Internet IPO Underpricing and Underwriter Prestige
Sebastiaan Hersmis 10263470
Bachelor Economics and Business Specialisation Economics and Finance Supervisor R. Matta PhD
2
Abstract
This thesis investigates the relation between underpricing of internet Initial Public Offerings -IPOs - and the presence of a prestigious underwriter. When firms go public, an initial stock price is set and a (syndicate of) underwriter(s) sells the initial stock to investors. On average, this stock price rises on the first day of public trading. This means the IPO was initially
underpriced. The data used in this thesis are obtained
from the Thomson ONE database. The
sample consists of 1367 firms that have traded stocks in the period between 01-01-1999 and 01-04-2012 and are classified as U.S. firms. Firstly, we find that the level of underpricing is higher for internet IPOs, among other things because internet IPOs carry more risk. Secondly, we find that the level of underpricing is higher if the IPO has a prestigious underwriter, arguably because of the objective of the prestigious underwriter. Thirdly, we find that IPOs during the Dotcom bubble were more underpriced than IPOs after the Dotcom bubble. Concluding, the final results of this thesis confirm our expectations formed on the basis of prior research. We find that if a prestigious underwriter is present for an Internet IPO, the IPO is even more underpriced than without a prestigious underwriter. Finally, the validity of the result is discussed and some follow-up questions are being asked.
3
Acknowledgements
Upon the completion of this work, I wish to thank the following people who have helped me considerably. I would like to thank Professor J. Ritter, who provided a lot of useful data on IPOs on his website, which have proved to be very useful for the research for this thesis. I am most grateful to Dr. R. Matta for the help he provided during this period. After he had piqued my interest for IPOs in the lectures of Advanced Corporate Finance, his valuable advice and comments have contributed greatly to this thesis.
4
Table of contents
Abstract ... 2
Acknowledgements ... 3
List of tables and figures ... 6
List of tables ... 6
List of figures ... 6
Chapter 1 Introduction ... 7
Relevance of topic ... 7
Underpricing and underwriter prestige ... 7
Research questions ... 8
About this paper and outline ... 8
Chapter 2 Previous Research ... 10
2.1 IPOs ... 10
IPO process and important factors ... 10
Underpricing in general ... 10
2.2 Underwriters ... 11
The underwriter and determining the pricing ... 11
2.3 Mechanics of underpricing ... 11
Premise 1 Internet IPOs have higher levels of underpricing than non-internet IPOs ... 11
(1) Underpricing makes up for risk ... 11
(2) Internet IPOs are riskier as the internet market is uncertain ... 12
(3) Internet IPOs are riskier as the issuing firms are younger ... 12
Premise 2 IPOs during a bubble have higher levels of underpricing than IPOs after that bubble ... 13
(1) A ‘hot’ market has higher underpricing ... 13
Premise 3 Prestigious underwriters have higher incentives to underprice than non-prestigious underwriters ... 13
(1) Underwriters have an incentive to underprice ... 14
(2) Prestige of the underwriter influences underpricing ... 14
Premise 4 Prestigious underwriters underprice internet IPOs more ... 15
Chapter 3 Data ... 17
Chapter 4 Assumptions and Hypotheses ... 18
Assumptions ... 18
Premises ... 19
5
Chapter 5 Research Methodology ... 21
5.1 Variables ... 21
5.2 Multicollinearity ... 22
The Correlation of Variables method ... 22
The Variance Inflation Factor method ... 22
5.3 Regression model ... 23 Testing hypothesis 1 ... 23 Testing hypothesis 2 ... 23 Testing hypothesis 3 ... 23 Testing hypothesis 4 ... 24 Chapter 6 Results ... 25 6.1 Descriptive statistics ... 25
6.2 Results from regressions ... 25
Results from regression 1 ... 26
Results from regression 2 ... 26
Results from regression 3 ... 26
Provisional conclusions from regressions 1, 2 & 3 ... 27
Results from regression 4 ... 28
Chapter 7 Conclusions ... 29
Sub question (i) ... 29
Sub question (ii) ... 29
Sub question (iii) ... 30
The main research question ... 30
Summarizing our conclusions ... 31
Chapter 8 Discussion and Further Questions ... 32
Validity of the assumptions ... 32
Possible shortcomings ... 32
Further Questions ... 33
List of references ... 34
6
List of tables and figures
List of tables
Table 1 (a, b & c) Underpricing of Internet IPOs and other IPOs, 1999-2012) ... 36
Table 2 (a, b & c) Average age of Internet IPOs and other IPOs, 1999-2012 ... 38
Table 3 Construction of the sample of all IPOs, 1999-2012 ... 40
Table 4 Correlation matrix ... 41
Table 5 Variance Inflation Factor ... 42
Table 6 Descriptive statistics ... 43
Table 7 Results from regressions ... 44
List of figures Figure 1 Number of Internet IPOs per year, 1990-2012 ... 16
7
Chapter 1 Introduction
Relevance of topic
From 1999 until 2012 more than 3000 firms in the United States went public through an Initial Public Offering (IPO). A substantial part of these firms is the group of internet firms. These firms conduct their entire business online, offer the majority of their products online and/or are highly internet-related at the time that the firms went public.
After the unusually high number of internet-firms IPOs during 1999-2000, the number of internet IPOs returned to its level from before 1999. Despite the ‘normal’ level of IPOs, much media attention has been given to the valuation of the internet companies at their Initial Public Offering (Woo et al., 2001, Worthen, B., 2012). The great deal of attention in the media for internet companies going public and the lack of recent academic literature on underpricing of specifically internet companies give rise to this thesis. In this thesis the effect of the presence of a prestigious underwriter in relation to underpricing of internet IPOs will be tested.
