IPO underpricing and family controlled firms : is IPO underpricing different for family controlled firms?

Master Thesis

IPO underpricing and family controlled firms

Faculty of Economics and Business

MSc in Business Economics

Specialization: Finance

- Is IPO underpricing different for family

controlled firms? -

Name: Britt Dams

Date: July 6

2016

Student Number: 10214364

Statement of originality

This document is written by Britt Dams, who declares to take full responsibility for the contents of this document.

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

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

Abstract

This thesis examines whether there is a difference in IPO underpricing between family- and non-family controlled firms. We argue that some firm characteristics related to corporate governance may affect information asymmetry, agency problems and management incentives, which in turn affects underpricing. Performing regressions on a sample consisting of both family- and non-family

controlled U.S. firms that conducted an IPO between 2000 and 2015, we try to find empirical evidence for a difference in underpricing. Our models include variables on corporate governance, variables related to the issue and variable on other firm characteristics. From our estimated models, we only find evidence that direct ownership is the single independent variable able to explain a variation in

underpricing. Our results indicate that higher direct ownership increases underpricing for family controlled firms

Table of Contents

I. INTRODUCTION

………..………..………...5

II. LITERATURE REVIEW

………..………..………...8

Ownership………8

Corporate governance and underpricing……….8

Owners’ wealth and underpricing………..……….9

Competitive advantage and firm performance………..10

Ownership and control………..………..………..11

Determinants of underpricing………...12

Asymmetric information………...13

A. Winner’s curse………..………..……….….13

B. Information revelation………..………13

C. Signalling………..………..………..14

D. Agency problems………..………14

Behavioural theory……….………..……….15

Control theory………..………..………..……….15

Agency problems of ownership and control……….16

III. HYPOTHESES

………..………..………...17

IV. DATA

………..………..………..………19

A. Collection……….19

B. Variables……….………..20

C. Descriptive statistics……….21

V. METHODOLOGY

………..………..………..………29

VI. RESULTS

………..………..………..………32

VII. ROBUSTNESS

………..………..………..………39

VIII. CONCLUSION

………..………..………..………..41

IX. REFERENCES

………..………..………..………..43

X. APPENDIX

………..………..………..………...46

A. Technology dummy………..………..………..…………46

B. Variables……….………..46

C. Mean % direct ownership by type of control………..……….47

D. Differences in mean underpricing by type of control………..47

E.. RVF-plot………..………..………..49

F.. Subsamples by Proceeds………..49

I. Introduction

Private firms seeking for external capital can make the decision to become publicly traded. When a private firm goes public and sells its shares for the first time, this is called an initial public offering (IPO). The path towards an IPO and the IPO itself can be quite complicated. This is where the issuing firm gets help from an underwriting firm. The underwriter acts as an intermediary and advices the issuer in determining the type of shares, the offering price and time of offering (Loughran and Ritter, 2004).

IPOs allow firms to acquire new capital and expand shareholder base. However, a potential problem in IPOs is mispricing the shares. For instance, setting the offering price too low, resulting in a smaller amount of raised capital. Another problem might be loss of control or ownership for firms that wish to keep ownership within the family (Daugherty and Jithendranathen, 2012).

When Twitter went public, it underpriced its stock (www.newyorker.com)1_{. The company sold 70}

million shares for $ 26.00 whereas it could have sold the shares for $45.00. This underpriced IPO meant that Twitter brought in less capital than it could have.

The puzzle of the underpricing of IPOs is a well-studied phenomenon in financial literature. The many theories trying to explain underpricing can be split into several categories: asymmetric information, control theories and behavioural theories (Ljungqvist, 2007).

There is an extensive amount of literature that focuses on explaining the determinants of IPO underpricing. Nevertheless, there is not much literature that combines IPO underpricing with

characteristics of firm structure. Literature on firm performance and family ownership does exist, but to my best knowledge, high quality literature that examines the relation between IPO underpricing and family controlled firms is scarce. Presumably, whether a firm is family controlled or non-family controlled makes a difference for underpricing due to different firm characteristics.

Andrew Steen (2014) argues that the impact of going public is different for family controlled firms. Just like non-family controlled firms they experience positive impacts as increased liquidity, better valuation of the company and increased business performance. New capital might enable the firm to realize the desirable growth (Steen, 2014).

However, due to their close ties with the firm it is plausible that the possible drawbacks of going public will be experienced differently for family controlled firms than non-family controlled firms. It is inevitable that the family controlled firm will lose at least some of its control. Outside

shareholders will also have a say. This involvement of outsiders can lead to a loss of identity of the firm. Family controlled firms have a more strategic long term focus and care about the success of the

firm when the next generation will manage the firm (Kenyon-Rouvinez2_{, 2016). Therefore, the} possible loss of identity or changed strategy will affect family controlled firms more.

Additionally, family controlled firms tend to be less transparent and more information sensitive (Deloitte, 2015)3_{. Going public forces the firm to increase transparency and therefore sharing} information. Family controlled firms may have more trouble meeting disclosure requirements. Concluding, there are reasons to believe the IPO process and therefore underpricing will be different for family controlled firms. It is likely that family controlled firms have more and different incentives to mitigate the negative impacts of going public and therefore undergo a different IPO process.

This thesis studies the effect of ownership structure and the related governance on IPO underpricing for US firms between 2000 and 2015. We start with a literature review discussing the firm

characteristics of family controlled firms, that may be related to IPO underpricing. Also, we present multiple theories from prior literature, concerning the determinants of underpricing that will be used in the empirical analysis as control variables.

Then, to see whether the theories found in the literature apply in practice, we will perform a data analysis to empirically test for differences in underpricing between family- and non-family controlled firms. Using ownership data of publicly listed firms from the United States, a distinction between family controlled firms and non-family controlled firms is made. Additional data on the managers and percentage of ownership of the firm provides insights on the governance structure of the firm.

Furthermore, we gather data regarding the initial public offering of the firm and other firm characteristics of interest to conduct our analysis. Ownership data and IPO data is combined to perform a cross-sectional comparison of family controlled firms and non-family controlled firms. Using this data, a multiple linear regression model is formed and tested using the OLS technique. The regression results will tell us more on the relation between underpricing and firm structure. My research will provide insights to investors on the valuation of IPO performance of family-controlled firms. The rationale is that firm structure can affect information asymmetry and agency problems between owners, underwriters and investors.

Prior literature has been concerned with providing explanations of why underpricing occurs. Underpricing might be explained because it is a tool to collect information (Benveniste and Spindt, 1989), because underwriters sell a differentiated product (Liu and Ritter, 2011), because it is used to signal the true value of the firm (Ibbotson, 1975 and Allen and Faulhaber, 1988) or because it is a means to retain control (Brennan and Franks, 1997). Yet no particular distinction in the type of firm

2 _{Dr. Denise Kenyon-Rouvinez, Head of International Family Business & Philanthropy at RBS Coutts Bank Ltd} in Switzerland

3_{http://www2.deloitte.com/content/dam/Deloitte/xe/Documents/AboutDeloitte/mepovdocuments/mepov17/}

and its governance is made. There is no focus on the role of firm characteristics that can affect agency problems, information asymmetry and the firm’s incentives regarding underpricing.

Therefore, I would like to further investigate the determinants of IPO underpricing and the role of firm structure on IPO underpricing. The research question to this will be: Is IPO underpricing different for family controlled firms?

