The influence of family ownership on the information provision

(1)

The influence of family ownership on the

information provision

Name: Jan Brakenhoff Student number: 10657622

Thesis supervisor: Dr. P. Ghazizadeh Date: June 25, 2018

Number of words: 16,044

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

(2)

Statement of Originality

This document is written by student Jan Brakenhoff 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)

3

Abstract

Prior research emphasizes the potential differences in information provision between family firms and non-family firms due to divergent incentives. This thesis examines the effect of the presence of founding family ownership on the information provision in comparison with non-family ownership by using a different approach. The abnormal volatility reaction around earnings announcements is utilized as proxy for the information provision for the largest 2,000 domestic firms from the United States from 2001-2010. The results provide no direct evidence that there is a difference in the information disclosure (i.e. different volatility reaction) by listed family and non-family firms. However, the descriptive statistics and the closely related dual class share structures variable suggest a negative influence of family ownership on the volatility around earnings announcements (i.e. better or more information provision). To confirm this suggested relation, additional analysis has been performed. These additional analysis provide evidence for the suggested negative relation by showing that family firms are associated with less research and development activities, less cash holdings and better financial performance. Taken all the findings into consideration, there is an increased likelihood that family firms have less severe volatility reactions around earnings announcements and maintain higher levels of disclosure compared to non-family firms.

(4)

1 Introduction

Accounting represents one of the most under developed streams of research in family firm studies, despite the fact that accounting research is one of the eldest business disciplines and they represent the prevalent form of ownership structure in the world (Songini, Gnan & Malni, 2013). This prevalence is shown in the fact that about one third of the firms listed on the Standard and Poor’s 500 are family firms. In these largest U.S. firms, the families have held their stakes on average for over 78 years, and control over 18 percent of their firm’s ownership rights. In those cases, in which the family does not have the majority share ownership rights, they directly control 2.8 times as many board positions as their ownership share provides. Also, family members serve as top executives or CEO in 63% and serve on the board of directors in 99% of the cases (Burkart, Panunzi & Shleifer, 2003; Anderson & Reeb, 2003). These figures show that family ownership and influence are quite prevalent and significant in U.S. firms. Hutton (2007) and Ali, Chen and Radhakrishnan (2007) emphasize the importance of studying the influence of family firm incentives, especially on the information provision.

The aim of this study is to examine if there is a difference in the information disclosure between listed family and non-family firms. This paper attempts to answer the following question: RQ: How does the presence of family ownership affect the volatility around earnings announcements?

Providing an answer to this research question is relevant because it gives additional insight into potential differences in information disclosure and transparency by listed family and non-family firms by using another proxy for information provision and transparency than the previous studies. The contribution of this paper is the combination of two distinct research streams in accounting. The first research stream emphasizes the difference in information provision between family and non-family firms through their unique position within the firm (Ali, Chen & Radhakrishnan, 2007; Anderson, Duru & Reeb, 2009). The second stream emphasizes the value relevance of information disclosure and the corresponding share price reaction i.e. volatility (Beaver, 1968; Landsman & Maydew, 2002). The combination of the two is relevant because it gives a new insight into the difference in information disclosure by family and non-family firms. This is useful because of the inconsistency in the literature about the influences of the founding family on the corporate information provision. After the definition of a family business is been clarified, this inconsistency will be discussed.

(6)

In this research a family firm is defined as a listed entity in which multiple members of the same family are involved as large share owners or hold management positions, either contemporaneously or over time. This enables a number of variations: in the level of ownership and voting control, in the managerial roles played by family members, and in the family generation of key family members (Miller et al, 2007). Those characteristics of family firms results in unique position for the family in comparison to non-family firms.

According to Jiraporn & Dadalt (2009), this unique position leads to significant long-term investment horizons. They pretend that founding families view their holdings as an asset to be passed along to heirs rather than a mean of funding current consumption. Their long tenures with the firm gives the founding family greater levels of firm-specific knowledge, making them potentially better monitors of management. Due to better monitoring, it is assumed that there is convergence of interest between the controlling family and external investors. This will reduce investors' need to control insiders, reducing the need for financial disclosure. Lakhal (2005), Zhao and Millet-Reyes (2007) corroborate that family firms disclose less information because of the combination of ownership and management positions held by the family. This combination could also have a contradicting influence because the large concentrated equity positions can entrench families, causing the presence of information asymmetry problems between the family and outside investors. When investment decisions are more likely to be made to maximize the families’ wealth at the expense of outside investors, outsiders will find it necessary to oversee family managers by increasing the extent of information disclosures (Chau & Grey, 2010). Ali et al. (2007) confirm this reasoning of enhanced information disclosure by family firms. This reasoning can be strengthened by the fear of reputational damage that could lead to more openness (Jiraporn & Dadalt, 2009).

Taken the previous mentioned incentives of family firms into consideration, there is discrepancy in the literature about the influence of the incentives on the information disclosure of family firms. The aim of this study is to provide additional insight into this relation by analysing the stock price reaction around the earnings announcements for the largest 2,000 domestic firms from the United States from 2001 till 2010. The difference in price reaction by family and non-family firms is analysed and controlled for side effects to ensure the results are not explained by another factor. This study utilizes the volatility as proxy for the information environment to investigate the difference between family and non-family companies. The volatility is the standard deviation (i.e. dispersion) from the share price returns from a firm (Beaver, 1968). This price reaction displays the dispersion of information availability by investors. The disclosure of the earnings announcements results in a diverse response from different investors, caused by the deviation in precision or the availability of information (Kim & Verrecchia, 1991; Harris & Raviv,

(7)

7

1993; Bailey, Karolyi & Salva, 2006). This price reaction is characterized by the amount of new information incorporated in the disclosure (Beaver, 1986; McNichols & Manegold, 1983; Landsman & Maydew, 2001). Increased volatility at the moment of disclosure is considered evidence that new information is disclosed because investors are more vigorously revising their expectations. Kim and Verrecchia (1991) show that the amount of pre-disclosed information has a mitigating influence on the volatility around the earnings announcements, because this information is already incorporated in the stock price and expectations.

This mitigation effect is relevant for this research because it shows that when a company publishes more timely information over a period of time, the company is supposed to be more transparent (Kim & Verrecchia, 1991). The findings from this research provide no direct evidence for the influence of the presence of the founding family in the firm on the information provision because there is no influence found on volatility reaction. However, the other results suggest a negative association between family firms and the volatility, indicating that family firms potentially provide more or better information. To reinforce this suggestion, additional analysis has been performed. The results of this additional analysis confirm the suggested relation. Thereby, this research provide evidence for the increased likelihood that family firms have less severe volatility reactions around earnings announcements and provide better or more information compared to non-family firms.

