Are Dutch-Auctions really the answer?

“Are Dutch-Auctions really the answer?

MSc Business Economics, Finance track

Master Thesis

A e Dut h-Au tio s eall the a s e ?

Made by: Durk Kooi

July 2014

Abstract

This thesis investigates the impact of the Dutch-Auction method of IPO on the severity of Underpricing in the U.S. in the period January 1999 to May 2014 and 1974 firms within. Where most previous research regarding Auction theory overlooked the issue of endogeneity, this thesis corrects for it. The research is based on the idea that the choice for Dutch-Auction is correlated with Firm and Issue characteristics which are correlated with the level of underpricing. A modified Heckman Selection model is used to correct for self-selection within the firms who chose Dutch-Auction. There are two major findings in this thesis. First it is found that there is indeed a high level of self-selection within the choice of choosing a Dutch-Auctioned method for IPO. For firm characteristics the only significant variable that affects the probability of Dutch-Auction is that young firms tend not to engage in a Dutch-Auctioned IPO. As for issue characteristics, the smaller offerings will increase the probability of choosing a Dutch-Auctioned method of IPO. Secondly, the dummy for Dutch-Auction indeed reduces underpricing after correcting for selection bias. This result survives multiple robustness checks and is significant. However due to the tiny amount of firms that choose Dutch-Auction method for IPO there still exist some questions surrounding this topic.

Table of Content

Abstract ... 2

Table of Content ... 3

1. Introduction ... 4

2. Literature Review ... 7

2.4 Empirical research IPO Auctions ... 10

2.5 Reduction of Underpricing through Dutch-Auction ... 12

3. Methodology & Hypotheses... 14

3.2 Objectives ... 14

3.3 Model ... 15

3.3 Firm and Issue Characteristics and Control variables ... 18

4. Data and descriptive statistics ... 21

4.1 Data Sources & Sample ... 21

4.2 Data Description ... 21

5. Results... 25

5.1 Empirical Results First Stage ... 25

5.2 Empirical Results: Main Effect of Dutch-Auction on Underpricing ... 27

6. Robustness Checks... 28

7. Conclusion: ... 31

References ... 33

1. Introduction

When Twitter placed its initial public offering (IPO) in the morning of the 7th of November for an offer price of $26 everyone involved held his or her breath. However when Twitter closed at $44.90, it was obvious that the security was not priced correctly. A lot of money was left on the table by the issuers and a phenomenon that is referred to as IPO underpricing occurred.

The literature suggests that an adequate solution against underpricing would be to reduce the power of the underwriter. In a survey among 336 Chief Financial Officers (CFOs) it is even found these CFOs indicate that the second most important reason for underpricing is the desire of underwriters to obtain the favor of institutional investors (Brau and Fawcett 2006). One way of reducing underwriter power would be the IPO via an Auction Method. In the U.S. in the year 1999 a new method of taking a compa pu li had ee i t odu ed: The Dutch-Auction.1 This was method of taking a company public was introduced by the investment bank WR Hambrecht Co. under the name of OpenIPO. The OpenIPO process is a multi-day Dutch-Auction over the Internet where interested bidders can enter the offer prices and quantities desired.

In 2004, this method of IPO was even adopted by Google for their IPO. In this method of pricing an IPO the highest clearing price for all the shares will be the price all the investors pay. In an Economist article in June of 2002 the author of the article even proposes to the make the method of IPO required. The qualitative literature however suggests that this new pricing mechanism is not necessarily more fair to all parties involved and may in fact lead to less efficient capital markets (Anand 2005, Jagannathan et al. 2009)

What is quite interesting is that most of the previous empirical work draws no solid or unified conclusions. Some find that auctions may be less desirable while others state that Dutch-Auctioned add value. For Example Sherman (2005) finds that auctions may be far less desirable while Degeorge et al. (2010) come to the conclusion that Dutch-Au tio IPO’s add alue. Even more so, the

1

The Dutch-Auction is a type of auction where the price is set high by the auctioneer and lowered until a participant is willing to accept. The name originates from the Dutch flower markets.

literature even hints to the fact that an explanation the Dutch-Auction method might not be functioning correctly is because investors are not yet entirely familiar with the concept (Robinson and Robinson 2012).

Because the method of taking a company public is a choice there is likely to be self-selection within the companies who choose for the Dutch-Auction of IPO. The literature is unclear about this aspect of the Dutch-Au tio ed IPO’s. Furthermore most of the underpricing empirical research2 regarding underpricing of this method of IPO has been done by an ordinary least squares analysis (OLS). For OLS to be valid however the error terms of the regression variable have to be uncorrelated with the out come variable. I however believe that due to self-selection this is not the case. More specifically I think that there are certain, observables and unobservable variables, that both influence the choice for Dutch-Auction and the level of Underpricing at an IPO.

In this thesis I will test if there is self-selection among the firms choosing the Dutch-Auction method of IPO and if there is a causal effect of this Dutch-Auction method of IPO and underpricing. This will be tested by means of using of a Heckit model. A Heckit model is a modified form of the Heckman (1978) sample selection model adjusted for analyzing treatment effects. The sample used in this thesis consists of firms who conducted an IPO in the U.S. in the period January 1999 to May 2014. The number of firms in this sample who performed an IPO is 1974, with 24 of those firms who chose the Dutch-Auctioned IPO leaving the ast ajo it of IPO’s ei g ook uilt.

The results of this thesis are as that, as hypothesized, that there is indeed self-selection with the choice for the Dutch-Auction method of taking a firm public. Smaller offerings tend to significantly increase the probability of a Dutch-Auctioned IPO. On the other hand young firms, in terms of firm age at the time of the IPO, tend to significantly decrease the probability of a Dutch-Auctioned IPO. Furthermore after correcting for self-selection bias, it is found that there is indeed a

2

_See

causal effect of the Dutch-Auction method of IPO and the level of underpricing. This effect is found to be negative and is consistent through multiple robustness checks.

The remainder of this thesis is organized as follows. Section 2 gives a review of the literature ega di g IPO’s, IPO u de p i i g a d au tio ed IPO’s. “e tio states the h pothesis, e plai s the methodology this thesis will use and reviews the variables used and their expected signs. Section 4 will give background on the data and will describe the data used. Section 5 will present the main results of this thesis. Section 6 will present robustness checks and Section 7 concludes.

2. Literature Review

2.1 What is an IPO?

When a firm is in need of capital there are three activities a firm can engage in, namely gathering internal funds, issuing debt or issuing new equity. The pecking order theory as developed by Myers (1984) states that firms prioritize their sources of financing by first depleting internal funds, then issuing debt till the level that it is still viable and finally equity is issued. The focus of this thesis however will be on issuing equity. More specifically this thesis will focus on the first time a company issues equity publicly, the Initial Public Offering (IPO).