The remainder of this paper is structured as follows. Section II contains a literature review, section III states the hypotheses, section IV covers the data collection, variables used and descriptive statistics, in section V the methodology is described, section VI contains the regression results, robustness checks will be discussed in section VII and finally, section VIII concludes, discusses on our findings and provides directions for further research.

II. Literature Review

Ownership

The main difference between family controlled firms and non-family controlled firms is the separation of control between the owners and management. A family controlled firm is defined as an enterprise where the founder or acquirer of the firm (or its relatives) owns at least 25% of the decision making rights (European Commission, 2009).

In contrast to a non-family controlled firm, family members in a family controlled firm will participate as an owner as well as in a management position. This difference in corporate governance

characteristics and its influence on IPO performance has not been covered much in the IPO

underpricing literature. This thesis debates that different corporate governance structures of family- and non-family controlled firms exhibit different agency relationships and information asymmetry in the stage towards an IPO.

During the process of going public, evaluating the ‘true’ value of the firm is subject to information asymmetry, which potentially leads to agency problems (Filatotchev and Bishop, 2002). As stated by Ding and Pukthuanthong-Le: “behavioural differences between family and non-family firms further complicate the investors’ efforts to evaluate the quality of family firm IPOs” (2009, p.56). Thus, we argue that these potential agency problems manifest themselves in the degree of underpricing and are different for family controlled firms.

Families embody a unique class of shareholders, according to Schleifer and Vishney (1986). They hold poorly diversified portfolios, have senior management positions and have a long term perspective regarding investment, due to the multiple generations. These characteristics of family firms,

potentially lead to differences in performance in comparison to non-family firms.

The subsequent part describes some characteristics of (family) firms that can affect firm value and the IPO process.

Corporate governance and underpricing

In their research, Filatotchev and Bishop (2002) question what factors affect the corporate governance mechanism of a firm in the phase of an IPO. The relationship between corporate governance

characteristics and the performance of an IPO is studied. The extent of IPO underpricing is possibly affected by strategic ex ante board selection and financial interests, as is claimed by the authors (Filototchev and Bishop, 2002). Moreover, these factors are not seen as exogenous, but related to characteristics of top executives. Agency problems arise in the IPO process due to information asymmetry about the true value of the firm. These problems are costly and are revealed in the amount

of underpricing. In line with the assumption of non-exogenous factors, agency costs and thus

underpricing can be reduced by board diversity and ownership of non-executive directors (Filototchev and Bishop, 2002).

Owners’ wealth and underpricing

Ownership and underpricing is studied by Habib and Ljungqvist (2001), who believe that owners care about underpricing to the degree they are affected by it. According to the authors, owners can affect the level of underpricing by promotion. Habib and Ljungqvist (2001) combine the model of Rock (1986), which states that underpricing is necessary for uninformed investors to participate in the offering, with Carter and Manaster’s (1990) view that promotion of the issue helps increasing the fraction of uninformed investors to participate. Consequently, promotion mitigates adverse selection problems and as a result, less underpricing is necessary. It follows that promotional activities and underpricing can be seen as substitutes. Hence, when the costs of promotion are relatively higher compared to the benefits, it is more likely that the issue will be underpriced. Therefore, Habib and Ljungqvist (2001) argue that owners seek to minimize their wealth losses when making a tradeoff between underpricing and promotion of the issue. In other words, issuers optimize at the margin. Habib and Ljungqvist (2001) provide the argument that the amount of shares sold by the owners at the IPO forms a link with how much owners care about underpricing. When owners sell many shares they will suffer substantially from underpricing and vice versa.

In contrast with Rock’s (1986) model, Habib and Ljungqvist (2001) assume that the fraction of

informed and uninformed investors is not exogenous. Moreover, the issuer is assumed to influence this fraction, making it endogenous. The issuer can exert greater effort and higher promotion costs such as making use of an underwriter with higher reputation (Habib and Ljungqvist, 2001). This will support more uninformed investors in their decision to participate in the IPO. Underpricing will then decrease. Lower underpricing implies higher promotion costs for the issuer. The issuer will make a comparative assessment to choose the combination of underpricing and promotion costs and how this affects his wealth (Habib Ljunqvist, 2001). Their empirical research shows, the larger the number of shares being offered, the more promoting costs are being made, which reduces underpricing.

Leiterstorf and Rau (2014) address the noneconomic utility of a family firm, known as the socio-emotional wealth (SEW) to explain IPO underpricing and combine this with the behavioural agency model. Going public is likely to lead to reduced influence of the family and therefore harms the family firm’s SEW. On the other hand, the increased economic utility of becoming a public firm might offset the loss of non-economic utility.

The trade-off between economic and non-economic utility of family firms can be assessed through IPO underpricing. Namely, existing shareholders will usually sell shares for the highest price possible, however there are reasons to underprice shares, that serve non-economic utility. These non-economic

utility advantages of underpricing come from reduced ownership concentration, reduced risk of

lawsuits and the reduced risk of a failed IPO. Ibbotson (1975) developed the “litigation risk hypothesis” which means that underpricing reduces the probability of a lawsuit. The informational cascade theory of Welch (1992) states that without underpricing, some investors won’t participate in the IPO and therefore other investors abstain as well. A lack of underpricing thus implies a risk of a failed IPO. The last two theories both imply that higher underpricing reduces risk and therefore reputational damage which is of greater importance to family firms (Leiterstorf and Rau, 2014). The most important advantage to preserve a family firm’s SEW of retained control comes form ownership dispersion (Leiterstorf and Rau, 2014). To generate an oversubscription, the issuer can willingly underprice newly issued shares (Brennan and Franks, 1997). This leads to a reduction in non-family ownership concentration. Leiterstorf and Rau (2014) conclude that in the perspective of SEW, the above mentioned behavioural agent models insinuate that, by underpricing their shares, family firms give up part of their economic utility in order to minimize the threats to their SEW. Therefore, the authors (2014) state that underpricing is higher for family firms.

Competitive advantage and firm performance

Learning about firm performance of family controlled firms is important since it is an indicator of the firm’s value. The more transparent the firm’s value, the easier it is for the underwriter and investors to determine the ‘true’ value of the firm. Consequently, new shares can be priced better, according to their ‘true’ value.

Prior literature mainly portrays family firms as self entrenching at the cost of minority shareholders. In family controlled firms, owners draw scarce resources away (Demsetz, 1983), the controlling

shareholders extract private benefits (Schleifer and Vishny, 1997) and management serves family interests (DeAngelo and DeAngelo, 2000). In addition, Anderson and Reeb (2003) state that using a restricted labour pool, by keeping management positions within the family, can be disadvantageous. However, family influence can also provide competitive advantages concerning firm performance. Family owners have considerable incentives to reduce agency conflicts and firm value maximization (Demsetz and Lehn, 1985, cited in Anderson and Reeb, 2003, p. 1305). These incentives are attributed to the close link between firm welfare and family wealth. As a result, they wish to maximize firm value -and consequently their own utility- and have strong incentives to monitor management. Anderson and Reeb (2003) declare that families are potentially good in monitoring management due to their long term tenure in the firm, giving them superior oversight. The long term presence of family members potentially results in long term management horizons compared to non-family controlled firms. It suggests that family firms are more motivated to invest in long term projects. Moreover, survival of the firm is important for families as they plan to pass on the firm to next generations. Therefore, family members will be concerned with maintaining and improving firm value, rather than

consuming their wealth. Furthermore, family members in particular feel responsible for firm

performance, as they experience it to be a substantial part of their own well-being (Davis, Schoorman and Donaldson, 1997, cited in Anderson and Reeb, 2003, p. 1307). Anderson and Reeb (2003) also mention that family firm ownership is closely related to reputation effects. In family controlled firms, it is very likely that the the same persons will deal with third parties for a longer period compared to non-family controlled firms. It follows that economic consequences formed by the reputation of a family controlled firm are of a longer lasting nature. As a consequence, the firm is more concerned with its reputation, which gives incentives to improve performance of the firm (Anderson and Reeb, 2003).