The findings of this study provide an additional insight into the influences of family ownership on the provision of information. This contributes to the existing literature by reducing the inconsistency in the literature by providing evidence that it is more plausible that families influence results in more or better information provision (i.e. less volatility around earnings announcements). In addition, contributes this research to the existing literature by examining the information provision using a different proxy of information provision compared to e.g. Anderson, Duru and Reeb (2009) and Ali, Chen and Radhakrishnan (2009).

In the next section, the theoretical background of family ownership and the relation between information disclosure and price movements of share prices are discussed in the literature review. In paragraph 3 the hypothesis will be developed. The methodology and research design are explained in paragraph 4. The results will be discussed in section 5. Paragraph 6 consists of the conclusion and the limitations of this research.

(8)

2 Literature review

In this paragraph the theoretical and empirical background of this thesis will be discussed. Within the accounting literature the family firm research is a relatively young research field. In the last decade the amount of research done in the field of family firms is increased, those relatively new studies focus on the involvement of the family in corporate governance and management (Songini, Gnan & Malmi, 2013). This research focuses on the influence of the involvement of the family on the information provision of the company. In the literature there is inconsistency about the impact of family ownership on the transparency and quality of the disclosure information. Although regulators mandate extensive disclosure for listed companies, there is still considerable variation between the transparency of the companies, depending on the disclosures, the levels of private information and the coverage of financial analysts. Companies therefore have a tendency to maintain different types of information dissemination (Anderson, Duru and Reeb, 2009). This illustrates the discretion used by management to disclosure information and affecting their transparency.

The distinction between founding family and non-family firms originates from the following characteristics. The family represents a persistent type of large undiversified shareholders in combination with influence through disproportionate board control, management positions, dual-class share structures and long-term affiliation with the firm (Anderson, Duru & Reeb, 2009). The family has a combination of share ownership and management positions that can change simultaneously or in the course of time. This enables multiple variations: at the level of ownership and voting rights, in the management functions of family members and in the family generation (Miller et al., 2007). Those characteristics result in different incentives, objectives and priorities when it comes to information disclosure by family firms in comparison with non-family firms.

Previous studies discuss these differences in information dissemination between family and non-family businesses based on different incentives. In these studies, the distinction between family and non-family businesses is explained on the basis of the agency theory (Ali et al, 2007; Hashim, 2011; Wan-Hussin, 2009). Two types of agency problem are distinguished in the literature. According to Wan-Hussin (2009), Ali et al. (2007) and Anderson and Reeb (2003) family firms face less severe agency problems that arise from the separation of ownership and management (i.e. agency problem type I). However, they face more intense agency problems between the controlling family and the non-controlling shareholder (i.e. agency problem type II). In other researches the competing incentives by the firms are described as the entrenchment- and alignment effect. This research stream elaborates on the previous mentioned agency problems. The entrenchment effect

(9)

9

contains the less transparent arguments by family firms and the alignment effect includes the more transparent arguments (Hashim, 2011).

The combination of ownership and management (i.e. agency problem type I) creates a unique monitoring position for families. They have prime knowledge about their firms’ activities, which enables them to provide superior monitoring of managers. Thereby, they tend to have significant longer investment horizons compared to other investors. In comparison with non-family firm, non-family firms face less intense hidden-information problems due to the separation of ownership and management (Anderson, Mansi & Reeb, 2003). The alignment between the family and the shareholders should lead to more information disclosure.

In the literature there is a plausible alternative to the aforementioned argument that focuses on the unique position of the family to monitor managers (Anderson & Reeb, 2003). This extensive monitoring by the family results in the provision of control and oversight that can substitute the role of information disclosure relative to diffusely held firms (Anderson, Duru & Reeb, 2009). This contradicts the previous mentioned argument and shows that superior monitoring perspective also suggests a negative relation between family ownership and transparency (Anderson, Duru & Reeb, 2009). This is confirmed by Lakhal (2005), which shows a reverse relationship between the concentration of ownership and management and the amount of transparency.

The large, non-diversified equity holdings by the family lead to substantial control (agency problem type II) in their companies, their voting rights exceeds their dividend rights, usually resulting in dominance in the board of directors. This gives the family the power to extract private benefits. Ho and Wong (2011) find evidence that the percentage of family members on the board of managers is inversely related to the disclosure practices. Anderson, Duru and Reeb (2009) support this point of view and argue that the information disclosure can vary substantially in the presence of family ownership. In their study they argue that founders’ unique control positions provide particularly strong incentives to diminish corporate transparency, because of the possibility to entrench family benefits from this control position. This hidden-information between the family and the outside shareholders, results in less transparency (Fan and Wong, 2002). This incentive can be reduced by the expectation of outside shareholders who will find it necessary to oversee family managers by increasing the extent of information disclosures (Chau & Grey, 2010)

The incentive to withhold information will also be mitigated by the reputational concerns of the family (Anderson & Reeb, 2003). This argument is supported by the notion that the family has a sustainable presence in the firm and their intention to preserve the family name and company (Wang, 2006). The sustainable presence of the family within the firms results from the incentive

(10)

of succession of the founder by its heirs. This long-term focus results in the alignment of the incentives of the family with outside shareholders. The sustainable incentive within the firm results in increased information disclosure through the stewardship role of the family (Anderson, Duru and Reeb, 2003).

The above-mentioned incentives from family businesses are ambiguous about the impact on disclosure practices. The possible difference in disclosures between family and non-family firms could be measured with share price movements around earnings announcements, in specific the volatility reaction. The volatility is a statistical measure of the dispersion (i.e. standard deviation) of share price return for a specific share. The change in share price measures the average market reaction of all traders (Beaver, 1968).

Based on empirical studies, volatility is seen as a measure of information asymmetry and differential information processing by investors (Bailey, Karolyi & Salva, 2006). Francis, Schipper and Vincent (2002) describes that the volatility reaction around a disclosure is an increasing function of the level of new information provided by the firm. This reaction can be mitigated by more pre-disclosed information given by the firm. This implies that more transparent companies respond with less severe volatility due to less newly disclosed information in the announcement. When the earnings announcement contains new information, better informed investors make smaller revisions in their expectations of the value of the share (e.g. the family) than less well-informed investors (e.g. external investors) with less precise private information. This precise information is in the literature described as private information which is basically defined as information known by the managers of the firm and not by investors. On the contrary, public information is the information already available to investors. This implies that private information has more value relevance, because it gives additional insights. As mentioned earlier, due to the unique position within the family business, the family is likely to have private information in accordance with managers.

