dataset. After the merge, however, not all observations in the full dataset have available data for the CEO variable. As a result, a smaller sample is created for running the founder-CEO regressions of this study. This smaller sample is called the S&P1500 subsample, which consists of 1,221 firms, yielding 4,002 firm-years or observations.
3.1.3 Dependent variables: ROA and Tobin’s Q
The two dependent variables are return on assets (ROA) and Tobin’s Q, which measure accounting performance and market performance, respectively. For all firms in the sample, data for the dependent variables are collected using the Compustat North America database of WRDS. The main measures of ROA and Tobin’s Q are similar to those of Anderson and Reeb (2003). ROA is obtained by dividing net income by the book value of total assets. Tobin’s Q is calculated by dividing the market value of total assets (total book value of debt plus the market value of equity) by the book value of total assets. Alternative measures of ROA and Tobin’s Q are included in the robustness checks, which are ROA based on EBITDA and industry-adjusted Tobin’s Q. These alternative measures are described in appendix table A1.
3.1.4 Control Variables
Several important control variables are included in this study, to control for firm- and industry-specific characteristics that could possibly affect family ownership and performance. These control variables are mainly based on those included by the highly regarded family-owned firm performance study of Anderson and Reeb (2003), in order make a correct and insightful comparison between the findings of this research and theirs. Additionally, as
18 Villalonga and Amit (2006) is an important and comparable study executed during relatively normal financial times, additional control variables of their study will be incorporated as well.
Important control variables of Anderson and Reeb (2003) are firm size, growth opportunities, firm risk, capital structure and firm age. These variables are also standard features of many other well-regarded family firm papers such as McConaughy et al. (1998) and Villalonga and Amit (2006) and therefore, these variables are incorporated. The inclusion of these variables ensures that this research controls for factors that might influence family ownership and performance. In addition, sales growth, investment intensity, dividends and industry are also added as controls, based on Villalonga and Amit (2006). Sales growth and investment intensity are likely to affect performance, as well as high dividend payout rates by limiting the expropriation risk, as dividend payouts remove wealth from family-owners (Villalonga and Amit, 2006). Moreover, as family firms are not uniformly distributed across industries, it is important to control for industry effects (Villalonga and Amit, 2006).
The abovementioned control variables are constructed based on their respective studies, Anderson and Reeb (2003) and Villalonga and Amit (2006), unless stated otherwise. Firstly, firm size is included as the logarithm of the book value of total assets. The growth opportunities variable is measured as research and development (R&D) expenses scaled by the book value of total assets. Comparable to Villalonga (2004) and Miller et al. (2007), missing R&D expense data are replaced with zero. Firm risk is measured similar to Anderson and Reeb (2003), namely by the standard deviation of monthly stock returns. This study uses the previous 24 months for the standard deviation calculation. The capital structure control is the leverage ratio, which is total long-term debt divided by the book value of total assets. This study measures firm age as the logarithm of the amount of active years since the firm’s first fiscal year of available accounting data reported by the “names” database from Compustat.
Furthermore, the control for investment intensity is defined as capital expenditures divided by book value of total assets (King and Santor, 2008; Maury; 2006). Growth of net sales is measured by the average growth rate of net sales over the last two fiscal years, which is a similar measure as in King and Santor (2008). Dividends is defined as total dividends paid divided by the total book value of equity. Lastly, this study controls for industry by controlling for firm fixed effects in the regressions, which includes the industry effects.
The data needed for constructing all the control variables are collected through the Compustat North America database of WRDS, except for firm risk. For the firm risk variable, 24 months of historical monthly stock returns are collected for each firm-year using “Security Monthly” from the CRSP/Compustat merged database of WRDS. The returns are used to
19 calculate the standard deviation of the returns over the 24 months prior to the fiscal year. A more detailed description of all control variables is presented in table A1 of the appendix.