Family firm performance can be best explained in a qualitative- rather than a quantitative way, according to Habbershon & Williams (1999). They apply a resource based view (RBV) to assess the strategic advantages of family firms. RBV claims that in a sense, firms are heterogeneous and it is the idiosyncratic recourses of the firm that can possibly lead to a competitive advantage. In family firms, family involvement leads to a collection of resources that are unique to the firm, referred to as ‘familiness’ (Habbershon and Williams, 1999). In line with RBV, the competitive advantage of idiosyncratic resources come from the ‘familiness’ of the firm. Hence, when comparing firms at a behaviour level, family firms can perform better in some ways, due to their different behaviour and social impact (Habbershon and Williams, 1999).

Carney (2005) also uses qualitative corporate governance characteristics to explain the competitive advantage of family controlled firms. Organizational qualities able to enhance value creation are incorporated in the firm’s corporate governance structure. The governance system of unification of ownership and control creates three propensities which the author describes as parsimony, personalism and particularism. Parsimony comes from the fact that management decisions are made with their own money, the family’s personal wealth (Carney, 2005). In general, people tend to be more prudent if it concerns their own money. We therefore assume that, in a governance system where ownership and control are combined, agency problems are reduced. Personalism is described by Carney (2005) as the personalization of authority, meaning that the firm is less constrained in decision making since to authority lies with the person that is a manager as well as an owner.Particularism refers to the owner-managers’ view of the firm as their ‘own business’. Family control rights allow the family to

discriminate and intervene in certain affairs. (Carney, 2005). Owner managers exercise their authority in the way they want.

Ownership and control

According to Fama and Jensen (1983), the decision process of a firm has the components of decision management and decision control. Decision management consists of initiation and implementation, decision control consists of ratification and monitoring (Fama and Jensen, 1983).

When decision managers do not bear a substantial amount of the risk of their decisions as major residual claimants, agency problems emerge. If not controlled for, they are likely not to act in the interest of the residual claimants (Fama and Jensen, 1983). The authors investigate when it’s more efficient to separate risk-bearing and decision functions in stead of combining these functions in the same agents.

In a company where relevant information for decision making is concentrated in one or a few agents, it is efficient to assign both decision control and decision management to these agents (Fama and Jensen, 1983). A family firm is an example of a firm with concentrated information. To overcome the agency problem of the combined decision functions of agents, residual claims should be restricted to decision agents. Fama and Jensen (1983) argue that this restriction can be an alternative for control devices, which are costly. Decision management and control does not need to be separated to reduce agency problems if there are agents - holding residual claims- that have a special relationship with decision agents (Fama and Jensen, 1983), for example within family firms. In family firms, due to the close ties of family members, there is a great scope of long term exchange with one another. This enables better control for agency problems in the form of monitoring and correcting the decision agents (Fama and Jensen, 1983).

Determinants of underpricing

The puzzling phenomenon of underpricing of newly issued shares has been given much attention in the IPO literature. Many researchers have tried to rationalize the underpriced shares of initial public offerings. One of the researchers that is part of the founding literature on this topic is Roger G. Ibbotson. Ibbotson (1975) already stated that new issues are underpriced, given their positive initial performance. High positive first day returns of IPOs averaging 24% during 1993-2008 were also found by Liu and Ritter (2011). According to Loughran and Ritter (2002), US firms that went public between 1990 and 1998, left over $27 billion ‘on the table’, while investment bankers received $13 billion in underwriting fees. Also in more recent years, Ritter (2016) finds the average first day returns of IPOs between 2001-2015 to be 13.9%.

Since these returns are so high, it cannot be assumed to be compensation for risk, as the period is only one day (Grinblatt and Hwang, 1989). This indicates there should be other reasons underlying the underpricing of initial public offerings. Underpricing is a well studied manifestation in the IPO literature. Many studies aim to explain underpricing by developing new models and theories. These theories can be mainly categorised into the following: asymmetric information, behaviour and control. Majority of the literature points out the first-order effect of information frictions on underpricing (Ljungqvist, 2007). Yet, there is no exclusive answer to this phenomenon as it may be a combination

of several factors including factors that are not easy to measure.

To understand how underpricing may be different for family controlled firms, we should look at the determinants of underpricing and how these determinants are related to firm structure. In order to make the connection between firm structure and IPO underpricing, it is important to understand what causes underpricing. That way we can make better assumptions on which determinants play a different role or have a different weight in the underpricing of family controlled firms versus non-family controlled firms.

The following part of this literature review will discuss the main theories of IPO underpricing.

1. Asymmetric information theory

The three main parties in the IPO process are the issuer, the underwriter and the investors. Asymmetric information theories assume that one of these parties has additional information, unknown to the other parties. This asymmetry may contribute to underpricing of the IPO. A. Winner’s curse

One of the most famous asymmetric information models comes from Kevin Rock (1986). His model considers informed- and uninformed investors that have different information on an asset to be issued with an uncertain value. Informed investors will bid on good priced IPOs and not on the bad ones, while uninformed investors will bid randomly. Ultimately, this enforces the winner’s curse on uninformed investors, where informed investors will get majority of the ‘good priced IPOs’ and the uninformed will get the more unattractive IPOs (Rock 1986). In the most extreme case, uninformed investors will be rationed in underpriced IPOs but will receive total allocation in overpriced IPOs (Ljungqvist, 2007). The latter leads to negative expected results, leaving the uninformed investors unwilling to participate. IPO underpricing is a way to attract uninformed investors and can be seen as a compensation to ‘the uninformed’, accordcing to Ljungqvist (2007).

In line with the winner’s curse theory and the information asymmetry theory of Kevin Rock (1986), Michaely and Shaw (1994) find that there is a relation between the heterogeneity of investors'

information and IPO underpricing. Underpricing is found to be significantly smaller in such a market, where the winner's-curse for the uninformed investors is not that pronounced (Michaely and Shaw, 1994).

B. Information revelation

In general, informed investors have an incentive to misrepresent their information. This way, the investor expects to benefit from the lower offer price set by the underwriter. Therefore, the

underwriter should come up with ways to incentivize the investors to reveal their information (Ljungqvist, 2007). In contrast to Rock (1986), Benveniste and Spindt (1989) state that underpricing can be lower by using the underwriter’s access to informed investors to collect information in advance. Benveniste and Spindt (1989) define the IPO process as a repeated game where underwriters can ‘change the rules’ to obtain information and thereby reduce underpricing. Share allocation can be used as a tool to mitigate the incentive to misrepresent positive information. When investors are regularly given priority in the allocation of shares, this can be used against them as a lever (Benveniste and Spindt, 1989). Investors should reveal information or otherwise they will lose their allocation priority. Nevertheless, some underpricing is necessary to compensate investors for revealing positive

information. How much this compensation should be, depends on how much profit is expected to be foregone by revealing the information (Benveniste and Spindt, 1989).