It is assumed that investors are informed differently and differ in their information positions. This leads to deviant reactions from different investors, depending on their availability of information. The information provided may be a confirmation (i.e. public information) of what is already known to the investors or the information is may be a surprise (i.e. private information). When the information disclosed is a surprise, the return on the share price is heavier compared to the conformance of the information, because this information is already included in the share price. Thereby increases the importance of the disclosure when the information disclosed is more precise and decreases with the precision of the previously disclosed information. In addition,

(11)

11

recently published information is relatively more important for investors, resulting in a greater impact on their expectations (Kim & Verrecchia, 1991; DeFond, Hung, Trezevant, 2007).

The choice for the volatility reaction (i.e. the volatility of the share price) stems from the fact that the volume reaction reflects the sum of the differences in investor response and the price reaction measures the average response (Kim & Verrecchia, 1991). As a result, the volume is proportional to both the price response and the degree of differential information precision. Therefore, the precision of the volume response is not discernible, suggesting that the trade volume is a noisier indicator of the value relevance of the disclosed information than the price response. (Kim & Verrecchia, 1991).

This research aims to contribute to the existing literature by investigating the difference in information provision by family firms. In the prior literature there is inconclusiveness about the effect of the presence of the founding family ownership on the information disclosure or transparency. This research aims to find evidence on this relation and finds a significant different volatility effect. The earnings announcements are used as proxy for the information content of the disclosures. The disclosure of earnings announcements is assumed to contain value relevant information. Beaver (1968) and Landsman and Meydew (2002) confirm this by providing evidence that earnings announcement has increased in informativeness in the last decade. The volatility around those announcements indicate the amount of new information disclosed by the announcement. This reaction has an inverse relation with the information provision or transparency of the firm (Bailey, Karolyi & Salve, 2006).

(12)

3 Hypothesis development

Recent research on founding family incentives illustrates several implications how investors respond to earnings announcements. This paragraph appeals to the literature mentioned in the previous section to identify how the presence of founding family within the firm is likely to influence the information content of earnings announcements. In addition, we develop a hypothesis that predicts how the presence of founding family ownership is likely to affect volatility reaction resulting from the information content of annual earnings announcements.

The researches of Chau and Grey (2002); Chen, Chen and Cheng (2008) and Anderson, Duru and Reeb (2009) indicate that the level of information disclosure is likely to be less by family-controlled companies. Zhao and Millet-Reyes (2007) confirm these results by referring to the lack of incentive to report timely and relevant earnings to minority investors by family firms. In addition, the studies of Ho and Wong (2001) and Lakhal (2005) provide evidence on the negative relation between the percentage of family board members and the extent of voluntary disclosure by the family firm. The results of these studies indicate that founding families provide less disclosures compared to non-family firms, resulting in less transparency.

The previously mentioned evidence is contradicted by Hashim (2011). In his research Hashim (2011) finds that the agency theory prediction is confirmed for non-family firms, but evidence from family firms is not significant or conflicting with the agency theory prediction. Also, Anderson and Reeb (2003), Wang (2006); and Ali et al. (2007) find that family-controlled firms disclose better accounting information compared to their non-family-controlled counterparts. Chau and Grey (2010) show that the amount of disclosure depends on the family shareholdings. When the family holds moderate levels (i.e. less than 25 percent) of family shareholding, the convergence of interest effect is dominant, and the extent of transparency disclosures is relatively low. At increased levels of family shareholding, the entrenchment effect dominates and is associated with higher transparency. The results of Ali et al, (2007), Wang (2006), Hashim (2011) and Chau and Grey (2010) indicate that the incentives of family firms could also result in increased information disclosures compared to non-family firms, leading to more transparency.

The above-mentioned research results show the inconclusiveness of the effect from the incentives on the transparency of family firms in comparison with non-family firms. Nevertheless, there are relatively more studies suggesting the negative influence of family ownership on the provision of information and the transparency. But there is also significant evidence that family firms provide better accounting disclosures. This makes it difficult to form an expectation based

(13)

13

on this prior literature, because of the ambiguousness of the incentives. This research seeks to give an additional insight into the effect of family firms on the information provision.

The magnitude of the price reaction stems from the average change in investors’ expectations and this depends on the magnitude of the new information and the average precision of investors’ information. If the information availability diverges between family and non-family firms because of their deviating incentives, the volatility of the share prices should be more (less) severe in case of less (more) transparent family firms, the information content of the earnings announcement is higher (lower) and the precision of the prior private information should be smaller (greater). Accordingly,

Hypothesis 1. The presence of founding family ownership affects the volatility around earnings announcements.

(14)

4 Methodology and Research Design

In this paragraph the methodology of this empirical research will be explained. The aim of this section is to describe the actions taken to investigate the information disclosure by family firms. This paragraph will first illustrate sample, secondly, the origin of the main variables used in this thesis. Third, the techniques used to measure the volatility will be discussed. Fourth, the regression model to measure the effect of family ownership on the volatility will be explained. Finally, the control variables added to the model will be clarified.

4.1 Sample selection

In this paragraph the composition of the data sample used will be explained. This study builds on the family firm data sample of Anderson, Duru, and Reeb (2009) and Anderson, Reeb, and Zhao (2012). This data sample is obtained from the website of R.C. Anderson.

It consists of the 2,000 largest domestic firms based on total assets from the United States. They exclude the following firms from the sample; regulated public utilities, financial firms, firms listed as master limited partnership, firms with a share price less than 0.25 U.S. dollar and non-domestic firms. The raw data is obtained from the Compustat database from January, 2001. To control the survivorship bias Anderson, Reeb and Zhao (2012) allow firms to exit and re-enter the sample. The total sample consists of 2,000 largest firms from 2001 and stretches from 2001 till 2010, consisting of 40,1 percent family firms and 59,9 percent non-family firms.

While refining the sample used in this research, attention was guided by a wide variety of prior research from e.g. Anderson and Reeb (2004), Ali, Chen and Radhakrishnan (2007), Block (2010) and Jaskiewicz et al. (2017). In their researches they use the Standard and Poor’s 500 (hereafter S&P 500) as sample. The choice for the S&P 500 has the benefit of making the sample more homogeneous with respect to firm size (Ali et al. 2007; Hutton, 2007). However, there are disadvantages as well. Hutton (2007) and Chen, Chen and Cheng (2008) criticize the decision because the focus on the S&P 500 limits the generalization of the findings. The broader the sample, including smaller firms, the larger variation in the sample in both family ownership and disclosure behaviour. This leads to potentially more powerful tests, and more importantly, it allows this research to generalize to a broader segment of the economy. Thereby the information availability at the S&P 500 is relative high (Anderson, Duru and Reeb, 2009). Firms listed on the S&P 500 undergo a great deal of market scrutiny and maintain higher levels of information availability. This mitigates the potential influence of family ownership on provision of less (more) information disclosure. Based on these arguments the full sample of the Anderson, Reeb and Zhao (2012)

(15)

15

consisting of the largest 2,000 domestic firms of the United States is used as the foundation of this analysis.