3.1.5 Criteria for sample and variable construction
In order to create a dataset that is suitable for running all the regressions, additional criteria for the sample and the variables are applied. Firstly, financial firms and regulated public utilities are excluded, due to potential difficulties in calculating Tobin’s Q for financial firms and because government regulations could affect accounting performance of the firms. This is done frequently in studies on family firm performance (Anderson and Reeb, 2003; Villalonga and Amit, 2006; etc.). In addition, only firms with available data for every consecutive year they appear in the sample are included. Moreover, all duplicates are dropped from the sample. Furthermore, after all variables are generated, firm-years before 2006 and after 2010 are removed from the sample. Observations with missing values for one of the variables used in the regressions are also dropped from the sample to make sure that all variables have the same amount of observations. Lastly, all non-binary variables are winsorized at the 1st and 99th percentiles. This is comparable to previous studies, which winsorize the variables in order to reduce the weight of extreme values (Maury, 2006; Lins et al., 2013). The firm size and firm age variables are winsorized using their non-logarithm values.
3.2 Descriptive statistics: summary statistics, correlations and differences of means tests This section provides descriptive statistics of the main dependent, independent and control variables. Table 1 shows summary statistics of all variables. Table 2 presents outcomes of the difference of means tests for all variables between family-owned firms and non-family owned firms. Finally, Table 3 provides a correlation matrix, including all variables.
Table 1 presents the summary statistics of the variables. The table shows that 30.5 percent of the firms in the sample are family-owned. This percentage is roughly comparable to that of Anderson and Reeb (2003), who find that 35% of S&P 500 firms are family firms. However, they measure family firms as S&P 500 firms in which families have an equity ownership stake and also when families have board seats. The table also shows that in 8.1 percent of the firms in the smaller S&P 1500 subsample, a founder-CEO is present. In addition, in 9.8 percent of all firms, dual-class shares are used. The average firm in the sample has a Tobin’ Q of 1.68. Moreover, firms on average have a return on assets (ROA) of 1.8 percent, with a median of 4.5 percent. This ROA is lower than in the Anderson and Reeb
20 (2003), which could be explained by worse performance during the financial crisis, as many firms were negatively affected by the difficult financial situation in the sample period.
Table 2 tests the difference in means for all variables. Comparable to Anderson and Reeb (2003), the difference of means tests use the time-series averages for each firm. The variables are measured in the same way as in table 1, apart from firm size and firm age Instead of using the logarithms, firm size is defined as the total value of total assets in millions ($) and firm age is defined as the amount of years since the first fiscal year of available accounting data. Difference of means between family-owned and non-family-owned are tested for a total, a pre-crisis and a crisis period. Lins et al. (2013) use 2006-2007 as non-crisis years and 2008-2009 as financial non-crisis years. This research uses comparable periods, although 2010 is also used in the crisis period. Boria (2014) shows that the contraction phase of the financial cycle, which is closely associated with the financial crisis, lasts multiple years. Thus, firm performance is likely to be affected by the crisis for several years and 2010 is therefore also included in the crisis period. Later, in the robustness checks, the same crisis period as in Lins et al (2013) is employed, to check whether comparable results are found.
Table 2 shows that a founder-CEO is present in 15.5 percent of the family firms in the S&P 1500 subsample during the total period. This is comparable to the 14.5 percent found by Anderson and Reeb (2003). However, they also measure family firms when a family has board seats and therefore do not find founder-CEOs presence in non-family firms. By contrast, this study also finds founder-CEO presence in 4.5 percent of non-family-owned firms, as the family’s ownership stake is used as the only measure. The table also shows that family-owned firms use dual-class shares to a much larger extent, represented by highly significant mean differences, which is consistent with King and Santor (2008), who state that control-enhancing mechanisms are more likely to be used by family-owned firms.
In addition, Table 2 shows that there is no significant difference in the accounting performance measure (ROA) between family-owned and non-family-owned firms in the total period and neither in the pre-crisis and crisis periods. The ROA decreased for both family- and non-family-owned firms during the crisis period. ROA also appears to be higher for family-owned firms relative to other firms during all periods, although the differences of the means are not significant. By contrast, there is a significant difference in market performance between family-owned firms and other firms. Tobin’s Q is smaller for family-owned firms in all three periods and decreases substantially for all firms in crisis years. The significantly lower valuations for family-owned firms contrasts with Anderson and Reeb (2003), who show that family firms have a significantly higher Tobin’s Q than non-family firms.