C. Signalling

Within the literature on the influence of asymmetric information on underpricing of IPOs, there is also a theory on signalling. To construct this theory, Rock’s (1986) assumptions on asymmetric

information between issuers and investors are reversed. In case a firm has better information on its financial prospects than investors have, for instance on future cash flows and risk, underpricing can be used to ‘signal’ the firms true value (Ljungqvist, 2007). Although underpricing is a costly signal, it may yield the firm better conditions to raise capital in case of a secondary offering. This is in line with one of Ibbotson’s (1975) explanations of IPO underpricing, which suggests that underpricing is a way to “leave a good taste in investors’ mouths”.

Following Ibbotson (1975), Allen and Faulhaber (1988) assume that the firm itself is better informed over its prospects than the investors. Their article sketches a situation in which there are ‘good’ and ‘bad’ firms. ‘Good’ firms using underpricing as a costly signal to reveal their type, are believed to recoup this cost from subsequent issues. The risk of detection as well as the inability to earn back the costly signal, will abstain ‘bad’ firms from mimicking the ‘good’ firms (Allen and Faulhaber, 1988). This results in a separating equilibrium where only ‘good’ firms use underpricing as a signal.

D. Agency Problems

The involvement of several parties in the IPO process is likely to lead to frictions over the share price, either caused by the allocation of shares or by the level of information revelation. Ultimately,

underpricing can result in principle agent problems.

The paper of Baron (1982) examines the relationship and interaction between an issuing firm and an investment banker (underwriter), based on an information asymmetry. The underwriter is assumed to have better information about the market demand for the new shares to be issued (Baron, 1982). This

demand will for the most part depend on the costly ‘distribution effort’ made by the underwriter to persuade investors to buy shares and to influence their expectations by providing information to the market (Baron, 1982). Since the underwriter has an incentive to misreport its superior information, the issuer will not be able to reach a first-best price solution. The underwriter must be compensated for using its superior information and consequently, the issuer must lower the offer price (Baron, 1982) and therefore engage in underpricing.

Ljungqvist (2007) translates the findings of Baron in stating that when there is more uncertainty of a firm’s value, there is more information asymmetry resulting in a higher valuation of the underwriter’s superior information and thus underpricing will be greater (Ljungqvist, 2007).

2. Behavioural theory

The importance of other theories -besides asymmetric information- is emphasized by Ritter and Welch (2002). The authors declare these theories unable to fully explain average first-day returns of 65 percent during the Internet bubble. Therefore they believe that agency conflicts, share allocation issues and behavioural explanations should be better investigated in future literature.

Loughran and Ritter (2002) apply prospect theory trying to explain the phenomenon of underpricing and in particular why issuers don’t seem to be bothered with it.

According to prospect theory, it is the increase in wealth, rather than the level of wealth that matters for the issuer. Applying the theory to the IPO market, it predicts that the wealth loss from underpricing and the larger wealth gain from a price jump will be combined by the issuer, resulting in a net increase of wealth (Loughran and Ritter, 2002). Issuers put the opportunity costs of underpricing in the

perspective of their increase in wealth. From that point of view, Loughran and Ritter (2002) state that underpricing can be seen as the indirect cost of having an underwriter and issuers don’t get upset about leaving money on the table.

3. Control theory

Taking the firm public can have major consequences for current management. Additional regulation and obligations and the presence of external shareholders can affect the company’s structure and therefore control of current management a lot. This gave rise to another theory from the IPO literature, namely underpricing as a means to retain control.

Brennan and Franks (1997) state that underpricing is an effective mechanism for directors of the firm wishing to maintain control after the IPO. In their paper, it is assumed that the directors of the firm

want to prevent a hostile takeover when going public. Brennan and Franks (1997) argue that underpricing the issue will lead to oversubscription. As a consequence, the issuer can strategically ration the allocation of shares and discriminate against large applications, thus preventing a hostile takeover. In addition, it causes shareholder dispersion, which is favourable for management whishing to maintain their private benefits of control. This dispersion reduces the incentives of new

shareholders to monitor the management and is called the reduced monitoring hypothesis (Brennan and Franks, 1997). The researchers (1997) find empirical evidence for underpricing to be a mechanism to achieve a diffuse outside shareholding.

Agency problem of ownership and control

For a firm to go public, this means the transformation to a situation where ownership and control will be (partly) separated. Transfer of ownership has its implications on management’s incentives

(Ljungqvist, 2007). Management may only be concerned with maximizing their own expected private benefits rather than maximizing shareholder value (Jensen and Meckling, 1976, cited in Ljungqvist, 2007, p. 409). Agency problems can arise between managing and non-managing shareholders.

In contrast to Brennan and Franks (1997), Stoughton and Zechner (1998) argue that monitoring by outsiders raises firm value. Monitoring will be beneficial to all shareholders, but only large investors are capable of monitoring and will do so as far as is optimal for them (Stoughton and Zechner, 1998). Small investors on the contrary, do no monitor the firm and free-ride on the monitoring activities of large investors. Ultimately, underpricing may allow a favourable share allocation for large investors, so they have a large stake in the firm which encourages monitoring and therefore raises firm value, as is stated by Stoughton and Zechner (1998). The amount of underpricing is thus compensated by higher intrinsic firm value.

III. Hypotheses

The next section aims to test our assumptions on the relation between IPO underpricing and firm structure. Based on the different theories described in the literature section, we investigate whether there is a difference between IPO underpricing of family controlled firms and non-family controlled firms. Also, we try to identify the determinants of underpricing and if these account for the differences in underpricing between family controlled firms and non-family controlled firms. Several hypotheses are formed to empirically test these differences.

We classify family controlled firms as firms where the owning family members can influence the decision making process through certain control rights and have at least 25% of these rights (European Commission, 2009)4_{. This classification is consistent with the classification of the Osiris database,}

where control is traced by calculating voting rights.

First, a very general hypothesis will be tested in order to see whether there is any difference at all. Then some hypotheses will be formed based on asymmetric information and agency problems that affect underpricing.

1. Family controlled firms experience lower underpricing than non-family controlled firms.

As described in the preceding literature section, the overlap between owners (known as the principles) and managers (known as the agents) in a family controlled firm, reduces the costs of agency problems that are manifested in the amount of underpricing. Family controlled firms are also known for their long-term perspective and according to Habib and Ljungqvist (2001), they care about maintaining the family’s wealth. Consequently, it is reasonable to believe they have more incentives to lower

underpricing as much as possible.

It may be the case that the opposite is true, namely family controlled firms experience higher underpricing. The conclusions from the literature on underwriting are rather divided. Some theories suggest that family firms willingly underprice. Leiterstorf and Rau (2014) conclude that family firms underprice to minimize the threats to their SEW. Also, family controlled firms typically wish to maintain their control over the firm and according to the control theory of Brennan and Franks (1997), underpricing is an effective mechanism for directors to do so.