TABLE 1

Sample Development

FULL SAMPLE FAMILY FIRM NON-FAMILY

Obtained from Database

Firms 2,000 802 1,198

Observations 67,336 26,870 40,466

Dropped by Eventus

Firms (19) (10) (9)

Observations (4,803) (1,458) (3,345)

Dropped during merging

Firms (54) (23) (31)

Observations (4,552) (2,217) (2,335)

The Final sample

Firms 1,927 769 1,158

Observations 57,981 23,195 34,786

The numbers between brackets are dropped/deducted from the obtained firms or observations.

During the research process, 73 companies from dropped out the sample, resulting in a sample of 1,927 companies. The development process of the sample is shown in Table 1. The number of observations display the earnings announcements.

4.2 Dependent variables

This section discusses the main variables used in this thesis, the origin of the data and the modifications made to the data. The variables used in this thesis have two sources: the website from Anderson, Duru and Reeb (2009) and the Wharton Research Data Services (hereafter WRDS). The WRDS system has access to several financial databases e.g. Compustat Capital IQ and CRSP database. First, the family ownership variable will be explained. Second, the origin of the earnings announcements will be clarified. Third, the abnormal share price returns will be discussed.

(16)

The family firm data (founder and/or heir ownership) from Anderson, Duru and Reeb (2009) retrieved from their website consist of manually collecting the data from individual company websites and from corporate proxy statements (i.e. 10-k’s) for the years 2001-2010. The complete dataset consists of the number, the data-year and the company name. The GVKEY-number is the firm specific identifier for the firms. The dataset consists of a dummy-variable FAMFIRM that equals one when the family owns at least 5 percent of the company shares and zero if this is not the case. By the composition of the sample Anderson, Duru and Reeb (2009) did not include shares held by charitable foundations as part of the family holdings. This is the case in less than 1 percent of the sample where foundations hold substantial ownership rights with the express intent of promoting public welfare rather than economically benefiting family members.

The reporting date of the quarterly earnings announcements is obtained from the Compustat Capital IQ. The Report Date Quarterly Earnings (RDQ) is retrieved from the beginning of 2001 till the end of 2010 for the entire database. Combining the RDQ-data and the sample resulted in 67,336 RDQ’s for the 2,000 firms from the sample. This is shown in Table 1 under “obtained from database”. In this research the earnings announcements are used as proxy for value relevant information disclosure. Prior literature emphasizes the value relevance of the quarterly earnings announcements (Landsman & Maydew, 2002).

The security price reaction resulting from the earnings announcement needs to be measured. This type of empirical research is called an event study. An event study is an empirical investigation of the impact of an economic event, i.e. the firm specific quarterly earnings announcement, on the security price return. (Yadav, 1992). The main concept is to measure the abnormal return attributable compared to the earnings announcement by adjusting the return that stems from the price fluctuation of the total market. This abnormal return method is commonly used by the prior studies from e.g. Beaver (1968), Yadav (1992), Landsman and Maydew (2002) and Bailey, Karolyi and Salve (2006).

Beaver (1968) explains the importance of removing the effects of market-wide events upon the price fluctuations in more detail. The motivation is twofold. First, it is possible that the abnormal high returns may be caused partly by the market-wide pieces of information that are released at the same time as the earnings announcements. For example, the quarterly earnings announcements are released almost uniformly throughout the year. However, removing the market wide effects should diminish the probability that external effects account for the results. Second, the analysis will serve to reduce noise in the price data. Noise is any movements in price due to

(17)

17

unspecified factors, one of which is market-wide events that would cause increases in price (Beaver, 1968).

In this thesis the abnormal returns are retrieved from the Eventus Query software available on WRDS. Eventus is a software application which performs event studies using the CRSP database in combination with user-collected data. The software is appropriate for this research because it deals with similar events occurring at different times for different firms. The software application takes the calculating part of the event study for its account. The specific type of Eventus query used in this research is the cross-sectional analysis (daily). The output provided by Eventus are the firm and announcement specific abnormal returns for different time windows.

The main input which needs to be provided (the user-collected data) contains two variables: the firm specific number (PERMO) and the reporting date of the quarterly earnings announcement (RDQ) (i.e. YYYYMMDD), as described previously. Other important input for the query are the event windows/event period and the estimation period. The studies from Bamber (1986), Landsman and Maydew (2002), and Bailey, Karolyi and Salva (2006) emphasizes that most of the price reaction occurs during the days -1, 0 and +1 whereby day 0 is the announcement date. They argue that the price reaction will dilute when they choose for a longer event period. In line with their reasoning the event windows used in this study matched the windows used by Bamber (1986), Landsman and Maydew (2002) and Bailey, Karolyi and Salva (2006). The estimation period used in this study -200 and -11 with respect to the announcement date is also based on this prior research. This period is used to adjust the return from market trends previous explained by Beaver (1968). The output generated from Eventus is used to calculate the abnormal volatility reaction around the specific earnings announcement date. As indicated in Table 1, Eventus is not able to provide abnormal returns for 4,803 earnings announcements because: the date outside of period 2001-2010 (4,470), too few estimation period days with data (220) and too many event period days with missing data (113).

4.3 The abnormal volatility

The volatility is a statistical measure of the diffusion of the returns for a given share. In this thesis the volatility is used as proxy for the information provision of firms in specific, family firms. The volatility can be measured by the variance or the standard deviation from the same share, but the volatility is common interpreted as the standard deviation of returns (Landsman & Maydew, 2002). In this thesis the abnormal volatility 𝐴𝑉𝑂𝐿(&,() is calculated based on the output of eventus and

(18)

𝑨𝑽𝑶𝑳(𝒇,𝒊) = 2₄₅₃3 ∑ (𝑅<(=8 (&,(,8)− 𝑅:::::::)(&,8) ; (1)

𝑹_𝒇,?

::::: = ₄3∑4 𝑅_(&,(,8)

8=( (2)

𝒍𝒐𝒈_𝑨𝑽𝑶𝑳_(𝒇,𝒊) = 𝑙𝑛(𝐴𝑉𝑂𝐿_(&,()) (3)

Where the 𝐴𝑉𝑂𝐿(&,(,8) is calculated over the abnormal returns obtained from Eventus are divided

into different daily windows, these windows 𝑅_(&,(,8)represent the daily abnormal returns for days t = -1, 0, +1. This is done for firm f and for i - reporting date quarterly earnings announcement. N stands for the number of observations/windows used to calculate the volatility and is equal to 3. 𝑅_(&,F)

:::::: is the average of the daily abnormal return windows for t = -1, 0, +1. The mean is calculated according to the following formula 2. Hereafter, the 𝐴𝑉𝑂𝐿(&,() variable will be mathematically

transformed into the natural logarithm because of the skewed distribution of the original AVOL. This is for statistical purposes.