The existing academic literature suggests four motivations for an IPO. The first motivation can be described as the cost of capital motivation as first published by Modigliani and Miller (1963). This stream of literature states that a firm will go public if this minimizes the cost of capital. Furthermore this is the backbone of the pecking order theory. Second there is the thought that an IPO allows insiders to cash out and that this is a good harvesting strategy for Venture Capital firms because an IPO can be seen as option to bail out. Third it is shown that IPOs may serve as facilitation for takeover activity. Furthermore it is argued that the i po ta e of IPO’s fo takeo e a ti it stating that when a fi goes pu li its sha es a se e as u e fo takeo e s ith optio s. Finally a motivation for going public as a firm is that an IPO may also be a strategic move. Not only is the firm benefitted from the extra publicity and public attention, analyst recommendations are also biased upward after an IPO3.

2.2 IPO Underpricing

In the U.S. there are two ways of taking a firm public. There is the auction method, commonly referred to as Dutch-Auction, and book building. Book building is by far the most well-known and well-documented method of conducting an IPO. This is most likely because it is also the most used method of taking a firm public. For example in the sample used in this thesis there are 24 firms who

3

See Zingales (1995), Black and Gilson (1998), Brau et al. (2003) and Ritter (2003) for further e pla atio of these oti atio s fo IPO’s

used an auction method and 1950 firms in the sample period who have used book building as a method of IPO.

IPO underpricing occurs when shares are significantly underpriced on the first day of trading in order to make a significa t ju p Logue . Fou streams within the theory explaining IPO underpricing have been documented academically: asymmetric information, institutional reasons, control considerations, and behavioral approaches.

The asymmetric information theory, which states that one of the parties involved in an IPO knows more than the other, holds the most ground in the literature. The parties involved being the issuer, the lead underwriter and the investors. In this stream of models it is mostly assumed that the lead underwriter holds the most information and therefore has the most influence if an IPO will be underpriced or not (Baron 1982). Within the asymmetric informatio theo odels Ro k’s

i e ’s u se odel is p o a l the ost well known. This theory states that informed investors o l id o att a ti el p i ed IPO’s hile u i fo ed i esto s id i dis i i atel . This i tu has as an effect that uninformed investors while make a loss on average. On the other hand there are the informed investors whom ho e e ill o l id o att a ti el p i ed IPO’s. In order to have a fu tio i g a ket fo IPO’s o ditio al e pe ted etu s ha e to e at least g eate tha ze o so the uninformed investors at least break even. In other words, this means that IPO’s ha e to e underpriced in order to have a market all together.

Furthermore within this stream there is the information revelation model. This model assu es that the i esto s a e the ost i fo ed pa t i this ga e . This odel st i es to accomplish a mechanism that induces investors to reveal their information truthfully and book building can, under certain conditions, be such a model4. This theory states that in principle the underwriting bank will reward the investors who are truthful with their information. For example if

4

Benveniste and Spindt (1989), Benveniste and Wilhelm (1990), and Spatt and Srivastava (1991) show the conditions under which book building is such a model.

an institutional investor is willing to reveal all its information so the underwriting bank can come to an offer price, which fully reflects the value, this investor might be rewarded for its truthfulness by giving a somewhat lower price. Although this leads to underpricing the issuer will be better off because the offer price now reflects the intrinsic value.

The next big stream in the IPO underpricing literature consists of the institutional explanation models. First and foremost is the basic idea developed by Logue (1973) and Ibbotson a ed the legal i su a e app oa h . This app oa h states that o pa ies deli e atel sell their stock at a discount in order to reduce the likelihood of future lawsuits by disappointed shareholders over the post-IPO pe fo a e of the sha es. A othe app oa h is the p i e sta ilizatio app oa h , this app oa h states that u de ite s i te d to edu e p i e d ops i the aftermarket and therefore engages in underpricing. The reason this is done is that underwriters when performing a book built IPO generally guarantee a certain price in the aftermarket. If this is not reached they will buy the shares themselves trying to drive up the price. This will have as an effect that the underwriter will make a loss. Fi all the the e is the ta ad a tages app oa h hi h state that there may be certain tax benefits accompanied by underpricing a stock and therefore this becomes a trade-off.

The third stream within the IPO underpricing literature is the ownership and control stream. The two most important streams within this model are that managers use underpricing as way to keep control of the company, this because if they value of the shares held by the managers decreases less relatively to the value sold publicly they retain control (Brennan & Franks 1997). The next theory within this stream is that managers use underpricing in order to avoid agency costs by large stake-holders who suddenly want to know all about the company.

Finally there are the behavioral explanations models. These state that some investors on the aftermarket are irrational and therefore the large price jump of the first day of the stock is there.

The next approach within this model is however that issuers fail to put pressure on the underwriting banks in order to have underpricing reduced.

2.3 OpenIPO

Robinson & Robinson (2012) explain that in the OpenIPO process after the bids on the auction are received there can be a below market-clearing price as the final offer price. If the issuer chooses this approach of sanctioning the shares, she creates an excess demand. WR Hambrecht Co. then uses a distribution algorithm to divide the shares among the bidders. By employing such an algorithm the issuer has the choice to ration the shares among institutional investor or not. Because the Dutch-Auction as offered by WR Hambrecht in the form of OpenIPO is online, the retail investor now too has a chance to bid on the shares. So by employing this distribution mechanism, and that retail investors now have the opportunity to invest the issuers can now redistribute the shares to what investors they want instead of what the underwriter wants. Thereby eliminating what CFOs indicated to be the second most important reason for IPO underpricing (Brau and Fawcett 2004), that underwriters underprice to gain the favor of institutional investors has now disappeared due to the Dutch-Auction method of IPO. Furthermore because the shares are now sold to everyone who wishes to bid auction theory predicts that the price will be closer to the intrinsic value of the stock and thus underpricing will be reduced.

2.4 Empirical research IPO Auctions

The literature above states suggests that IPO auctions will reduce underpricing. However the empirical evidence regarding this aspect of Dutch-Au tio ed IPO’s is o t adi ti g. A uestio that iti s ight ask is: If au tio ed IPO’s ai l ha e ad a tages ho o e the ha e ee abandoned in other countries?

In Japan where IPO auctioning was previously the most used method for taking a firm public it has been driven out by book building as researched by Kutsuna & Smith (2004). Their results indicate that if all issues are equally weighed the issue cost of the auction model is less than that of book building however when issues are weighted by Issue size, the estimated aggregate costs of

auctioning and book building are similar. The reason they find that book building has driven out IPO auctioning is because the auction method of IPO does not reflect opportunity costs related to underinvestment. Furthermore estimates of issue costs overlook the more accurate pricing benefits that book building affords. Moreover Jagannathan and Sherman (2004) find that in nearly all the 47 countries they studied the auction methods of IPO, book building is now commonly used. In a later study Sherman (2005) models both book building and auction methods after which he finds an explanation for this result of why these IPO Auctions have been abandoned. The overall result of this study is that book building is less risky, less risky for both investors and issuers which leads to, in general, less underpricing. The explanation for this smaller level of risk is that someone, in this case the underwriter, is managing the process. By selecting book building method the underwriter can guarantee that enough investors carefully evaluate the IPO in contrast to an auction method where no such guarantee can be given. These results reinforce the reinforce of Jagannathan and Sherman (2004) who offer evidence of when a large number of bidders participated in an auction this lead to inaccurate pricing and high aftermarket volatility. However in some countries the opposite has happened with regard to book built and au tio ed IPO’s. In Israel for example, the book building way of performing an IPO is forbidden. However this is a su ess as Ka del et al. (1999) point out there is only 4.5% underpricing on average, which is significantly lower than the level of underpricing in the U.S.