21 Table 1: Summary Statistics
Variable Mean Median SD p25 p50 Min Max N
Family-owned 0.305 0 0.460 0 1 0 1 6,722
Founder-CEO 0.081 0 0.273 0 0 0 1 4,002
Dual-Class 0.098 0 0.297 0 0 0 1 6,722
Tobin’s Q 1.683 1.444 0.830 1.132 1.975 0.665 5.314 6,722
ROA (Net Income) 0.018 0.045 0.135 -0.001 0.082 -0.661 0.277 6,722
Firm Size (log) 7.431 7.315 1.520 6.379 8.356 1.992 13.590 6,722
Growth Opportunities 0.032 0 0.058 0 0.038 0 0.304 6,722
Firm Risk 0.105 0.093 0.054 0.067 0.130 0.030 0.312 6,722
Capital Structure 0.204 0.176 0.188 0.029 0.306 0 0.742 6,722
Firm Age (log) 3.140 3.045 0.584 2.639 3.714 1.792 4.094 6,722
Investment Intensity 0.047 0.031 0.048 0.017 0.057 0.002 0.278 6,722
Sales Growth 0.069 0.052 0.246 -0.049 0.154 -0.564 1.141 6,722
Dividends 0.029 0 0.064 0 0.034 0 0.440 6,722
This table provides the summary statistics of dependent and independent variables used in the performance regressions. It shows summary statistics of all variables, consisting of the means, medians, standard deviations, maximum and minimum values, 25 and 75 percentiles and the number of observations. The full sample is composed of 6,722 observations of 1,574 large US industrial firms. The S&P 1500 subsample is composed of 1,221 firms and contains 4,002 observations that have data available for the founder-CEO variable. Family ownership, the main independent variable, is a dummy variable with the value of 1 if the firm is family-owned. Dual-class shares use is also indicated with a dummy, which takes the value of 1 when dual-class shares are employed in the firm, and 0 otherwise. The next two are the dependent variables, Tobin’s Q and ROA. Tobin’s Q is calculated by dividing market value of total assets (total book value of debt plus the market value of equity) b y the book value of total assets. ROA (Net Income) is obtained by dividing net income by the book value of total assets. Next, the descriptive variables for the control variables are given. Firm size is included as the logarithm of the book value of total assets. The variable for growth opportunities is measured as research and development expenses scaled by the book value of total assets. In addition, firm risk is measured by the standard deviation of monthly stock returns over the prior 24 months. Capital structure is the leverage ratio, measured by total long-term debt divided by the book value of total assets. Firm age is measured as the logarithm of the amount of years since the first fiscal year of available accounting data. Furthermore, investment intensity is measured as capital expenditures divided by book value of total assets. Growth of net sales is measured by the average growth rate of net sales over the last two fiscal years. Lastly, dividends represent dividends paid by firms, defined as total dividends paid scaled by the total book value of equity.