4.4 Regression model

In this section the primary quantitative test of the paper will be defined. The purpose of the paper is to examine the influence of family ownership on the volatility, which is used as a proxy to examine the potential difference in information provision. First, the model will be introduced. Hereafter, the control variables will be explained.

The model is similar to the one used by Landsman Maydew (2002) and Anderson, Duru and Reeb (2009), added the family firm variable and various control variables used in other studies e.g. Ali, Chen and Radhakrishnan (2007) and Anderson, Duru & Reeb, 2009. There is an ordinary least squares (OLS) regression analysis performed to estimate the unknown parameters (𝛽₍). Accordingly,

𝒍𝒐𝒈_𝑨𝑽𝑶𝑳(𝒇,𝒊) = 𝛽I + 𝛽3 𝑭𝑨𝑴𝑭𝑰𝑹𝑴 + 𝛽; 𝑭𝑰𝑹𝑴𝑺𝑰𝒁𝑬 + 𝛽Q𝑺𝑯𝑨𝑹𝑬𝑺 + 𝛽S 𝑫𝑼𝑨𝑳𝑪𝑳𝑨𝑺𝑺 +

(19)

19 4.5 Control variables

In this section the control variables used in this research are illustrated. These variables are included because prior empirical research has shown associations with the volatility around earnings announcements. The incorporation of the control variables is an attempt to control the influence on the main results of the investigation. When considering the influence of one variable, the effects of all other variables are taken into account by including them in a regression to separate their influence from those of the explanatory variable of interest (i.e. family firm variable). The control variables used will be elaborated below.

The first control variable is the firm size (FIRMSIZE). Prior research suggests there is more information available before the earnings announcement (i.e. pre-disclosed information) by larger firms (Bamber, 1968; Landsman & Maydew, 2002). Previous studies suggest that this is an increasing function, in the sense that the value relevant information is known before earnings are announced in relation with the size of the firm (Bamber, 1986). The more information is available from alternative sources, the less important the earnings announcement and the smaller value relevance of the disclosed information. The market capitalization is calculated by multiplying the shares outstanding and the daily closing stock price, the data is obtained from the Compustat dataset.

The second control variable (SHARES) embodies the annually traded shares by the firms. The relative large undiversified stock holdings by the family from on average about 20 percent leads to the fact that this portion of shares permanently is not traded (Anderson & Reeb, 2003). The volatility originates from the imbalance between trade orders i.e. buy (sell) orders with little or no sell (buy) orders, then share price will increase (decrease) for a particular share. This relationship illustrates the relation between the trading volume and the price movements (i.e. volatility). If the shares trading volume is high, but there is a balance between the orders, then the volatility is low. When there are significant less shares traded annually the probability of more volatile stock prices is lower, which can potentially explain the designated lower volatility of family firms shares. Earlier research form Anderson, Duru and Reeb (2009) show in their descriptive statistics that family shares are less traded compared to diffusely owns firms (i.e. non-family firms).

The third control variable is the dual class structure (DUALCLASS). The presence of a different type of shares, with different allocated rights (e.g. voting right) results. Based on different prior studies, the presence of a different type of shares with different allocated rights results in the skewed distribution of the control and dividend rights. This results in more severe agency problems as described in the literature section (Lakhal, 2005). The family possess a significant

(20)

control position resulting in less transparent information disclosure. Other research from Khalil, Magnan and Cohen (2008) emphasize this influence on the transparency by non-family firms. The dual class shares variable is a dummy-variable DUAL that equals one when the firm has a dual-class share structure and zero if there are no different types of shares. This variable is obtained from the dataset from Anderson, Reeb and Zhao (2012).

The fourth control variable is the INDUSTRY dummy. This dummy-variable equals one when the firm operate in a high volatile industry and zero for the other business operating in a less volatile susceptible industry. Cavaglia, Brightman and Aked (2000) and; Ferreira and Gama (2005) emphasize that some industries are more volatile and provide evidence in their research on the effects of controlling for industry influence. Based on their results is controlled for the following industries: technology, media, telecommunication, financial, energy, health, minerals and heavy industry. The variable INDUSTRY is manually created based on the four-digits Standardized Industry Classification codes (SIC) obtained from the Compustat. The retrieved SIC numbers with company classification code are matched with the initial dataset. The corresponding firms and SIC-numbers are compared with the overall list of four-digit SIC-SIC-numbers. The sample consists of 432 different industry sectors, whereof 143 operate in more volatile subindustries. The 143 SIC-numbers of the industries indicated as more volatile are displayed in the table in Appendix A.

The fifth explained variable is the listing on the S&P 500 which is frequently used to investigate the influence of founding family ownership. Anderson and Reeb (2009) maintain that there is a supplementary influence of the listing on the S&P 500 on the information provision. Firms part of this major market index are heavily followed by analysts and media outlets. These listed firms receive more market research due to distrust and maintain higher levels of information availability compared to other not on the S&P 500 listed large firms. This results in a potential fading effect on the volatility. The variable developed to indicate the listing is a dummy variable (SP500) that equals one when firms are included in the S&P 500 and zero otherwise.

The last control variable is the CRISIS variable. The sample used stretches from the beginning of 2001 till the end of 2010, but the years 2008 and 2009 are characterized as the financial credit crisis. This resulted in an increase of market scrutiny due to the loss of confidence in the financial markets in that period. Camodeca, Almici, and Bernardi (2013) confirm in prior literate that there is a higher demand of information during times of economic crisis. Other research emphasized the increased volatility during the financial crisis through the stock market crash (Schwert, 2011). This leads to the incorporation of a dummy variable CRISIS turning one for the year 2008-2009 and zero for the other years within the sample.

(21)

21

5 Results

5.1 Descriptive statistics

In this paragraph the quantitative characteristics of the sample and the variables will be illustrated. First, the relevance of the earnings announcements (RDQ) will be discussed. Secondly, the other summary statistics will be explained. Finally, the Pearson Correlation Matrix will be discussed.

The foundation of this empirical research is the market reaction around the quarterly earnings announcements. This reaction is important for this research because the share price fluctuations are used as proxy for the information provision of firms. Because of this importance, this study examines whether the abnormal volatility around the earnings announcements actually deviates from the daily volatility of the shares and therefor has explanatory power to use this volatility reaction as a proxy for information disclosure.

Table 2

Mean Abnormal Market Returns - relative to Reporting Data Quarterly Earnings

Days _{Mean Returns} _{Mean Volatility} F-Value

-3 0.02% 0.99% 0.58 -2 0.04% 1.01% 0.596 -1 0.10% 1.69% 1.681* 0 0.21% 2.31% 3.134* +1 0.05% 1.89% 2.118* +2 -0.01% 1.12% 0.731 +3 -0.03% 0.89% 0.467

The F-Value indicates the significance of the mean volatility.