Another drawback of an auctioned IPO could be the free rider problem. A characteristic of an IPO is that it is difficult and time consuming to evaluate the value of an IPO through the eyes of an investor. This lack of price discovery is mainly because there are no analysts reports, in the U.S. at least, before the offering of a company and thus there is a free rider problem. Within the information revelation academic literature it is generally assumed that investors are compensated for their revealing of information in the form of underpricing. Therefore it is expected that when shares are not discretely offered to investors, commonly institutional investors, who are willing to put in the research their reward in the form of underpricing fades away. However in later empirical

research its is found that WR Hambrecht Co. using the OpenIPO process received 84% of their demand in dollar value from institutional investors.5 Concluding on the empirical section, all the research is contradicting and not all the research has been done in an appropriate manner.

2.5 Reduction of Underpricing through Dutch-Auction

Because the role of the underwriter is limited in the Dutch-Auction process it is expected, and shown empirically, that the promotion of the IPO will be limited as well. However because the Dutch-Auction method of IPO is open to retail investors there will be price discovery and word will spread among retail investors that this is expected to heighten the offer price towards nearer to its intrinsic value and thus reducing underpricing. As stated efo e the u de ite po e i IPO’s is sig ifi a t and large. Chen & Ritter (2000) even find that an investment bank sets its fee at 7%. Furthermore in their research they develop the hypothesis that investment banks even further try to gain from an IPO. The indirect gain from underpricing may even be that investment banks underprice the IPO to get further business from the institutional investors who stand to benefit. Moreover underwriters also underprice to encourage “to k Flippi g in order to make stocks liquid. His reasoning for this is that in order to give groups a reason to trade. So by engaging in underpricing the aftermarket liquidity of a stock is raised. Furthermore by allocating IPO shares to encourage an active secondary market, the underwriter gains profits from market making. Moreover underpricing may be considered the cost of providing liquidity in the issue6. Finally in a more recent study it shown that Dutch-Au tio ed IPO’s edu e underpricing however after a multivariate analysis it is shown that the coefficient for the Dutch-Auction is not significant.

Summarizing the literature has multiple explanations for why IPO underpricing occurs. None of these theories however are conclusi e o a gold sta da d a d the efo e IPO u de p i i g e ai s o e of the ost esea hed topi s ithi a ade i lite atu e. F o a la e ’s pe spe ti e

5

Sherman and Titman (2002) show the characteristics of an IPO and link the free rider problem while Degeorge et al. (2010) find that this is not in line with empirical esea h ega di g IPO’s.

6

For more in depth analysis regarding liquidity and market making see Boehmer (2000) and Ellis et al. 1999)

Dutch-Au tio ed IPO’s a e e pe ted to edu e u de p i i g, afte all follo i g au tio theo a d that underwriter power is reduced this should be apparent. However the empirical research to IPO auctions contradicts this but it also contradicts itself.

3. Methodology & Hypotheses

3.2 Objectives

This thesis will try to fill in the gap of the literature in review above. What the literature does not document, but seems a large part of the Dutch-Auction IPO discussion is the issue of self-selection within the firms who choose Dutch-Auction IPO. Furthermore the empirical literature is also not clear on whether or not it reduces underpricing. Even more so, in light of endogeneity most previous research regarding this topic seems statistically invalid.

More specifically this thesis will try and investigate whether there are certain determinants for a firm to choose, the uncommon, Dutch-Auction method of taking the firm public. The focus is specific on the period January 1999 to May 2014 because the OpenIPO process, the Dutch-Auction version of IPO as offered by WR Hambrecht Co, was introduced in 1999. Since 1999 only 24 companies have chosen the Auction method of going public. A reason that not many Dutch-Auctioned issues have been done might be that issuing companies were not yet at ease with the new method and that this would develop over time. Although an increase in the number of firms who performed a Dutch-Auctioned IPO was expected, it barely did. Therefore, in comparison to the latest empirical research about Dutch-Auction IPOs that employed 21 firms, the sample used by this thesis will only improve this number by three. This on the other hand can be explained by that the number of IPO’s is decreasing7_{. This is in line with the academic literature stating companies prefer to do an}

IPO i hot IPO a kets.

This leads to the two hypotheses to be researched by this thesis:

Hypothesis 1: Certain firm characteristics will determine the likelihood of a firm performing a Dutch-Auctioned IPO, thus firms “elf-select .

Hypothesis 2: The Dutch-Auctioned method of IPO reduces underpricing.

7

The latest empirical research was done by Robinson & Robinson (2012) and Ritter (2012) finds that number of IPOs in general is increasing.

3.3 Model

The main contribution of this paper will see if there is self-selection within companies that go public via the Dutch-Auction method and if this indeed reduces underpricing. Underpricing being the first day return of the company in the aftermarket. In other words underpricing is the difference between the offer price and the closing price on the first day of trading.

Then the next step is to determine the causality between Dutch-Auction and underpricing. Until now the literature has only used OLS and multivariate regressions (Robinson and Robinson 2012). This overlooks the issue of endogeneity, one of the most common issues in corporate finance. More specifically to the topic at hand, if Dutch-Auction reduces underpricing is it also not likely that the firms who select the Dutch-Auction method were less prone to underpricing to begin with? Statistically speaking I believe the error terms of Dutch-Auction and Underpricing are related. Or as shown in the formula below:

Y=Underpricing

D=Dummy for Dutch Auction U=Error term of regression V=error term of error term.

Because the research performed in this thesis presumes self selection for the Dutch-Au tio ed IPO’s it assumed that D is correlated with error term and therefore when performing an Ordinary Least Squares (OLS) regression of Y to check for the effect of D it will provide the estimates ß1 & k1. Differently speaking when regressing Dutch-Auction dummy on underpricing the error terms will be correlated and therefore an inconsistent estimator of how Dutch-Auctions are supposed to reduce underpricing.

A solutio fo this is a odified fo of He k a ’s , , sa ple sele tio model. Heckman based his model on incidental truncation, which is an effect of data gathering rather than data collecting. Heckman (1979) described self-selection bias in the following way:

One observes market wages for working women whose market wage exceeds their home wage at zero hours of work. Similarly, one observes wages for Union members who found their nonunion alternative less desirable. The wages of migrants do not, in general, afford a reliable estimate of what nonmigrants would have earned had they migrated. The earnings of

manpowertrainees do not estimate the earnings that nontrainees would have earned had they opted to become trainees.

He then modeled a two-step OLS equation based on a regression equation, which is only observed when the variable of interest is larger than one and a selection equation accompanying it. However as stated before the error terms are related and therefore a propensity analysis to determine the probability and then this will be entered in the regression equation.