22 Table 2: Difference of means tests
Total sample period (2006-2010) Pre-crisis period (2006-2007) Crisis period (2008-2010) Family-owned Non-family- owned Family vs. non-family Family-owned Non-family- owned Family vs. non-family Family-owned Non-family- owned Family vs. non-family
Variable Means Means t-statistic Means Means t-statistic Means Means t-statistic
Family Ownership 0.305 0.695 0.320 0.680 0.293 0.707
Founder-CEO 0.155 0.045 12.27*** 0.159 0.049 7.86*** 0.153 0.042 9.42***
Dual-Class 0.288 0.014 38.80*** 0.271 0.015 23.88*** 0.304 0.013 30.84***
Tobin’s Q 1.568 1.734 -8.61*** 1.714 1.917 -6.05*** 1.454 1.594 -5.59***
ROA 0.021 0.017 1.50 0.038 0.032 1.30 0.009 0.005 1.02
Firm Size ($ millions) 4,258 6,484 -6.19*** 3,925 5,996 -4.08*** 4,545 6,841 -4,51*** Growth Opportunities 0.018 0.039 -13.82*** 0.019 0.039 -9.05*** 0.018 0.038 -10.08***
Firm Risk 0.104 0.106 -1.06 0.095 0.099 -1.89* 0.112 0.111 -0.48
Capital Structure 0.193 0.209 -3.35*** 0.190 0.206 -2.28** 0.197 0.211 -2.15**
Firm Age (years) 25.275 28.323 -7.24** 23.617 26.703 -4.92*** 26.576 29.568 5.26***
Investment Intensity 0.049 0.045 3.51*** 0.054 0.049 2.19** 0.046 0.042 2.33**
Sales Growth 0.056 0.074 -4.96*** 0.109 0.132 -2.76*** 0.016 0.030 -2.84***
Dividends 0.031 0.029 1.66* 0.029 0.028 0.14 0.032 0.029 1.90*
This table presents the difference of means test between family-owned and non-family-owned firms for the dependent, explanatory and control variables used in the multivariate analysis. Performance is measured by Tobin’s Q, which is the market value of total assets divided by the book value of total assets, and by ROA, which is net income divided by total assets. In this table, firm size is defined as the total value of total assets in millions ($), instead of using its logarithm. Growth opportunities is R&D expenses scaled by total assets. Firm risk is measured by the standard deviation of monthly stock returns over the prior 24 months. Capital structure is total long-term debt divided by total assets. Moreover, firm age is the amount of years since the first fiscal year of available accounting data, also not expressed as a logarithm. Investment intensity is capital expenditures divided by total assets. Growth of net sales is the average growth rate of net sales over the last two years. Lastly, dividends is defined as total dividends paid scaled by the total book value of equity. The means of the variables are presented for both family-owned and non-family-owned firms. The difference of means for the variables are tested for three periods. The first period is the total sample period, which covers the years 2006-2010. The second period is the pre-crisis period, consisting of the years 2006 and 2007. The third period concerns the crisis period used in this study, which are the years 2008, 2009 and 2010. Comparable to Anderson and Reeb (2003), the test of difference of means use the time-series averages for each firm. The t-statistics of the tests are presented in the third column of each period. The asterisks, namely *, **, and *** denote statistical significance at the 0.10, 0.05, and 0.01 significance levels, respectively.
23 Table 3: Correlation Matrix
Variable Family- Owned Founder-CEO Dual-Class Tobin’s Q ROA Firm Size Growth Opp. Firm Risk Capital Struct. Firm Age Investm. Intensity Sales Growth Dividen ds Family-Owned 1 Founder-CEO 0.191 1 Dual-Class 0.428*** 0.062*** 1 Tobin’s Q -0.090*** 0.032** -0.096*** 1 ROA 0.016 -0.030* -0.057*** 0.243*** 1
Firm Size (log) -0.142*** -0.085*** -0.016 0.043*** 0.267*** 1
Growth Opp. -0.160*** 0.089*** -0.098*** 0.280*** -0.291*** -0.221*** 1
Firm Risk -0.018 0.042*** -0.028** -0.046*** -0.168*** -0.337*** 0.172*** 1
Capital Struct. -0.037*** -0.012 0.088*** -0.084*** -0.153*** 0.149*** -0.147*** 0.018 1
Firm Age (log) -0.074*** -0.211*** -0.018 -0.078*** 0.141*** 0.379*** -0.190*** -0.212*** -0.026** 1
Investm. Intensity 0.040*** 0.041*** -0.055*** 0.019 0.097*** 0.084*** -0.180*** -0.022* 0.076*** -0.007* 1
Sales Growth -0.027** 0.063*** -0.048*** 0.230*** 0.192*** 0.070*** 0.060*** 0.053*** 0.000 -0.114 0.093*** 1
Dividends 0.014 -0.059*** 0.013 0.198*** 0.143*** 0.176*** -0.084*** -0.190*** 0.065*** 0.215*** -0.007 -0.058*** 1
This table presents the correlation between the variables used in the multivariateanalysis for all 6,722 observations. Family ownership, is a dummy variable which takes the value of 1 if the firm is family-owned, and 0 otherwise. Founder-CEO is a dummy that is equal to 1 if the founder is CEO of the firm and 0 otherwise. Dual-class shares use is also indicated with a dummy, which takes the value of 1 when dual-class shares are employed in the firm, and 0 otherwise. Tobin’s Q is market value of total assets divided by the book value of total assets. ROA (Net Income) is net income divided by the book value of total assets. Firm size is the logarithm of the book value of total assets. Growth opportunities is measured as research and development expenses scaled by the book value of total assets. Firm risk is measured by the standard deviation of monthly stock returns over the prior 24 months. Capital structure is total long-term debt divided by the book value of total assets. Firm age is measured as the logarithm of the amount of years since the first fiscal year of available accounting data. Investment intensity is capital expenditures divided by book value of total assets. Growth of net sales is measured by the average growth rate of net sales over the last two fiscal years. Dividends is defined as total dividends paid scaled by the total book value of equity. The asterisks *, **, and *** denote statistical significance at the 0.10, 0.05, and 0.01 significance levels, respectively. Significance is measured using Stata’s pwcorr function.