The symbol * shows the significance of the F-value at a 0.01 percent level (two-tailed) The bold printed parts of the table indicate the used day windows of this study. The mean statistics in this table are based on the 62,533 observations (RDQ)

The raw daily returns are obtained from the Eventus event-study software application on the WRDS-website.

The results in Table 2 present the abnormal market reaction obtained from Eventus. This table displays the average abnormal volatility and market return for the firms used in this research over a seven-day window; -3 days before the reporting date, the reporting-date 0 and +3 days after the reporting date of quarterly earnings. As displayed in the table, the largest abnormal volatility in the days are the -1, 0. +1, relative to the reporting date of the quarterly earnings. This confirms the notion in the papers of Beaver (1968) and Landsman & Maydew (2002) that most of the

(22)

price reaction occurs between this period. Besides this, the statistics in Table 2 validate the choice for these windows as foundation of this volatility analyses, because the most significant price reaction occurs in this period. The choice of a different (e.g. longer) window would have resulted in a mitigating effect on the volatility reaction.

The F-value column in Table 2 indicates that the daily average abnormal volatility deviates significantly from the abnormal volatility for estimation period. The estimation period used is the 7-day windows displayed in the table with the average volatility of 1.3039%. This confirms the pre-assumed notion from the papers of Beaver (1968) and Landsman & Maydew (2002) that the quarterly earnings announcements are highly value relevant disclosures. This is important for this study because the quarterly earnings announcements are used as a proxy for value relevant information disclosure to examine if family firms differ in the disclosure of information.

TABLE 3

Differences of Mean test – key Variables

Mean Values

FULL SAMPLE FAMFIRM NON-FAMFIRM T-STAT

(N=57,981) (N=23,195) (N=34,786) AVOL 0.6877 0.6775 0.6946 -5.585** SIZE 20.9256 20.4948 21.2129 -54.862** SHARES 18.5143 17.9364 18.8996 -68.620** DUAL (%) 0.1008 0.2186 0.0223 69.443** INDUSRY (%) 0.3650 0.3337 0.3856 -12.875** SP 500 (%) 0.2198 0.1267 0.2819 -47.696** CRISIS (%) 0.1711 0.1715 0.1709 0.188

The symbol *, ** indicates a significance level of 0.10; 0.001 percent, respectively.

The AVOL, FIRMSIZE & SHARES variables are winsorized at the 1% and 99% levels to correct for outliers. The AVOL, FIRMSIZE & SHARES variables are displayed in natural logarithm; e.g. AVOL=ln(x)

% indicates the percentage of the observations/firms matching with the variable All numbers in this table are rounded up at four digits

As reported in Table 3, the mean abnormal volatility (AVOL) for the full sample is 1.989% (0.6877). The abnormal volatility as indicated in the table shows that family firms have lower volatility reaction around earnings announcements than the non-family firms, 1.969 % (0.6775) and 2.003 % (0.6946), respectively. The mean difference of the volatility reaction is significant at

(23)

23

0.01 percent level. This shows that our sample family firms are less volatile on average. Based on prior research this finding is not unexpected because family businesses are seen as operationally more conservative compared to non-family firms. Anderson and Reeb (2003) argue that family firms in the United States are commonly perceived as less profitable or efficient, because families forgo profit maximization instead of opting for a more sustainable operational strategy. Since they have the desire to pass the company to the next generation and see the company more as an asset than means for current consumption. This difference of mean test illustrates also that on average the family firms are more transparent in their information provision.

As described in Table 3, the number of non-family companies with dual class (DUAL) share structures is relatively small compared to the full sample, with the mean being 2.23%. When you compare the full-sample of 10.08% with the percentage of family firms is 21.86%, this difference is highly significant at 0.01 percent. The reason why dual class structures are more common by family firms is because they provide e.g. founders, families and executives the ability to control the majority voting rights with a relative small percentage of total equity. This is beneficial for them because they have control of the firm but bear personally significant less financial risk. This explains why the inverse relation between dividend- and control rights is more severe by family firms compared to non-family firms. The inverse relation result potentially in less transparency by family firms based on the previous study of Ho and Wong (2011).

Table 3 presents the mean number of shares traded for the full sample is 108.25 million (18.5). The average traded shares indicate that family firms have considerably less traded shares compared to non-family firms, 61.8 million (17.9) and 161.4 million (18.9), respectively. This difference is highly significant with a t- statistic of -68.6. This difference must be interpreted with caution, as it is not checked for company size, but it is in line with the expectation that the shares of family businesses will be traded less. This expectancy originates from the large undiversified shareholdings by the family in the family business. Anderson, Mansi and Reeb (2003) show that families hold on average about 23% of the family company shares. Less traded shares lead to a smaller probability of more severe volatility, because the volatility originates from the imbalance in trade orders a particular share on the stock exchange.

In terms of market capitalization (FIRMSIZE) the average market capitalization within the sample is (𝑒;I,b;Wc) 1,224.26 million US dollars. The mean of the market capitalization of the family firms is 795,7 million US dollars and the average non-family firms is 1,631.7 million dollars. This shows that the mean of family firms is significant smaller than the non-family firms at 0.01 percent level. This is consistent with prior research from Chen, Chen and Cheng (2008) and;

(24)

Anderson and Reeb (2003). They also show in their descriptive statistic section that family firms are on average smaller than the non-family firms.

This has a logic relation with the listing on the Standard and Poor’s 500, because only the largest 500 domestic firms list on this market index. The mean S&P 500 listing, as represented by the percentage of firms listed on the S&P 500, shows that 12.67% from the family firm lists on the S&P 500. The average of the full sample is 21,98% lists on the S&P 500, in comparison with the non-family firms 28.19%. The mean difference between the family and non-family firms is significant at 0.01 percent level. This corroborates the notion that on average less family firms list on the S&P 500 compared to the non-family firms (Chen, Chen and Cheng, 2008). The significant difference between the firm size and the listing on the S&P 500 suggests also less public information availability by family firms (Bamber, 1986).

In terms of INDUSTRY there is a significant divergence between the means. Family and non-family firms operate on average 33.3% and 38.6% in more volatile industries. This shows that on average family firms operate more in less volatile industries. This confirms the suggestion that family firms are in essence more conservative compared to non-family firms, because they have a long-term focus through their aspiration of succession by their heirs. Therefore, it could be argued that this desire would facilitate the choice of a less volatile industry. The results displayed in Table 3 corroborate this reasoning.