It is important to understand the basics of this sample selection model, because this thesis will research an endogenous treatment effect. This treatment effect, the choice of a Dutch-Auctioned IPO, is suspected to be an endogenous binary variable8. In the treatment model derived from the Heckman Selection model a Probit model of predicting can be made as Greene (2003) adequately named these Heckit models. This treatment effect model differs from the sample selection in the manner that a dummy variable indicating treatment or not, here Dutch-Auction, is directly entered into the regression equation and the outcome variable being underpricing is observed is observed for both the treatment group and the non treated group. More specifically the treatment model takes the form of a regression equation and a selection equation as described below:

8

The choice for a Dutch-Auctioned IPO is either yes or no (0 or 1) and most likely related to firm or issue characteristics, thus this variable is an endogenous binary variable.

9 Probabilities: | | 10

The importance of this Heckit model becomes obvious because the treatment effect of Dutch-Auction is estimated and the choice for Treatment is, as previously stated, endogenous. So to estimate the effect of taking a company public through a Dutch-Auction method of IPO this two-stage model has to be followed. In short, the model states that underpricing is dependent on Firm and Issue Characteristics, to be defined later, of the IPO. Furthermore there are certain control variables, also to be explained later, that will also influence the severity of Underpricing.

9

Please note that i represents the firm, s the industry and t the year.

The likelihood of the treatment effect, Dutch-Auction, is to be estimated by a form of probability equation or in the Heckit model the selection equation. As a first step the likelihood of a Dutch-Auction IPO, given the firm and issue characteristics, is estimated through the selection equation. Then the outcome of this selection equation is switched11 in to the regression equation. The model explained above can be estimated by a two-step procedure same with the Heckman Selection model or a maximum likelihood estimator for the coefficients can be used. In this thesis the maximum likelihood estimator12 is used because it increases the efficiency of the model employed. For more information see Greene (2003).

3.3 Firm and Issue Characteristics and Control variables

There are certain variables surrounding an IPO that influence the severity of underpricing. Even more these same variables might influence the choice of method of a taking a company public, the self-selection as assumed by the model. Following the structure of the model and the essence of the research the Firm and Issue Characteristics and Control variables that affect underpricing, and thus the Regression equation will be explained.

A firm characteristic that will be employed as a control variable is Firm Age at the time of IPO. As defined in table 3 Firm age measures difference between the Offer year, the year of the IPO, and the year the firm was founded. Older firms with a longer history, more common to the public and more information available tend to have lower uncertainty surrounding them. As Ritter (1984) shows, there is a positive relationship between the level of underpricing and the ex ante uncertainty surrounding the value of a firm. So because older firms have lower ex ante uncertainty they will be likely to have less underpricing.13

11_{As defi ed Qua dt’s , fo of the eg essio odel that states t o out o es.} 12 _{A maximimum likelihood estimator takes the mean and variance of sample as parameters and}

then finds the paramaters that make the observed model the most probable.

13

Empirically this negative relationship between underpricing and firm age has been shown by Su and Fleisher (1999), Loughran and Ritter (2004) and Chahine (2008).

The literature on the relationship between underpricing and the Size of the Equity Offering is contradicting to say the least. The going literature, as developed by the most well known academics o the su je t, suggest that la ge IPO’s ill e u de itte the ulge-bracket investment banks. This is supposed to reduce the risk of perceived risk for the IPO. This phenomenon has even been described as something that al a s a e esta lished e pi i all , o a e pi i al egula it . The state that smaller offerings are, on average, accompanied with higher levels of uncertainty than their larger counterparts. Therefore one would expect an inverse relationship between Size of Equity Offering and underpricing. Other studies however show empirically that the opposite is true14. Therefore I am ambiguous on the expected relationship between the variables Size of Equity Offering and underpricing in the regression equation.

Then there are the control variables, or the fixed effects in this research. They will only be included in the regression equation, because there is no empirical or qualitative evidence whatsoever to hint that certain fixed effects will increase or decrease the probability of a firm performing a Dutch-Auction method of going public. Ibbotson et al. (1994) find empirically that the se e it of IPO u de p i i g is g eatl e la ged du i g hot IPO a kets. Fu the o e the variability of IPO returns varies greatly over time. Therefore the use of yearly fixed effects seems reasonably justified. Another control variable to be included will be the industry fixed effects. This is justified by Benveniste (2003), who prove the significance of an industry dummy.

Finally there is the Selection equation and the question how Firm and Issue characteristics will be of influence. First I expect older firms to be more likely to chose a Dutch-Auction method of taking the company public. The reasoning behind this expected correlation is that older firms, with an existing customer base, will be less benefitted from the auxiliary services15 that are normally

14

Ritter (1986) and Dunbar (2000) show the inverse relationship between size and underpricing while Habib & Ljunqvist (1998) and Daily et al. (2013) show a positively correlated relationship between size and underpricing.

15

Auxiliary services are the additional services an underwriter usually offers. What can be thought of is the promotion of the issue of the price guarantee.

offered by the underwriter when book building is chosen to take the company public. Secondly I expect firms planning to do a relatively small equity offering also more likely to engage in a Dutch-Auctioned IPO. This is because, even though empirically the literature is contradicting, smaller equity offerings are still perceived to be more risky due to larger ex ante uncertainty. This perception combined that Dutch-Auction is a choice leads me to believe that the probability of choosing a Dutch-Auctioned IPO is higher for firms doing a small IPO who are not in the beginning of their firm life cycle. Concluding to do compute all these coefficients and run the model data is needed on the Offer Price, Closing Price at the fist day of trading, Year in which the firm is founded, Date of the IPO, Method of IPO, and the SIC code.

4. Data and descriptive statistics

4.1 Data Sources & Sample

To gather the data needed various data sources have been used. Mainly this thesis has been built on data gathered from the ThomsonOne Database. ThomsonOne is a platform for financial data from Thomson Reuters. Furthermore it is the leading database on M&A and Equity offerings as well as firm characteristics. From this database the Offer Prices, Offerdates, Size of Equity Offerings was gathered. This was all gathered for the sample period of interest, namely January 1999 to may 2014 within the U.S. Then data was needed on which companies performed the Dutch-Auction method of IPO. This is done by looking at the website of WR Hambrecht Co.16 and matching the client list to the data of ThomsonOne. After this the aftermarket data had to be collected. This is done by collecting data from the Center for Research in Security Prices (CRSP), part of the Wharton Research Data Services (WRDS) database. CRSP has all the aftermarket data on firms in the sample however only the closing price on the first trading day was needed. The data in ThomsonOne on the founding years of firms is very limited. Therefore another database had to be consulted for this. Although it is not a common database, as offered by the UVA, Jay Ritter has made a quite significant database on IPO’s. I this data ase a e the fou di g ea s, of all o pa ies that pe fo ed a d IPO o hi h he could sufficiently collect data. Although this slightly thins the sample, this database is combined with the above-mentioned data to have the final sample of 1974 observations.