24 Table 2 also shows significant differences in the means of nearly all control variables. Similar to Anderson and Reeb (2003), family-owned are significantly smaller than non-family-owned firms in terms of total assets. Family firms also differ in terms of the capital structure, measured by a significantly lower leverage ratio during all three periods. This is in accordance with the theory of Shleifer and Vishny (1986), who explain that large shareholders might have risk aversion, due to their frequent undiversified holdings. Large family-owners may therefore prefer raising finance that has a lower default probability, which could explain the lower leverage ratio of family-owned firms (Anderson and Reeb, 2003).
Moreover, the results in table 2 indicate that family-owned firms have less growth opportunities and lower sales growth in all periods, which is likely to affect firm performance. According to the firm age measure, family-owned firms are also less old, which agrees with firm ages measures in other studies (Anderson and Reeb, 2003; Villalonga and Amit, 2006). The table shows that in the total period and during the crisis, family firms pay more dividends. In addition, according to table 2, family-owned firms appear to be less risky during the pre-crisis years, as there is a small, but significant difference in firm risk for this period only. The family-owned firms also appear to invest more than non-family-owned firms in all three periods. The differences in variables emphasize the need for an examination using a multivariate analysis, in order to determine how the variables and their differences contribute to performance differences between family-owned firms and non-family-owned firms.
Lastly, table 3 shows the correlations between the variables for the full sample, which consists of 6,722 observations. It also reports the correlations for the 4,002 observations of the founder-CEO variable. Nearly all variables in the table are significantly correlated, although most of the correlations are not strong. Overall, there are no worrisome correlations among the independent variables that could indicate potential multicollinearity threats. This notion is confirmed in the robustness checks. Furthermore, appendix Table A2 shows the correlations of variables for the 4,002 observations in the S&P 1500 subsample used for the founder-CEO regressions. The correlations are largely comparable in terms of coefficients and significance.
Firstly, the family-owned variable does not appear to be strongly related with most explanatory and control variables. However, family-ownership is negatively correlated with firm size, comparable to Anderson and Reeb (2003), and also with growth opportunities. Family ownership is also highly correlated with dual-class. This is because almost all firms that use dual-class shares are family-owned firms. The founder-CEO variable is also not strongly correlated with independent and control variables, apart from its negative association with firm age. This is logical as older firms are less probable to have founder-CEOs present.
25 Other important correlations are the following. Family-ownership has a significant and negative relation with market performance, but this correlation is small. By contrast, family-ownership has a negative but insignificant correlation with accounting performance. The opposite is shown for founder-CEO, which has a negative and small association with ROA and positive and small association with Tobin’s Q. Dual class is negatively correlated with Tobin’s Q and ROA, which corresponds with the notion that using dual-class shares increases expropriation risk, which may negatively affect accounting and market performance.