4

2a. The higher the age of the firm, the lower will be the underpricing. 2b. Higher firm age lowers underpricing more for family controlled firms

The age of the firm could be a proxy of information asymmetry, as we expect information asymmetry to decrease with time, because it is easier to evaluate a firm based on more previous years. Therefore, the older the firm, the more information is available, lowering information asymmetry and therefore underpricing is lower. Besides higher information asymmetry, younger firms have higher ex ante uncertainty which also increases underpricing, as is reasoned by Beatty and Ritter (1986). Since family controlled firms are regarded to have higher information asymmetry, we believe the effect of reduced information asymmetry -due to higher age- on underpricing is bigger for family controlled firms.

3. Underpricing decreases for family controlled firms when the ultimate owner is also a manager

When the owner is also a manager, we assume that interests are more aligned, resulting in less agency problems and thus lower underpricing. Aggarwal et al. (2003) argue that managers wish to maximize their personal wealth in the long run and therefore engage in underpricing. However, in family controlled firms, family members are present in the management team and will hold their shares to maintain control. They have no motive to sell their shares for a higher price and for that reason, they don’t aim at an increase in the share price. Consequently, management has no incentive to underprice new shares.Contrary, the hypothesis may be rejected if the theory of Filatotchev and Bishop (2002) holds. They state that agency costs and therefore underpricing can be reduced by board diversity and ownership of non-executive directors.

4. Higher direct ownership of the ultimate owner decreases underpricing more for family controlled firms

We assume that a higher percentage of direct ownership by the family increases the personal ties with the firm, hence they will exert more effort to reduce underpricing. The direct ties with the firm will strengthen the incentives to sell shares for the highest possible value in order to maintain the family’s wealth. Additionally, in line with agency theory, a higher percentage of ownership lowers agency problems and thus underpricing.

IV. Data

A. Collection

Data on ownership is found in the Osiris database. From this database we extract a large sample of firms, that are publicly listed in the United States. Next, we follow the data description of Lins et al. (2013) and filter in this database for firms that have an ultimate owner. An ultimate owner is identified as an entity controlling the firm directly at a defined threshold or via a control chain whose links exceed that threshold. The database allows us to self select the definition, i.e. the threshold percentage of an ultimate owner. It can either be 25% or 50%. According to the paper of Lins et al. (2013) we also choose the ultimate owner to own 25% or more.

For the resulting sample we extract: company name, percentage of direct ultimate ownership, type of shareholder, ultimate owner also a manager, Central Index Key (CIK) number, ticker symbol, IPO date and date of incorporation.

An important aspect in our dataset, necessary to answer our main question, is the distinction of family controlled and non-family controlled firms. The data on type of shareholder will help us to identify the the type of ultimate owner in order to classify which firms are family controlled and which are not. Osiris makes a distinction between several types of shareholders including: bank and financial

companies, one or more named individuals, foundations, industrial companies, private equity firms etc. Just as Lins et al. (2013) do in their paper, we classify firms controlled by a family, as firms of which the shareholders are of the type: one or more named individuals. Hence, the remaining shareholder types are classified as non-family controlled firms. Consequently, we have a sample consisting of both family- and non-family controlled firms.

Subsequently, this dataset is sized down to a dataset of firms with an IPO date that lies within 1999 and the current months of 2016. We use a time frame that is slightly bigger than the intended 15 year frame of 2000-2015. This way we can correct for firms that are indeed in the 2000-2015 time frame according to the offer date denoted in the Thomson One database in the next phase. Once this is done, we are left with a primary sample of 1281 firms.

Since the company identifiers of the Osiris database do not match with the Thomson One database, we first convert the CIK numbers to Committee on Uniform Security Identification Procedures (CUSIP) numbers in Compustat, accessed through Wharton Research Data Services (WRDS). Then, we import the CUSIP numbers in the Thomson One database to extract IPO data. In addition to the CUSIP numbers we specify that the issuer’s nation is United States of America, the issue type must be IPO,

following Loughran and Ritter (2002) weexclude American Depository Receipts as well as REIT’s and the issue date must be between 2000 and 2015 (year-end).

From Thomson One we obtain the following variables: company name, main SIC code, ticker symbol, issue date, offer price, stock price at close of offer/1st trade, stock price after one week, total shares offered, shares outstanding after the offering, underwriting fee per share, date founded, type of security, Spin-Off flag and 9-Digit CUSIP.

A type of IPO that should be treated differently is an equity carve-out. In an equity carve-out, a firms sells a part of its business in a subsidiary through an IPO. This process is also known as a partial spin-off and allows the firm to retain some control. In the perspective of a family firm, it allows the firm to resign some of their non-core business without putting the rest of the firm at risk (Deloitte, 2015)5_.

We therefore eliminate these equity carve-outs from the sample as it is likely that they exhibit different underpricing behaviour.The ‘equity carve-out spin-off indicator’ variable allows us to manually exclude firms that are characterised as an equity carve out.

Lastly, both samples are merged to create the final dataset which contains both ownership and IPO data. Missing values of IPO data are completed, if possible, using Yahoo Finance6_{, NASDAQ}7 _{and the}

firm’s websites. Missing values that remained, were treated differently. For unknown percentages of direct ownership, we assumed there to be no direct ownership and replaced missing values by zero. The same replacement was done for underwriting fee. In case of an absent overhang ratio, due to missing values of shares outstanding after the issue or the number of shares offered, we assumed this ratio to be one and consequently replaced missing values for one.

We are left with a sample of 333 firms. Unfortunately, we lost quite some observations due to the necessary conversion of CIK numbers to CUSIP numbers, as well as the exclusion of firms with certain characteristics as mentioned above and missing data in the Thomson One database.

B. Variables

Dependent variable

-Underpricing: percentage of first day return.

5_{http://www2.deloitte.com/content/dam/Deloitte/xe/Documents/AboutDeloitte/mepovdocuments/mepov17/}

were-going-on-an-IPO-mepov17.pdf

6

Independent variables

-

-Manager: dummy variable for firms where the ultimate owner is also a manager. -Direct ultimate ownership: percentage of direct ultimate ownership.

Control variables

-Underwriting fee: underwriting fee per share in US Dollars.

-Overhang: ratio of shares outstanding after the issue to newly issued shares. -Age: age of the firm at the time of the IPO.

-Proceeds: proceeds from the issue in US Dollars.

-Market capitalization: market capitalization at the offer price.

-Technology dummy: dummy variable for internet- and technology firms. -Bubble dummy: dummy variable for the year 2000 of the Dot-com bubble.

A more detailed description including construction of all variables can be found in Appendix B.

C. Descriptive Statistics

Table 1 reports information on the number of observations, mean, median, standard deviation, as well as minimum and maximum values regarding our variables.