4.2 Data Description

In Table 5 the summary statistics for the variables of interest are given. The variables of interest being the basic variables used in the analysis are Underpricing, Size of Equity Offering and Firm Age at IPO. The first thing that meets the eye are the means and medians of the entire sample and subsamples. While the entire sample has a mean and median of respectively 32% and 11%, Dutch-Au tio ed IPO’s ha e a ea a d edia of espe ti el % a d %. This is a i di atio

16

that the Dutch-Auctioned IPO indeed decreases the level of underpricing. However when looking at the high standard deviations no conclusions past an indication can be made with this summary statistics. Furthermore table 5 shows the distribution of Size of Equity Offering and the Firm Age of IPO. This table again gives a slight indication that the expectations regarding that Dutch-Auctioned Issues are generally smaller than their book built counterparts. However regarding age, the distribution seems the same. Finally what can be read from table 5 is that there is a negative relationship between both the Size of the Equity Offering and the Firm Age at IPO and the dependent variable underpricing. Meaning that as firms get older the underpricing decreases and as the size of the Equity Offering gets larger the severity of underpricing decreases. Figure 1 shows the distribution of underpricing and shows that the most of underpricing ranges around the 0-10% range.

As described in the methodology section, suspected is that level of underpricing varies between years and industries. First, an analysis will be made if the yearly underpricing differs a lot over time. As Figure 2 shows the average of underpricing indeed varies greatly over time. More interestingly, underpricing seemed to have peeked in the beginning of the sample, 1999-2001, but e ai i g elati el sta le afte that. This pe iod is o o l efe ed to as the dot- o u le 17_.

Severe levels of underpricing characterized this period. Mainly the stock market prices of the tech firms that went public did not nearly represent the intrinsic value behind the actual stocks. Therefore this figure indicates that it is justified control for the difference of underpricing over time.

As the second fixed effect, industry effects might also effect underpricing. To estimate the industry fixed effects on underpricing, the sample is categorized by major division of industry. The major division of the company can be read from the first digit from the SIC code. These divisions are given in more detail in Table 1. The average of underpricing is then plotted in Figure 1 against each of these divisions. What can be seen from Figure 1 is an indication that, at least average,

17

underpricing also varies greatly over time. Again Figure 1 indicates that it is justified to control for the difference of underpricing spread amongst industries.

For the main analysis and the robustness checks performed in this thesis, quantiles have been made. These quantiles have been made in various forms from the variables Age of Firm at IPO and Size of Equity Offering over the entire sample. Furthermore these quantiles have all been transformed into dummy variables and an explanation of this together with the definition of all the variables is given in Table 3. However Table 2 Gives the range of each of quantiles and the number of o se atio s fo the e ti e sa ple a d the su sa ples of ook uilt IPO’s a d Dut h-Auctioned IPO’s. As a additio al ote fo the a ia les Age of Fi at IPO a d “ize of E uit Offe i g, the a ge can be seen as the minimum and maximum of these variables. When looking at the Half Quantiles in Table 2 no conclusions can be made from the number of observations or the distribution between the two subsamples. However moving on to the Tertiles, the distribution of Dutch-Auction becomes apparent. As is shown in Table 2 Dutch-Au tio ed IPO’s see to luste ithin the second Tertile of age, ranging from 6 till 14 years of Firm Age at the time of IPO. Furthermore Dutch Auctions mostly all seem to be small offerings. Another step down in the table shows that there is only one young firm that performed a Dutch-Auction IPO and that 16 of the Dutch-Au tio ed IPO’s e e s all i terms of Size of the Equity Offering. This indicates that the descriptive statistics seem to be in line with the literature and indicate that there might be signs of self-selection within the Dutch-Au tio ed IPO’s.

Where Table 2 hints of correlation, Table 7 provides the tetrachoric correlations between all the dummy variables and Dutch-Auction. Tetrachoric correlations are used because the variables are dichotomous18 and a normal correlation table will not correctly give the correlations. Table 7 shows that there is large positive and significant correlation of 0.4121 between small offering and Auction. Furthermore there is a large negative and significant correlation of -0.3862 between

18

Auction and Young firms which indicates that young firms are less likely to do a Dutch-Auction. Moreover with a positive and significant correlation of 0.2630 Table 7 indicates that medium firms will be the most likely to choose for a Dutch-Auctioned IPO. Finally the table confirms economic theory that younger firms have smaller equity offerings and older firms have larger equity offerings.

5. Results

Because the sample used by this thesis is a relatively small sample, at least the sub-sample with the Dutch-Auctions is, results will be added separately Heckit model as developed by Greene (2003). Correlations have been explained in the Description part as well as the validity of the model. All the results of the equations of Table 4 will be discussed below. Furthermore Table 4 should be read as first the regression equation and then the selection equation representing the two-step mechanism of how the Heckit model works.

5.1 Empirical Results First Stage

Hypothesis 1: Certain firm characteristics will determine the likelihood of a firm performing a Dutch-Auctioned IPO, thus firms “elf-select .

Analyzing Equation (2) it is found that none of the continuous variables are found to be significant. Even more the coefficient for Firm age at IPO even points the wrong direction from what was expected. Therefore for the analysis move on to Equation (2) with dummy variables. In Equation 2 the variables Size of Equity Offering and Age of the firm at the time of the IPO have been split in half according to two quantiles. This is done to create dummy variables to be entered into the selection Equation. When looking at equation (4) it is shown that the first half Quantile of Size of the Equity Offering is positive and significant at the 1% level. This means that all offering in the smaller half of the Equity Offerings being made positively increase the probability of a firm performing a Dutch-Auction. However with a probability to reject the Chi-Square test of 0.207, meaning that it is likely that the correlation of the error terms might be 0, this equation at this step is not of value to the research.

Next there is equation six (6) which employs tertiles to analyze effect of Size of Equity Offering and Age of IPO on the probability of Dutch-Auction. What can be read from this equation is that the coefficient for the first Tertile of Size is positive and significant at the 5% level. Furthermore the coefficient for the first Tertile of Age is negative and significant at the 5% level. These coefficients are the first sign towards the conclusion this thesis is hoping to find. Namely that the

choice for Dutch-Auction is influenced by certain characteristics surrounding the IPO. So speaking in terms of Equations (6) it can be said that the choice for Dutch-Auction is positively influenced by a smaller size and negatively by a younger firm at the 5% significance level.

Following is equation eight (8), which make use of quartiles for the selection equation. These are the regressions of choice for the Heckit model used. Even though in comparison to the previous equation (6) the coefficient for Age of the firm at IPO has a smaller T value, this equation makes use of quartiles instead of tertiles. By separating the Variables Size of Offering and Firm age of IPO into more divisions firms and Issue characteristics can be further singled out to draw conclusions about there impact on the probability for Dutch-Auction. In the Selection Equation the first Quartile for size is significant at the 1% level and First Quantile of Age is significant at the 10% level. These regressions have a high log likelihood19 of -2016, which can be compared to the other models because the sample is the same. Furthermore the Wald Chi square test of the model cannot be rejected at the 5% level.