Correlations in table 3 that are also noteworthy are those between Tobin’s Q and growth opportunities, sales growth and dividends, which are positive and relatively large. These positive correlations correspond with economic theories describing a positive effect on market performance of good projections and high dividends. Sales growth and dividends are also positively correlated with ROA, which is sensible, as higher growth and sales contribute to better accounting performance. However, growth opportunities are negatively correlated with ROA, potentially caused by the larger R&D expenses. In addition, firm size is positively correlated with Tobin’s Q and ROA, and firm age also appears to be positively associated with accounting performance. Furthermore, the firm risk and capital structure measures are both negatively correlated with both Tobin’s Q and ROA, which is in line with theories describing the relation between higher firm and default risk and performance. The correlations emphasize the need for inclusion of the control variables, as they may affect performance.
On the basis of the difference of means tests and the correlations, no conclusions regarding a potential causal relation can be drawn. The contribution of all explanatory and control variables to firm performance differences between family-owned and other firms are examined in a multivariate analysis. The methodology of this analysis is discussed next.
3.3 Methodology
This section presents the methodology for the multivariate analysis. It describes the regression models used to analyze the potential causal relation between family firms and both market and accounting performance. The multivariate regressions performed in this study are two-way fixed effects models. This is based on the results of Hausman tests that indicate that using fixed effects models instead of random effect models is appropriate. These models are also used in Anderson and Reeb (2003). The main difference with their paper is that this study features a pre-crisis and crisis period. This is done because the influence of the financial crisis
26 on family-owned and founder-CEO firm performance is a main focus of this study. Moreover, the effect of dual-class shares on performance is studied by including a dual-class dummy.
Firstly, the equations for the multiple fixed effect regressions models are given. This is done first for the total sample regressions that include the family-owned and dual-class share dummy’s. Thereafter, the founder-CEO regression equations are presented, which are run on the S&P 1500 subsample consisting of 4,002 observations with available data for the founder-CEO variable. For each regression, an explanation of the model is given. In this explanation, it is described which variables are included and what coefficients are of interest.
3.3.1 Performance regressions: family-owned firms and dual-class shares
The first main regression equation used in the multivariate analysis is regarding the effect on performance of family-ownership and dual-class shares during the total period (2006-2010), the pre-crisis period (2006-2007) and the crisis period (2008-2010). The equation is:
𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖𝑡 = 𝛽0+ 𝛽1𝐹𝑎𝑚𝑖𝑙𝑦𝑂𝑤𝑛𝑒𝑑𝑖𝑡+ 𝛽2𝐷𝑢𝑎𝑙𝐶𝑙𝑎𝑠𝑠 + 𝛽3𝐶𝑉𝑖𝑡+ 𝛼𝑖 + 𝜆𝑡+ 𝜖𝑖𝑡 (1) In this regression and in all other performance regressions, 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖𝑡 is the measure of
firm performance, which is either the ROA based on net income or Tobin’s Q. 𝐹𝑎𝑚𝑖𝑙𝑦𝑂𝑤𝑛𝑒𝑑𝑖𝑡 is a binary variable for each firm-year, which takes the value of one if the firm is family-owned and zero otherwise
.
𝐷𝑢𝑎𝑙𝐶𝑙𝑎𝑠𝑠𝑖𝑡 is also included as a dummy variable, with a value of one when dual-class shares are being used in the firm and zero otherwise. The regression is run for the total period the non-crisis period and the crisis period. The coefficients of interest are 𝛽1 and 𝛽2. The sign and significance of coefficient 𝛽1 are used to tests the first and third hypotheses, which expect outperformance of family-owned firms in both non-crisis and crisis years, respectively. The coefficient 𝛽2 indicates the effect of class shares on firm performance, thereby testing the fifth hypothesis that predicts that dual-class shares negatively affect performance during non-crisis and crisis years.In addition, 𝐶𝑉𝑖𝑡 in the equation represents all control variables that are described in
the data section. In addition, the model controls for firm fixed effects, denoted by 𝛼𝑖, and for year fixed effects, denoted by 𝜆𝑡. By including the firm fixed effects, the regressions also control for industry effects. Lastly, to ensure a correct estimation of the coefficients, clustered standard errors are employed in the regressions, which are clustered by firm (“gvkey” of each company). In all other regressions the same controls and standard errors are used. Including