Table 1. Descriptive Statistics - Complete Sample

(1) (2) (3) (4) (5) (6)

VARIABLES mean median sd min max N

Year of IPO 2010 2012 4.843 2000 2015 333

Family 0.492 0 0.501 0 1 333

Firm age (years) 11.73 6 16.94 0 101 332

Overhang 4.355 3.78 4.834 1 69.17 333 Underpricing 0.170 0.038 0.718 -0.403 12.33 333 Manager 0.205 0 0.404 0 1 332 Log of proceeds 18.69 18.69 1.173 12.61 22.65 304 Underwriting fee 0.110 0.084 0.120 0 0.630 333 % Direct-ownership 22.73 21.82 29.34 0 100 333 Log of market capitalization 20.00 20.00 1.452 13.27 25.12 308

Technology dummy 0.150 0 0.358 0 1 333

Bubble dummy 0.0841 0 0.278 0 1 333

The mean of the dummy variable Family, as reported in column 1, implies that 49% of our sample is a family controlled firm. This division of approximately half the sample being family controlled and the other half being non-family controlled, allows us to make a well established comparison between the

two groups. Column 5 of the variable Firm age shows the oldest firm to be 101 years. The average age of a firm at the time of an IPO is around 12 years old. Hence, our sample includes a broad rage of firm age, which we will use to test our hypothesis based on firm age in a later section. The minimum value of overhang can be partly attributed to replacing missing values by 1. Yet, this variable shows a high standard deviation, reported in column 3, of approximately 4.9, indicating a high variation of this ratio across firms. Direct ownership varies between 0 and 100% and shows a high variation, as displayed by the standard deviation of 29.34 in column 3. Column 1 shows that around 8% of the IPOs occurred during the ‘Bubble year’ 2000. The minimum value found for underpricing, stated in column 4, is -0.403. This value implies a negative initial return of 40.3% and thus ‘overpricing’ occurred in our sample. The maximum value of underpricing is 12.33, as shown in column 5, meaning 1233%, which seems to be an extreme outcome. We classify this value as an outlier. Since this outlier causes a skewed distribution and can affect the mean, this observation is eliminated from our dataset. Comparing the histograms before (graph 1a) and after (graph 1b), the elimination confirms that the distribution has improved, though skewness to the right remains.

Graph 1a. Distribution of Underpricing – before Graph 1b. Distribution of Underpricing - after

0 100 200 300 F re q u e n cy 0 5 10 15 Underpricing Graph 1a 0 50 100 150 F re q u e n cy -.5 0 .5 1 1.5 2 Underpricing Graph 1b

Subsample

Table 2. Descriptive Statistics - Family Sample

(1) (2) (3) (4) (5) (6)

VARIABLES mean median sd min max N

Year of IPO 2009 2010 5,015 2000 2015 163

Family controlled 1 1 0 1 1 163

Firm age (years) 13.61 9 16.62 0 101 163

Overhang 4.364 3.83 2.550 1 14.87 163 Underpricing 0.130 0.043 0.226 -0.281 1.011 163 Manager 0.411 0 0.494 0 1 163 Log of proceeds 18.47 18.41 1.246 12.61 22.65 148 Underwriting fee 0.103 0 0.123 0 0.630 163 % Direct-ownership 19.75 0 27.01 0 85.63 163 Log of market capitalization 19.83 19.82 1.528 13.93 25.12 156

Technology dummy 0.190 0 0.394 0 1 163

Bubble dummy 0.123 0 0.329 0 1 163

Table 2 provides summary statistics for the subsample of family controlled firms only.

Column 1 reports an average firm age of around 14 years old. Comparing this to the corresponding value reported in Table 1, we can say that family controlled firms tend to be older at the time of the IPO.

The values of direct ownership are not as we expected. We expect that family controlled firms on average have higher direct ownership, but eliminating non-family controlled firms from the sample does not lead to an increase of the mean. The percentage reported in column 1 of 19.75 is lower compared to the mean of 22.73 of the complete sample as can be seen in column 1 of table 1. This might be partly explained by a large number of missing values for family controlled firms which were replaced by 0. However, when we analyse Direct ownership before replacing missing values, we can tell from the number of observations reported in column 1 of Table 10 (Appendix C) that the known – and therefore also unknown– observations are relatively evenly distributed among family controlled and non-family controlled firms. Also, when we look at the sub-samples separately, the average percentage of direct ownership is still higher for non-family controlled firms as is shown by the mean values reported in column 2 of Table 10 (Appendix C). We attribute this to higher minimum and maximum values for non-family controlled firms, reported in column 4 and 5.

The different mean values raise the question why non-family controlled firms have higher direct ownership. Possibly, family firms have a more complex shareholder structure were multiple relatives have cross firm shareholdings in firms of family members instead of having a direct stake at the firm. Also, due to our classification of family controlled firms, non-family controlled firms consist of several types of shareholders. These include bank and financial companies, foundations, industrial

companies, and private equity firms. Perhaps, due to different characteristics of these types of shareholders, this has it implications for direct ownership.

Differences in mean underpricing

Next, we investigate the mean of underpricing per year and type of control. Table 3 and 4a report the mean for Underpricing, per year and control ‘group’ respectively.

Table 3. Descriptive Statistics – Underpricing by Year

(1) (2) (3) (4) (5)

Year Number of IPOs Underpricing sd min max

2000 28 0.126 0.269 -0.281 0.917 2001 3 0.109 0.105 0.036 0.229 2002 8 0.041 0.074 -0.052 0.145 2003 4 0.008 0.021 -0.011 0.038 2004 22 0.095 0.184 -0.110 0.606 2005 14 0.191 0.258 -0.028 0.946 2006 15 0.081 0.159 -0.1 0.552 2007 17 0.198 0.244 -0.048 0.759 2008 2 -0.074 0.131 -0.167 0.018 2009 7 0.065 0.115 -0.01 0.31 2010 2011 2012 2013 2014 15 19 29 55 54 0.120 0.022 0.164 0.143 0.108 0.192 0.040 0.218 0.244 0.266 -0.068 -0.05 -0.073 -0.085 -0.403 0.553 0.1 0.739 1.011 1.219 2015 40 0.240 0.451 -0.274 1.856

Graph 2 visualizes the contents of table 3. Both Graph 2 and column 1 of table 3 show the number of IPOs per year fluctuate heavily, with a peak between 2013-2015 and a low for the years 2001-2003 and 2008-2009. Most IPOs of our sample were conducted in 2013, whereas 2008 has the lowest number of observations. Highest average underpricing can be found in 2015, as stated in column 2 of table 3.

Graph 2. Average Underpricing and number of IPOs per year.

Table 4a column 6 reports the highest value of underpricing in the non-family controlled subsample. The mean value is slightly higher for the non-family controlled subsample as well, though the difference is very small compared to the family controlled subsample as listed in column 2.

Table 4a. Descriptive Statistics - Underpricing by control

(1) (2) (3) (4) (5) (6)

Control N mean median sd min max

Non-Family Family 169 163 0.136 0.130 0.018 0.043 0.295 0.226 -0.403 -0.281 1.856 1.011

The differences in mean for underpricing between family controlled firms and non-family controlled firms is further examined. Therefore, we conduct a t-test to test for differences in the mean. This test is constructed to test the null hypothesis that the difference between the population means is zero, i.e. to test whether the difference in underpricing between family controlled and non-family controlled firms can be associated with a ‘real’ difference in the population. The outcome of the independent t-test reported in table 4b, implies we can not reject that thepopulation means are equal. However, the violation of assumption8 _{5 concerning normal distribution (Appendix D) means the outcome of the}

independent t-test is invalid. Graph 3 and 4 display the distribution of underpricing for both groups (Appendix D).