The first Hypothesis can now be confirmed. Within Dutch-Au tio ed IPO’s there are certain characteristics increase or the probability of a firm choosing such a method of IPO. It is found that Small offerings significantly increase this probability and young firms significantly decrease this probability and thus there is Self-Selection. Now that this self-selection has been confirmed, and the conditions for the Heckit model are thus satisfied, this thesis will move on to testing whether Dutch-Auction indeed reduces underpricing.

19

Log likelihood is the predicted probability of an event happening in a certain case and the actual outcome in that case. Put differently: the log likelihood measures the explanatory strength of a model in logistic regression.

5.2 Empirical Results: Main Effect of Dutch-Auction on Underpricing

In Equation (1) the regression equation for the influence of Firm age at IPO, in units and the Size of Equity Offering is described. Firm Age at IPO has a negative sign and is statistically significant at the 1% level. Economically speaking this can be interpreted that the age of the firm negatively influences underpricing. Or even more that older firms experience less underpricing. This is also what is expected by economic theory that underpricing stems from uncertainty. What is important to note is that Dutch-Auction is significant in these results. This contrasts already with Robinson & Robinson (2012) who found the Dutch-Auction coefficient not to be significant. However, because none of the coefficients affecting Dutch-Auction from selection (first stage) equation are found to be significant this equation does not hold much ground. The same significance of the equation holds for equation (3) where the coefficient for Dutch-Auction is found to be insignificant.

Next is Equation (5) where the Coefficients for the Age of the Firm at the time of IPO and Size of Equity Offering are again significant and positive. Furthermore the Dutch-Auction coefficient is negative and significant implying that after correction for Selection Bias, Dutch-Auction reduces the level of underpricing. However the coefficient of Dutch-Auction in Equation (5) has a first stage of tertiles. Because Dutch-Auction in Equation (7) is composed quartiles, which singles out the firm characteristics even more and still yields comparable results the second hypothesis can also be confirmed. Thus Dutch-Auctioned IPOs reduce underpricing

6. Robustness Checks

In equations (9) and (10) the model controls Industry effects while in equations (13) and (14) the model controls for yearly effects. Finally equations (15) and (16) control for both yearly and industry effects. Industry effects are controlled for by round the SIC codes the first number. So for example for the Financial industry, consisting of SIC codes 6000-6999, the model controls for a SIC of 6. What can be seen from equations 9 and 10 is that controlling for the SIC codes does not add significance nor heightens the log likelihood estimator. However the signs remain the same. The same applies for equations (11), (12), (13) and (14). In all three sets of equations however the coefficient Dutch-Auction is significant at the 1% level and the sign is negative with more or less the magnitude indicating that Dutch-Auction reduces underpricing. Moreover in the last two sets of Equations the sign for the first quartile of Size of Equity Offerings is roughly the same and both are significant at the 5% level. Therefore it can be said that the results from this thesis, that firms self-select into Dutch-Auctioned IPO and that this reduces underpricing are robust for Industry and yearly effects.

Then the question of why Dutch-Auction has such significant effect arises. Because the Heckit, treatment effect model derived from the Heckman selection model, estimates the variables correcting for selection bias these coefficients can have seemingly high numbers. The Heckit treatment effect model estimates these numbers by adding them to equation if the significance passes a certain threshold. The standard threshold, also used by the model in this thesis, is 5% significance. To prove this magnitude an OLS regression will be performed which is common when analyzing underpricing OLS. Table 7 provides an explanation why the Dutch-Auction Coefficient so strongly reduces underpricing. As can be seen from table 7 the first quartile of size of Equity Offering has a decreasing and significant at the 10% level on underpricing. Furthermore the first Quartile of Age has an increasing effect on underpricing significant at the 1% level. These results are also consistent for both fixed as well as Winsorizing. Even more so the R squared even grows, thus the explanatory power of the model also grows, with added fixed effects, clustered standard errors and winsorizing. What is important to note is that the coefficient for Dutch-Auction is insignificant in all

these equations, except Equation (5) but the extremely low squared states that the model has barely any fit. This justifies the use of the Maximum Likelihood estimator in combination with the Heckit model, because the Dutch-Auction coefficient is indeed significant.

Another Robustness test is done in Table 8. Because the selection equation in table 4 determines the probability of Dutch-Auction it should reveal more or less the same signs in the same way and magnitude as in a Probit or Logit regression. Comparing Tables 4 and 8 shows that both in selection equation and the first Probit and Logit regression the coefficients for Size of Equity Offering and Age of Company at IPO are both not significant. The constant however is significant and negative indicating that there is a small chance of performing a Dutch-Auction as firm. This is obvious because the sample employed by this research shows that 1950 firms have chosen book building as a way to go public and 24 have only chosen Dutch-Auction for IPO. Next when comparing Equations (2) and (5) of table 8 with Equation (6) of Table 4 the direction of the coefficients is somewhat similar. However the Probit model in Equation (2) of Table 8 seems to be a closer replication of the Selection Equation used by the Heckit model. The constant term of the selection model is similar again to the Probit model and similar in direction to the Logit model. Finally when taking Fixed effects into consideration, Equation (14) of Table 4 and Equation (3) of Table 8 also point to the same direction. However with the Probit model of Table 8 in Equation (3) the constant term for the Dutch-Auction probability is not significant. On a final note, it is quite logical that the Probit model shows more similarities with Selection Equation because that also uses a modified Probit model.

As a final robustness check underpricing over time is evaluated in Table 9. The same heckit model is used as in the main research. From Equations (1) and (2) can be read that Dutch-Auction reduces underpricing over time and that firms planning to do a small IPO will be more likely to do a Dutch-Auctioned IPO. Ho e e the oeffi ie t fo You g Fi s does ’t see to e sig ifi a t anymore. These signs stay practically the same over Equations (3) and (4) where Fixed Effects have been added. When looking at underpricing over 2 weeks, Equations (5) and (6) the coefficients

mostly point in the same direction however now the negative sign for young firms performing Dutch-Au tio ed IPO’s e o es sig ifi a t at the % le el. Then underpricing after 60 days is reviewed. What is noticeable from Equation (7) is that the Coefficient for Dutch-Auction as well as the constant almost triple in size. This however can be explained that the price of stock rises and thus underpricing, which is the price of the stock relative to the Offer Price, also becomes larger. Furthermore the significance and the magnitude of the signs of the dummy variables vary a bit over time as shown in Equations (7), (8), (9), (10), (11) and (12) however even over time there still is self-selection within firms performing a Dutch-Auctioned IPO and this reduces underpricing.