The CRISIS variable stands for the Financial credit crisis between 2008-2009. The means of the three samples are, as shown in the Table 3 almost identical. This shows that on average our sample exist of the same number of observations in the years 2008-2009. This was presumed because the quarterly earnings announcement is mandated by the Security and Exchange Commission. However, this average difference gives an indication that the variety in the number of proxies (i.e. earnings announcements) between family firms and non-family firms is far from significant between 2008-2009. This fact contributes to the validity of the earnings announcements as proxy.

(25)

25

TABLE 4

Pearson Correlation Matrix

1 2 3 4 5 6 7 8 1 AVOL 1 2 FAM -0.023* 1 3 SIZE 0.089* -0.219* 1 4 SHARES 0.108* -0.279* 0.789* 1 5 DUAL -0.016* 0.319* -0.061* -0.115* 1 6 INDUST 0.026* -0.053* 0.064* 0.099* -0.035* 1 7 SP500 0.027* -0.184* 0.673* 0.527* -0.078* 0.001 1 8 CRISIS 0.056* 0.001 0.016* 0.151* 0.004 -0.001 0.053* 1 The symbol * indicate that the correlation is significant at 0.01 % level

The numbers in this table are rounded up from four decimal places to three.

The AVOL, FIRMSIZE and SHARES variables are winsorized at the 1% and 99% levels to correct for outliers.

Table 4 displays the relation between FAMFIRM and SIZE. The negative correlation (-.22) confirms the results from the mean difference test and predicts a negative association between the presence of the founding family and the firm size. Providing evidence that the family firms are likely to be the smaller than non-family firms.

In terms of volatility presents Table 4 the correlation between the FAMFIRM variable with the abnormal volatility (AVOL). This negatively correlation (-.023) suggests that family firms are less volatile. Based on the difference of mean analysis this negative relation was expected, but the correlation confirms the notion that family firms are from origin operationally more conservatism compared to dispersed non-family owned companies. This suggests that the results of the main regression analysis potentially also find a negative influence from family ownership on the volatility reaction. Providing evidence for more or better information disclosure by family firms compared to non-family firms. The negative correlation can potentially be explained by the negative correlation (-.28) between SHARES (i.e. the number of shares traded) and the family firm (FAMIFIRM) because the volatility reaction emerged from the imbalance of the traded shares on the stock exchange. This correlation confirms the suggested fewer traded shares by family firm by the difference of mean test, previously.

Displayed in Tables 4, this technical origin from the volatility reaction is affirmed by the correlation between SHARES and AVOL of 0.11, the highest relation with the volatility indicated

(26)

in the correlation matrix in Table 4. As explained earlier, the likelihood of more severe volatility reaction originates by the number of shares traded because of the probability of an imbalance between the traded shares. The considerable positive correlation (0.79) between the SIZE and SHARES (i.e. the trading volume). This implicates that the number of shares traded is positive in relation to the firm size. Hegde and McDermott (2003) emphasize this positive relation and specifically associate the listing on the S&P 500 with an enlarged liquidity of the shares traded. This is validated by the notable correlation of 0.53 between SHARES traded and listing on the S&P 500. The firms listed on the S&P 500 are the largest 500 domestic firms in the United States. Table 4 confirms this with a considerable dependence of 0.67 between SIZE and the S&P 500. Anderson, Dura and Reeb (2009) claim that more information is available by firms listed on the S&P 500 through more marker scrutiny and enhanced media attention. Bamber (1986) argues that in general large firms have more public information availability, because of enhanced media attention and market scrutiny by larger corporations (e.g. labour conditions at the Nike factories in Bangladesh).

Surprisingly, Table 4 shows a positive correlation between SIZE and AVOL (0.089). This indicates that larger firms have more volatile stock prices, suggesting that they provide relative less information disclosure compared to smaller firms. This contradicts the previous mentioned arguments from Bamber (1986) and; Anderson, Dura and Reeb (2009). Nevertheless, this correlation should be interpreted with caution because the correlations is not checked for possible side influences. An example of a side influence is the credit crisis from 2008-2009. The sample period overlaps this period which is characterized by a stock market crash due to loss of confidence in the financial markets. This influence is shown in Table 4 by the positive correlation (0.15) between CRISIS and SHARE; and between AVOL and CRISIS (0.056), providing evidence that during the credit crisis more shares are traded and positively affected the volatility. The previously discussed positive relation between firm size and the number of shares traded suggests that during the crisis relatively more shares are traded for larger firms, resulting in heavier volatility reactions. This proposed influence is confirmed by an additional mean comparison of the AVOL and SHARES for only the years 2008-2009. The average volatility for the full sample is 0.735 for family firms 0.722 and for non-family firms 0.744 (t-statistic from -2,86). The number of shares traded for the full sample is 19.07 for family firms 18.44 and for non-family firms 19.5 (t-statistic -32,32). These means differ considerably from the numbers in Table 3 confirming the reasoning.

Displayed in Table 4, the correlation between the SHARES and the existence of a dual class (DUAL) share structure presents a negative correlation (» -.12). This implicates that the

(27)

27

presence of dual class structures has a negative influence on the number of shares traded. This can probably be explained through the fact that the shares with extra rights (i.e. control or dividend rights) are considerably less likely to be traded and thus the overall portion of probably traded shares is smaller.

In terms of firm size and the existence of dual class share structures Table 4 shows a negative (-0.06) correlation between SIZE and DUAL. This shows a negative relation between larger firms and the existence of dual class share structures. This can potentially be explained through the controversy about dual class shares. As explained previously, dual class shares make it possible to be controlling shareholder with a significant less portion of shares. This can potentially result in a divergence between the interest of controlling and the minority shareholders which makes it less likely for investors to invest in firms with inverse control- and dividend rights or resulting in a more prudent valuation of the firm shares by external investors leading to a lower market capitalization. Earlier research form Villalonga and Amit (2006) confirm this negative relation between the valuation of company shares by investors and the existents of a dual class share regime.

Table 4 presents a positive correlation (»0.32) between the FAMFIRM and the DUAL variable. This confirms the suggestion that the presence of the founding family increases the likelihood of a dual share structure, as indicated previously in the mean comparison test. Thereby, DUAL share structure correlates negatively in relation to the AVOL. This suggests that there is a decline in the volatility reaction around earnings announcements if firms possess dual share structures. This finding contradicts prior research from Ho and Wong (2011) because they argue that the presence of dual class shares structures results in less information provision and less transparent firms should have more severe volatility reactions around the earnings announcements.

As depicted in Table 4, the suggestion that family firms are less likely to operate in industries designated as more volatile is confirmed. The negative correlation of » -0.05 between FAMFIRM and INDUSTRY provides evidence that family firms operate less volatile industries. This probably explains the impact on the share price volatility from family firms because they are associated with less volatile industries, which makes it more likely that their share prices also less volatile. The

With respect to correlation among variables illustrated in Table 4, the correlation tested in the study confirms that no multicollinearity exists between the variables since none of the variables

(28)

correlates above the 0.80 or 0.90. This would have affected the predictive power of the model used in this research within the sample used.