8

List of assumptions from: https://statistics.laerd.com/stata-tutorials/independent-t-test-using-stata.php

-. 1 0 .1 .2 .3 U n d e rp ri ci n g 0 20 40 60 N u mb e ro fI PO s 2000 2005 2010 2015 Year of IPO NumberofIPOs Underpricing Graph 2

Table 4b. Equality of mean, two-sample t-test - Underpricing by control

(1) (2) (3) (4) (5)

Group Obs Mean Se Sd 95% Confidence interval 0 – Non-family 169 0.136 0.023 0.295 0.091 0.181 1 - Family 163 0.130 0.018 0.226 0.095 0.165 Combined 332 0.133 0.014 0.263 0.104 0.162 Difference 0.006 0.029 -0.05 0.06 Diff = H0: H1: mean (0) – mean (1) Diff = 0 Diff ≠ 0 t = df p-value 0.203 330 0.839

In addition, because the independent t-test may be invalid, we conduct a Wilcoxon rank-sum test to test for equality of means. The results are presented in table 4c. Again, we find a high p-value, meaning there is not enough evidence to reject the null-hypothesis of mean equality.

Table 4c. Equality of mean, Wilcoxon rank-sum test - Underpricing by control

(1) (2) (3)

Group Obs Rank-sum Expected

0 – Non-family 169 0.136 0.023 1 - Family 163 0.130 0.018 Combined 332 0.133 0.014 Adjusted variance: H0: 759449.37 Underpricing(0) = Underpricing(1) z = p-value = -0.0475 0.635

Correlation

Table 5. Correlation Matrix

1 2 3 4 5 6 7 8 9 10 11 1-Underpricing 1.000 2-Family -0.011 1.000 (0.839) 3-Manager -0.026 0.511*** 1.000 (0.636) (0.000) 4-Dir.ownership 0.038 -0.095* 0.061 1.000 (0.496) (0.085) (0.263) 5-Underwriting fee 0.045 -0.064 -0.016 -0.050 1.000 (0.416) (0.245) (0.766) (0.367) 6-Overhang 0.195*** -0.016 0.035 -0.056 0.044 1.000 (0.000) (0.765) (0.529) (0.306) (0.424) 7-Ln(1+age) 0.095 0.233*** 0.162*** 0.033 0.063 0.172*** 1.000 (0.085)* (0.000) (0.003) (0.554) (0.255) (0.002) 8-Ln(proceeds) 0.063 -0.180*** -0.006 -0.029 0.302*** -0.104* -0.145** 1.000 (0.276) (0.002) (0.912) (0.612) (0.000) (0.070) (0.011) 9-Ln(marketcap.) 0.149*** -0.138** -0.016 -0.081 0.307*** 0.263*** 0.110* 0.821*** 1.000 (0.009) (0.016) (0.775) (0.158) (0.000) (0.000) (0.054) (0.000) 10-D.Technology 0.089 0.109** 0.150*** 0.007 -0.090 0.153*** 0.115** -0.020 0.148*** 1.000 (0.105) (0.048) (0.006) (0.901) (0.101) (0.005) (0.037) (0.726) (0.009) 11-D. Bubble -0.008 0.136** 0.097* 0.104* 0.216*** 0.093* 0.075 -0.097* -0.027 0.084 1.000 (0.885) (0.013) (0.078) (0.058) (0.000) (0.091) (0.173) (0.091) (0.634) (0.125) P-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1

In line with theories suggested in the literature section, table 5 shows that Underpricing and Family control are negatively correlated, as well as Underpricing and Manager, though not significantly. Market capitalization and Proceeds show a very high positive and significant correlation of 0.821, as reported in table 5. This is not surprising, since both variables are constructed using the offer price. Market capitalization and Overhang also show a significant positive correlation of 0.263. Again, we suspect this correlation to be caused by shares outstanding after the offer as common factor of both variables. Proceeds and market capitalization are significantly and negatively correlated to family. This fits with the expectation that family controlled firms conduct smaller issues from the perspective of keeping control. Worth mentioning is the substantial positive correlation of 0.511 between Manager and Family which is also significant at the 5% level. Apparently, it is common for family controlled firms to have an owner who is also a manager.

V. Methodology

This section elaborates on the empirical model and the methodology used to test whether underpricing is different for family controlled firms. We will use a multiple linear regression model including several control variables, which attempts to model the relationship between the dependent and independent (explanatory) variables.

We extracted several control variables from existing literature, known to affect underpricing.

The empirical analysis shows similarities with the regressions done by Liu and Ritter (2011) and Ding and Pukthuanthong-Le (2009). In their paper, Ding and Pukthuanthong-Le (2009) use the OLS estimation technique on a multiple linear regression equation to examine the association of underpricing with various family ownership- and IPO variables.

Even though Liu and Ritter (2011) investigate underpricing in a different light, namely local underwriter oligopolies, their multiple linear regression model is comparable. They regress the

percentage of first day return on variables associated with the underwriter, firm characteristic variables as control variables, as well as industry- and year- fixed effects using OLS. Several elements of their empirical analysis can be used for the purpose of this paper.

Accordingly, we will set first day underpricing as the dependent variable. Underpricing is measured as the percentage of first day return of a share. The return window is one day and thus we define

underpricing as the difference between the first day closing price and the offer price, divided by the offer price ((P1-P0)/P0). This definition of underpricing is also used by Habib and Ljungqvist (2001) and by many other researchers studying IPO underpricing.

The independent variables relate to firm structure and should provide insights to the main question whether underpricing is different for family controlled firms. We expect these variables concerning corporate governance to be different for family controlled firms and thus they might tell us more on the relation between family controlled firms and IPO underpricing.

The most important independent variable in our regression is “Family”. As described in section IV, we have classified family- and non-family controlled firms according to their type of ultimate owner. Following this classification, we construct a dummy variable that indicates whether the firm is family controlled or non-family controlled. It has a value of one if the firm is family controlled and zero for non-family controlled firms.

A positive sign of the corresponding coefficient would indicate that family controlled firms experience more underpricing, compared to non-family controlled firms.

The second independent variable that is interconnected to firm structure is “Manager”. This dummy variable has value one indicating that the ultimate owner is also a manager and zero otherwise.

Filatotchev and Bishop (2002) and Ding and Pukthuanthong-Le (2009) among others have used percentage of outside ownership in their regressions. However, since we were not able to obtain data on this variable, we have chosen for data on management ownership. Whether the owner is also a manager is in fact an indication of insider ownership. Thus, this variable denotes the effect of insider ownership on IPO underpricing.

Thirdly, the “Direct ownership” variable contains the percentage of direct ultimate ownership. We expect this variable to reveal more about the relation between how much a (family controlled) firm owns directly and underpricing. A higher percentage indicates the firm can exert more control over the firm and is therefore more able to influence underpricing.

To control for several other factors that are carried out by prior literature as to influence underpricing, we add control variables to our regression model. Among them is “Underwriting fee per share”. We expect the amount of underwriting fee to obtain information on valuation uncertainty. Though an investment bank is compensated by gross spread, this spread is compensation for more than the underwriting services. Habib and Ljungqvist (2001) therefore declare the underwriting fee charged for the underwriting cover as a better proxy for valuation uncertainty. The higher this valuation

uncertainty, the more the issue will be underpriced.

Another important variable to control for is “Overhang”, which is the ratio of total number of shares after the issue divided by the number of new shares offered. Overhang is also used by Leitershof and Rau (2014) who claim that a higher level of underpricing may be accepted when a lower proportion of wealth of existing shareholders is at stake (Dolvin and Jordan, 2008, cited in Leitershof and Rau, 2014). In addition, Habib and Ljungqvist (2001) provide the argument that the amount of shares sold by the owners at the IPO, forms a link with how much owners care about underpricing. Hence, if this ratio is low, meaning a relatively large fraction of shares offered, this increases incentives to reduce underpricing.