7. Conclusion:

Although the subject of IPO underpricing is one of the most heavily researched phenomena within the academic literature, both qualitively and quantitively, research about the Dutch-Auction is limited. Research performed towards whether Dutch-Au tio ed IPO’s edu e u de p i i g is contradicting to say the least. Jaganathan and Sherman (2004) find empirically that in all the 47 countries they use in their sample where IPO auctioning previously was the way to go, these methods have now been abandoned in favour of bookbuilding. However after empirically studying the French IPO market Derrien and Womack (2003) find that under certain conditions Auctioned IPO’s a e e de ease underpricing. The qualitative literature is even contradicting regarding Dutch-Au tio ed IPO’s. Whe e as “he a a d Tit a e plai that Dut h-Au tio ed IPO’s will harm price discovery because of the free rider problem Degeorge et al. (2010) found yet again contradicting evidence. Finally Robinson & Robinson (2012) found that Dutch-Au tio ed IPO’s a e ge e all less u de p i ed ho e e the did ’t fi d the oeffi ie t fo Dut h-Auction to be significant and therefore could not draw any conclusions.

As explained above, there is a big gap literature and from this the research question on which this thesis is built arises. Does Dutch-Auction reduce underpricing and do firms self select into this process of IPO? This thesis uses an U.S. sample of 1974 IPO’s f o hich 24 are Dutch-Au tio ed IPO’s. Fu the o e to o e t fo “ele tio ias a he kit odel, so e fo of the Heckman Selection model is used, to correct for Self-Selection within Dutch-Auction. The outcome value of this Heckman Selection model, IPO characteristics that increase the probability of Dutch-Auction, are then entered into a regression equation where the binary variable Dutch-Auction is regressed against underpricing. Furthermore for the IPO characteristics of Size of Equity Offering and Age of Firm at the time of IPO quantile dummies have been made from 2 to 4.

These tests delivered mainly statiscally significant results. For the first of Hypothesis that certain IPO characteristics influence the choice of Dutch-Auction, and thus there is Self-Selection within this model is found to be true. The smallest equity Offerings are found to have a statiscally

significant positive coefficient for Dutch-Auction. Furthermore young firms, in terms of firm age at the time of IPO, are found to have a statistically negative coefficient on Dutch-Auction. And finally when this Dutch-Auction coefficient from the Selection Equation is entered into the Regression equation it is found to have a strong positive effect, and thus reducing underpricing. What is most likely to be reason for this strong positive is that firms self-select into this form of IPO. The characteristics small offering and young firm however, after found by OLS, have a respectively strong positive and negative effect on underpricing. These results and models have all undergone extensive robustness tests as well as controlling for Industry/year effects however the signs remained the sign and largely significant.

Because the existing literature is as contradicting as it is these findings are neither in line nor out of line with it. It might complement the study by Robinson & Robinson (2012) by adding to their findings that Dutch-Auction coefficient is significant because of the Self-Selection. This also might add to the qualitative literature by adding reasons for Dutch-Auction. However in no way does it give a e pla atio fo Jaga atha a d “he a of h Au tio ed IPO’s ha e ee a a do ed over the world. What it adds on itself is that there is Self-Selection within the firms who chose for Dutch-Auction and that Dutch-Auction adds value.

Regarding policy implications of this study, there are definitely some for certain implied by this research. If a company would do a small IPO and is not relatively young, the Dutch-Auction method would definitely add value over normal IPO. Limitations of this study are the relatively small amount of Dutch-Auctioned issues and the availability of Data. For example, would the results still stand with 200 Dutch-Auctioned issues? Or do managers themselves know they are self-selecting and explicitly choosing for the Dutch-Auctioned method when they are within the small size of equity offering range and are a relatively young firm. These are all topics that could be research in future work.

References

Anand, Anita Indira. "Is the Dutch auction IPO a good idea." Stan. JL Bus. & Fin. 11 (2005): 233.

Baron, David P. "A model of the demand for investment banking advising and distribution services for new issues." The Journal of Finance 37.4 (1982): 955-976.

Benveniste, Lawrence M., et al. "Evidence of information spillovers in the production of investment banking services." The Journal of Finance 58.2 (2003): 577-608.

Benveniste, Lawrence M., and Paul A. Spindt. "How investment bankers determine the offer price and allocation of new issues." Journal of Financial Economics 24.2 (1989): 343-361.

Benveniste, Lawrence M., and William J. Wilhelm. "A comparative analysis of IPO proceeds under alternative regulatory environments." Journal of Financial Economics 28.1 (1990): 173-207.

Boehmer, Ekkehart, and Raymond PH Fishe. "Do underwriters encourage stock flipping? A new explanation for the underpricing of IPOs." Unpublished working paper. University of Miami (2000).

Brau, James C., and Stanley E. Fawcett. "Initial public offerings: An analysis of theory and practice." The Journal of Finance 61.1 (2006): 399-436.

Brennan, Michael J., and Julian Franks. "Underpricing, ownership and control in initial public offerings of equity securities in the UK." Journal of Financial Economics 45.3 (1997): 391-413.

Chen, Hsuan‐ Chi, and Jay R. Ritter. "The seven percent solution." The Journal of Finance 55.3 (2000): 1105-1131.

Daily, Catherine M., et al. "IPO underpricing: A meta analysis and research synthesis." Entrepreneurship Theory and Practice 27.3 (2003): 271-295.

Degeorge, François, François Derrien, and Kent L. Womack. "Auctioned IPOs: the US evidence." Journal of Financial Economics 98.2 (2010): 177-194.

Derrien, Francois, and Kent L. Womack. "Auctions vs. bookbuilding and the control of underpricing in hot IPO markets." Review of Financial studies 16.1 (2003): 31-61.

Ellis, Katrina, Roni Michaely, and Maureen O'hara. "When the underwriter is the market maker: An examination of trading in the IPO aftermarket." The Journal of Finance 55.3 (2000): 1039-1074.

Gao, Xiaohui, Jay R. Ritter, and Zhongyan Zhu. "Where have all the IPOs gone?." Journal of

Financial and Quantitative Analysis 48.06 (2013): 1663-1692.

Greene, William H. Econometric analysis. Pearson Education India, 2003.

Habib, Michel A., and Alexander P. Ljungqvist. "Underpricing and IPO proceeds: A note." Economics Letters 61.3 (1998): 381-383.

Ibbotson, Roger G., Jody L. Sindelar, and Jay R. Ritter. "Initial public offerings." Journal of Applied

Corporate Finance 1.2 (1988): 37-45.

Jagannathan, Ravi, Andrei Jirnyi, and Ann Sherman. "Why have IPO auctions failed the market test." Unpublished paper, Northwestern University (2009).

Kaneko, Takashi, and Richard H. Pettway. "Auctions versus book building of Japanese IPOs."

Pacific-Basin Finance Journal 11.4 (2003): 439-462.

Logue, Dennis E. "On the pricing of unseasoned equity issues: 1965–1969."Journal of Financial and

Quantitative Analysis 8.01 (1973): 91-103.

Lowry, Michelle, and G. William Schwert. "IPO market cycles: Bubbles or sequential learning?." The

Lowry, Michelle, Micah S. Officer, and G. William Schwert. "The variability of IPO initial returns." The

Journal of Finance 65.2 (2010): 425-465.

Ljungqvist, Alexander, and William J. Wilhelm. "IPO pricing in the dot‐ com bubble." The Journal of

Finance 58.2 (2003): 723-752.

Marshall, Beverly B. "The effect of firm financial characteristics and the availability of alternative finance on IPO underpricing." Journal of economics and Finance 28.1 (2004): 88-103.