5.2 The effect of family ownership on the information disclosure

TABLE 5

Regression Results - Abnormal Volatility

Expectations Coefficient t-stat p-value

b_𝟏 FAMFIRM ? .0049 1.49 .136 b_𝟐SIZE (-) .0133 7.50 .000*** b_𝟑 SHARES (+) .0168 11.19 .000*** b_𝟒 DUAL (+) -.0114 -2.18 .029** b_𝟓INDUSTRY (+) .0109 3.52 .000*** b_𝟔 SP500 (-) -.0414 -8.49 .000*** b_𝟕CRISIS (+) .0475 11.69 .000*** b_𝟎 .0962 -3.68 .000*** N 57,980 F-Value 129.78 Probability > F 0.0000 Adjusted R-squared 0.0154 (»1,5%)

The symbol *, **, *** indicate significance at 0.1; 0,05; 0.01 percentage level, respectively.

The FIRMSIZE, AVOL and SHARES variables are winsorized at 1% and 99% level to control for outliers.

In this section the main results of this analysis will be illustrated. Table 5 indicates the appropriateness of the regression analysis. The adjusted R-squared represents the portion of variance in the abnormal volatility that can be explained by the independent variables indicated in the first column. These variables explain 1.54% of the variation in the abnormal volatility. The model explains a small portion of the abnormal volatility, but the suitability of the regression model is indicated by the F-value. This shows that the model can statistically significant predict the dependent variables. Now the results for the independent variables will be discussed.

In the second column of Table 5, the expected influences on the volatility are shown based on prior literature. Three variables instantly confirm their influence. First, the results confirm the notion that firms faced more volatility during the financial crisis resulting from increased market scrutiny. Second, the findings show that the number of shares traded have highly significant influence on the volatility reaction around the earnings announcements. This confirms the theory that volatility originates from the imbalance of trade orders on the stock exchange. Three, the

(29)

29

industry influences as proposed by Landsman and Maydew (2002) have also a significant positive influence on the volatility reaction from firms.

Besides to confirmations, deviations were also found compared to the expected results. The first finding which deviates from prior literature is influence of the FIRMSIZE on the volatility. Bamber (1986) and Landsman and Maydew (2002) argue that the size of a firm has a mitigating effect on the volatility (i.e. information provision), but the results of this research indicate there is a positive relation (0.013) between SIZE and AVOL. In assessing this result is acknowledged that there is partial overlap between the control variables firm size and SP500 variable because both variables capture the firms size and the enhanced information availability. Nevertheless, the S&P 500 variable is included because prior research emphasized the additional market scrutiny and the maintenance of higher levels of disclosure compared to relative large non-S&P 500 listed firms. Accordingly, with the expectations shows Table 5 a negative significant effect between listing on the S&P 500. Surprisingly, this result highly contradicts the findings from the firm size on the volatility. In a supplementary regression analysis, the SP500 variable is excluded to eliminate the overlapping influence of the control variables. The results of this added analysis are displayed in Appendix B indicating a coefficient of .0057899 between FIRMSIZE and AVOL. In comparison with the results in Table 5, this shows a decrease in the positive influence of the firm size on the volatility. In Appendix C, only the S&P 500 firms are used as sample, the findings show a negative relation (-.0061, significant at 0.10 percent level) between the influence of firm size on the volatility. Taken this all together, these different results can be interpreted as follows: for firms listed on the S&P 500 the results confirm the increase of information availability and decrease of volatility around earnings announcements. The incorporation of both the firm size variable and S&P 500 variable separate the influence of the largest 500 firms from the firms below this threshold (i.e. firm 501 till 2,000). Providing evidence that for firms between the smallest of the sample and the firms just below the threshold of the S&P 500 there is an increase in volatility, thus contradicting the findings from previous studies, but when firms become large enough for the S&P 500 there is a decreasing function. This corroborates the notion of Anderson, Duru and Reeb (2009) of the additional increased market scrutiny and information availability for S&P 500 firms.

The second deviation from the expectations is the effect of the DUAL share structure on the volatility of the firm. Chau and Grey (2002) and, Ho and Wong (2011) show a positive relation between the presence of dual class share structure and less transparency (i.e. more volatile stock returns). The findings show there is a negative association between the presence of a dual share structure and the volatility reaction. The difference of mean test and the correlation matrix show

(30)

a positive relation between family firms corresponding with the regression analysis. This can potentially be explained through essence of dual class share structures; when a family firm grows, chances are that the company issues shares to finance its activities (e.g. investments/ mergers). This results in dilution of the portion of shares of the family in the company, but by creating dual class structures the family can keeps in control of the family firm when the company becomes larger. To confirm this relation, this research performed an additional test with an interactive variable between the dual class share variable and the family firm variable. Appendix B displays the results from the additional analysis. These results indicate a very slight difference between the coefficient of the interactive variable (-.0115) and the initial dual class variable (-.0114). This slightly difference suggests that the predicted combination results in increased transparency. However, the difference is too small to fully confirm this theoretical relation and thus provides no convincing evidence that family firms with a dual class structure provide more or better information, indicated by a lower volatility. This result only suggests the positive influence on the information provision.

The main purpose of this research is to investigate the difference in information provision between family firms and non-family firms. The results described in Table 5 of the regression analysis indicate there is no statistical difference between the abnormal volatility of family businesses and non-family businesses. The coefficient 0.0049 (almost nihil) indicates there is a very small positive relation between the abnormal volatility and the family firms. In contrast the correlation and the difference of mean test indicate a negative relation between family firms and the abnormal volatility, shown in Table 4 and 3. Thereby, looked at the results of the most affiliated variable with founding family ownership, as described in the previous paragraph, dual class share structures indicate this variable a significant negative influence, which suggests that family influence potentially leads to decreased volatility (i.e. more information provision) Notwithstanding this evidence cannot be concluded that family firms have less volatile share prices around earnings announcements, providing evidence for more or better information provision by family firms. This result only suggests that family firms are less volatile and thus disclosure more information in comparison with non-family firms in the United States.

In addition, several alternative explanations are explored to clarify the findings. The first potential explanation for the findings is the dispersion in the sample used. The skewed distribution between family firms and non-family firms in size could explain the findings because prior literature emphasizes the difference in information availability between relative small and large firms (Bamber, 1986; Landsman & Maydew, 2002; Chow & Wong-Boren, 1987). Therefore, the sample is changed to only S&P 500 listed firms to create a more homogeneous sample with regard to firm size and information availability (Anderson & Reeb, 2003; Hutton, 2007). The descriptive

The influence of family ownership on the information provision