Following Beatty and Ritter (1986), another control variable is the “Proceeds of the issue”. Since larger firms are considered to be less risky, we expect the size of the issue to be negatively correlated with underpricing. To control for firm size, the variable “Market capitalization” (at offer price) is constructed and included in the regression model. Leitterstorf and Rau (2014) note that there tends to be more information available for larger firms, thus lowering information asymmetry and this could reduce underpricing. Beatty and Ritter (1986) also argue that large firms can be seen as less risky and therefore they expect issue size in dollars to be negatively correlated with underpricing. In contrast, Baron (1982) argues firm size can increase underpricing because it is harder to market larger issues and thus underpricing the issue could compensate for that. Information asymmetry and therefore underpricing is also linked to the age of the firm, as it is easier to evaluate a company that is older and where more past data is available. So, in accordance with Ljungqvist and Wilhelm (2003) we control for “Firm age” at the time of the IPO too.

Consistent with Loughran and Ritter (2004) we will make use of a “Bubble dummy” for the dot-com bubble for the year 2000. The authors found that underpricing increased a lot in the bubble period, assumingly due to the valuation uncertainty.

The last control variable is the ‘Technology dummy’ for which we follow the specification of Loughran and Ritter (2004) to indicate internet and technology firms. These firms likely experience different underpricing as investors find them more difficult to value and their value is usually based on growth options (Liu and Ritter, 2011).

We run several regressions, each time adding more control variables to check for robustness of our results.

The first regression is based on a basic equation using only the dependent and independent variables.

(1) Underpricingi = β0 + β1Familyi + β2Manageri + β3Direct-ownershipi + ui

Next, we run a second regression including control variables.

(2) Underpricingi = β0 + β1Familyi + β2Manageri + β3Direct-ownershipi + β4Underwriting feei +

β5Overhangi + β6ln(Firm age)i + β7ln(IPO proceeds)i + β8ln(Market capitalization)i +

β9Technology dummyi + ui

To control for possible differences in IPO underpricing in 2000, we add a bubble year dummy.

(3) Underpricingi = β0 + β1Familyi + β2Manageri + β3Direct-ownershipi + β4Underwriting feei +

β5Overhangi + β6ln(Firm age) i + β7ln(IPO proceeds) i + β8ln(Market capitalization) i +

β9Technology dummyi + β10Bubble dummyi + ui

For regression four we add year fixed effects using IPO year dummies (T-1), in addition to the dependent-, independent- and control variables. By adding year fixed affects, we control for average differences across years and reduce the threat of omitted variable bias. We now have year fixed effects, so the bubble dummy is left out.

(4) Underpricingi = β0 + β1Familyi + β2Manageri + β3Direct-ownershipi + β4Underwriting feei +

β5Overhangi + β6ln(Firm age) i + β7ln(IPO proceeds) i + β8ln(Market capitalization) i +

β9Technology dummyi +∑dt Year dummyt + ui

Regressions 1-4 will be used to assess hypothesis 1.

The previous equation including control variables and year fixed effects is further elaborated using interaction terms. The variables Firm age, Manager and Direct-ownership are interacted with the

Family variable. We will use this regression to examine hypotheses 2-4. The interaction terms will show us how these variables (firm age, owner also a manager and direct ownership) affect

underpricing for family controlled firms.

(5) Underpricingi = β0 + β1Familyi + β2Manageri + β3Direct-ownershipi + β4Underwriting feei +

β5Overhangi + β6ln(Firm age) i + β7ln(IPO proceeds) i + β8ln(Market capitalization) i +

β9Technology dummyi + β10ln(Firm age)*Family i +β11Manager*Family i + β12

Direct-ownership*Family i + ∑dt Year dummyt + u i

Moreover, we will run regressions 1-4 also for the subsample of family controlled firms only. In stead of using interaction terms, the variables on their own may provide information on their relation to underpricing regarding family controlled firms. This allows us to compare the results using a different sample selection.

When there is heteroskedasticity, this has consequences for our OLS estimates. To detect possible heteroskedasticity we made a residual versus fitted plot for regression model 2. Graph 5 (Appendix E) shows this plot and we can see that the spread is not evenly distributed. This indicates that the variance of the error term is not constant, hence there may be some heteroskedasticity. To account for this problem, we made the regressions robust by using heteroskedasticity consistent standard errors.

VI. Results

This section aims to clarify the results found from the OLS regression models specified in section V. The results from the complete sample are reported in table 6. Table 7 displays the results from the regressions based on the subsample of family controlled firms only.

Table 6. IPO Underpricing regressions – Complete sample

The sample includes 331 US firms for column 1 and 284 US firms for column 2-5 from 2000-2015 and meet other criteria described in section IV of this paper. The dependent variable for all regressions is Underpricing, measured as the percentage first day return from the offer price to the closing price. Family controlled takes on a value of one (zero otherwise) if the firm is family controlled. The Manager variable equals one is the ultimate owner is also a manager (zero otherwise). Direct ultimate ownership is the percentage of direct ultimate ownership. Underwriting fee is the underwriting fee per share (in US dollars). Overhang is the ratio of shares outstanding after the offer to new shares offered. Ln(Firm age) is the natural logarithm of the firm’s age plus one. Ln(Proceeds) is the natural logarithm of the firm’s proceeds from the issue. Ln(Market capitalization) is the natural logarithm of market capitalization at the offer price. The technology dummy takes on value one for a technology firm (zero otherwise), based on SIC codes. The Bubble dummy has value one in the IPO occurred in 2000 (zero otherwise). Interaction terms on family control and firm age as well as family control and direct ownership are included. The interaction term for family controlled and manager was omitted due to collinearity. Year fixed effects based on the year in which the IPO was conducted are included, the coefficients are excluded from the table for brevity. T-statistics are computed using heteroskedasticity consistent standard errors, corrected for clustering across year.

(1) (2) (3) (4) (5)

VARIABLES Underpricing Underpricing Underpricing Underpricing Underpricing

Family controlled 0.00445 -0.000475 0.00150 0.0258 -0.0412 (0.0341) (0.0529) (0.0530) (0.0528) (0.0904) Manager -0.0216 -0.0378 -0.0380 -0.0406 -0.0480 (0.0357) (0.0345) (0.0348) (0.0377) (0.0352) % Direct-ownership 0.000357 0.000695 0.000732 0.000904 6.55e-05 (0.000466) (0.000533) (0.000572) (0.000609) (0.000677) Underwriting fee 0.00495 0.0203 0.0558 0.0468 (0.118) (0.132) (0.115) (0.107) Overhang 0.00447 0.00465 0.00558 0.00554 (0.00437) (0.00416) (0.00389) (0.00405) Ln(Firm age) 0.0117 0.0115 0.0102 0.00177 (0.0120) (0.0120) (0.0138) (0.0172) Ln(Proceeds) -0.0268 -0.0268 -0.0281 -0.0310 (0.0213) (0.0214) (0.0190) (0.0204) Ln(Market capitalization) 0.0501*** (0.0161) 0.0492*** (0.0156) 0.0402*** (0.0133) 0.0408** (0.0148) Technology dummy 0.0318 0.0335 0.0339 0.0366 (0.0313) (0.0316) (0.0365) (0.0391) Bubble dummy -0.0227 (0.0298)