Robinson, Richard M., Mary Ann Robinson, and Chien-Chih Peng. "Underpricing and IPO ownership retention." Journal of Economics and Finance28.1 (2004): 132-146.

Robinson, Mary Ann, and Richard Robinson. "Dutch-auction IPOs: institutional development and underpricing performance." Journal of Economics and Finance 36.3 (2012): 521-554.

Rock, Kevin. "Why new issues are underpriced." Journal of financial economics 15.1 (1986): 187-212.

Sherman, Ann E. "Global trends in IPO methods: Book building versus auctions with endogenous entry." Journal of Financial Economics 78.3 (2005): 615-649.

Appendix

Figure 1

This figure represents the relative frequency or distribution of underpricing in percentage. It is combined with a normal density plot.

Table 1:

SIC codes by first digit

This table represents the division represented by the first digit of the SIC codes.

First Digit

0 Agriculture, Forestry, Fishing

1 Mining

2 Construction

3 Manufacturing

4 Transportation, Communication, Electric, Gas, and Sanitary Services

5 Wholesale Trade

6 Retail Trade

7 Finance, Insurance and Real Estate

Figure 2

Average of underpricing by SIC code division

This figure represents the average of underpricing among the divisions as defined by the SIC codes. Underpricing is defined as the initial first day price jump or the difference between offer price and the closing price on the first day of trading.

Figure 3

Average of underpricing by Offer Year

This figure represents the average of underpricing over time. Underpricing is defined as the initial first day price jump or the difference between offer price and the closing price on the first day of trading.

Table 2

Name, Range and frequency of dummy variables regarding Age of Firm at IPO and the Size of Equity Offering.

This table represents the different quantiles made to represent all the dummy variables regarding Age of the Firm at IPO and the Size of the Equity Offering. The observations are all made between 1999 and 2014, from when the OpenIPO process was introduced in the U.S. The range represents the cutpoints, or range, of the quantiles. The following columns represent the frequency, or number of obse atio s, fo espe ti el the e ti e sa ple, the ook uilt IPO’s a d the Dut h-Auctioned IPO’s.

Entire Sample Book Building DutchAuction

Range # Of Observations # Of Observations # Of Observations

Age of Firm at IPO (0, 175) 1974 1950 24

Size of Equity Offering (0.1, 16006.9) 1974 1950 24

First Half Quantile Age (0, 8) 1025 1012 13

Second Half Quantile Age (9, 175) 949 938 11

First Half Quantile Size (0.1, 96.5) 987 969 2

Second Half Quantile Size (96.6, 16006.9) 987 981 4

First Tertile Age (0, 5) 670 667 3

Second Tertile Age (6, 14) 667 654 13

Third Tertile Age (15, 175) 637 629 8

First Tertile Size (0.1, 69) 666 650 16

Second Tertile Size (69.3, 140.2) 650 647 3

Third Tertile Size (140.5, 16006.9) 658 653 5

Young (0, 4) 537 536 1 Medium (5, 8) 488 476 12 Old (9, 19) 466 462 4 Oldest (20, 175) 483 476 7 Small (0.1, 56.6) 494 478 16 Moderate (56.9, 96.5) 493 491 2 Large (96.6, 183.3) 492 490 4 Huge (183.5, 16006.9) 495 491 24

Table 3 Variable Definitions

Table 3 gives the variable definitions as presented in the other tables with descriptive statistics and regression outcomes. Notes to be made are that Firm age is calculated as stated in the definition in contrast to some other literature, which calculate firm age as (Offer Year-Founding Year +1).

Variable Definition

Underpricing

First day trading return of a stock, or the percentual difference between the Offer Price and closing price on the first day of trading

Firm Age at IPO

The age of the firms at the time it performed its IPO. This can be measured as the difference between the Founding Year and the Offer year

Size of Equity Offering The size of the IPO

Year Fixed Effects Multiple dummies incorporating the yearly fixed effects on Underprcing Industry Fixed Effects

Multiple dummies incorporating the industry fixed effects, as defined by SIC divisions on underpricing

First Half Quantile age

A dummy that equals one if the observation is in the first half Quantile of Firm Age at IPO

First Half Quantile Size

A dummy that equals one if the observation is within the first half quantile of the Size of Equity Offerings

First Tertile Age

A dummy that equals one if the observation is in the first Tertile of Firm Age at IPO

First Tertile Size

A dummy that equals one if the observation is within the first Tertile of the Size of Equity Offerings

Second Tertile Age

A dummy that equals one if the observation is in the second Tertile of Firm Age at IPO

Second Tertile Size

A dummy that equals one if the observation is within the second Tertile of the Size of Equity Offerings

Young

A dummy that equals one if the observation is in the first Quartile of Firm Age at IPO

Medium

A dummy that equals one if the observation is in the Second Quartile of Firm Age at IPO

Old

A dummy that equals one if the observation is in the third Quartile of Firm Age at IPO

Oldest

A dummy that equals one if the observation is in the Fourth Quartile of Firm Age at IPO

Small

A dummy that equals one if the observation is within the First Quartile of the Size of Equity Offerings

Moderate

A dummy that equals one if the observation is within the Second Quartile of the Size of Equity Offerings

Large

A dummy that equals one if the observation is within the Third Quartile of the Size of Equity Offerings

Huge

A dummy that equals one if the observation is within the Third Quartile of the Size of Equity Offerings

Table 4 a-d Relationship between Underpricing, Dutch-Auction, Firm and issue characteristics and Fixed effects over the period 1999-2014 in the U.S.

This table reports all the outcomes of main model, Heckit model, as used in the research done by this thesis. All the coefficients in the tables are maximum likelihood estimates of the models. The tables 5a till 5d according to the how they are analyzed, some extra variables or fixed effects are added every set of equations for robustness and to get further insight to the estimates. Regression Eq. is short for Regression Equation and Selection Eq. is short for Selection Equation. Furthermore the log likelihood is shown. Another Coefficient to be reported is Probability to reject Chi-Square, of the LR test of independence. Because the treatment effect model assumes the level of correlation between the two error terms is nonzero, the thesis assumes the choice for Dutch-Auction is endogenous and thus error terms are related, a violation of this assumptions leads to biased estimation results. This ratio test compares the joint likelihood of an independent probit model for the selection equation and a regression model on the observed data against the likelihood of the treatment model used. If the test is reject the conclusion can be made that the error terms are correlated and thus the treatment effect model is a non-biased estimator. Finally all the standard errors have been computed using White’s (1980) heteroskedasticty consistent formula. The exact form of all the equations is given below:

(1): Underpricingi = (2) Prob(Dutch-Auction)= (3): Underpricingi = (4): Prob(Dutch-Auction)= (5): Underpricingi = (6): (7): Underpricingi = (8):Prob(Dutch-Auction)= (9): Underpricingi = (10):Prob(Dutch-Auction)= (11): Underpricingi = (12):Prob(Dutch-Auction)= (13): Underpricingi = (14):Prob(Dutch-Auction)=