Underpricing and underwriter prestige
The change in the price between the moment of issuing the stock and the end of the first-day of trading is the first-day return. The first-day return of all IPOs from 1999 until April 2012 was 26.6% on average (Table 1a). As 73%1 of all IPOs were underpriced, positive first-day returns were no exception. This phenomenon has been the subject of much research.
Especially during the Dotcom bubble (1999-2000) first-day returns were extremely high. A large share of these returns was due to internet IPOs. The average first-day return of internet-companies, in the period 1999-2000, was 111%, as opposed to an average of 42.11% for non-internet IPOs (Table 1c).
But also after the Dotcom bubble the first-day return of internet companies compared to non-internet companies remained significantly higher. After the Dotcom bubble the change in the first-day price of internet companies was on average 12.5% higher than the change in the price of non-internet companies (Table 1b).
8
Many theories about underpricing have been developed and tested. Some articles suggest that the prestige of the underwriter plays an important part. The underwriter is usually an investment bank that assists the issuing firm in going public. The effect of the presence of a prestigious underwriter in relation to underpricing of internet IPOs will be tested in this thesis.
Research questions
With the major media attention and the lacking academic literature about Internet IPO underpricing after the Dotcom bubble, questions arise.
i. Is there a significant difference between the level of underpricing for internet firms and non-internet firms?
ii. Is the level of underpricing for internet IPOs in the years after the bubble significantly different from the level of underpricing during the bubble?
iii. What is the effect of the presence of a prestigious underwriter on underpricing IPOs?
These questions are captured in the main research question:
“What is the effect of the presence of a prestigious underwriter on the level of underpricing of internet Initial Public Offerings?”
About this paper and outline
Although the motivations and factors of the underpricing of internet firms during the Dotcom
bubble have been well documented by academic literature, less has been written about the
underpricing of internet firms at an IPO in the period after the Dotcom bubble and the possible relation with underwriter prestige.
In this thesis the focus is on underpricing Internet stocks at an IPO level, for the period 1999-2012, where the Dotcom bubble is defined as the years 1999-2000. The effect of the level of prestige of the underwriter in relation to the level of underpricing is investigated. The chapters have been developed as follows:
Chapter 2 includes the characteristics and background of the internet industry and the IPO process and reviews previous academic research.
9
Chapter 4 introduces the assumptions and hypotheses.
Chapter 5 describes the research methodology, the relevant variables and the model we considered.
Chapter 6 deals with tests and results. Chapter 7 presents the conclusions
10
Chapter 2 Previous Research
In this chapter we review the prior literature on which our model is based. In particular, it includes literature on internet IPOs during the Dotcom bubble, literature on underpricing in general and literature on the effect of (prestigious) underwriters on the level of underpricing. 2.1 IPOs
In this section, characteristics of IPOs are explained. These characteristics include the IPO process and the pricing of an IPO, including underpricing.
IPO process and important factors
When a firm goes public, it sells its shares through an IPO. In order to do this, the issuing firm needs the help of an underwriter: an investment banking firm that organizes and designs the offer. This underwriter collects information from (informed) institutional investors, sets the IPO price and decides whom the shares are given to (allocation of the shares) (Derrien, F., 2005). After the IPO price is set and the shares are allocated, the initial investors can trade the shares on the stock public market. This can cause the price to change to a different first-day price (price at the end of the first day).
Underpricing in general
Underpricing (or first-day return; academics use the terms interchangeably) is a common feature of the IPO process. If a new stock has a higher closing price (price after the first day of trading) than the original IPO price, the stock is underpriced. This means that the market is willing to pay a higher price for the stock than the initial investors have paid.
Underpricing indirectly costs the issuing firm a lot of money, as the difference between the valuation on the basis of the IPO price and the valuation of the market (after the first day trading), is ‘money left on the table’. This money does not go to the issuing firm, but to the investor who got the stock allocated from the underwriter.
Ritter and Welch (2002) found that from 1980 until 2001 the average first-day return on IPOs is 18.8% and argue that underpricing is a persistent feature of the IPO market (p.1816). We follow this line of argument.
11
2.2 Underwriters
As said, when a firm goes public, it generally has a contract with several underwriters. One of those underwriters is the lead/main underwriter; in this thesis further referred to as ‘the
underwriter’. This section elaborates on the actions of the underwriter. The underwriter and determining the pricing
Pricing of an IPO is usually done by an underwriter or by a group of underwriters. Although there are multiple ways according to which underwriters can determine the price of an IPO, the ‘bookbuilding’ method is the one mostly used and considered as a standard practice. (Cornelli, F. & Goldreich, D., 2001, p. 2337).
With the ‘bookbuilding’ method, the underwriter(s), in cooperation with the issuer, sets a price range. After setting the price range, the stock of the firm is presented to
institutional investors to estimate the demand. Then, the issuing firm sets the IPO price and decides to whom the shares go. This illustrates the key role the underwriter plays in pricing (and underpricing) the stock.
2.3 Mechanics of underpricing
This section explains the mechanics of underpricing and elaborates on the possible reasons for this underpricing, including the prestige of underwriters. Based on previous research, we form four premises that lay out the foundation of this thesis
.
Premise 1 Internet IPOs have higher levels of underpricing than non-internet IPOs
(1) Underpricing makes up for risk
When a firm goes public, together with the underwriter it issues a prospectus which includes the basic financial data of the issuing firm. Naturally, the future of the firm is uncertain as not all possible information can be revealed. Especially when the future is unforeseen, investing in this stock is risky. In general, underpricing compensates initial investors for taking this risk. The higher the uncertainty, the higher the (first-day) return required by investors. Brau and Fawcett (2006, p. 425) indeed find that CFOs from companies feel that underpricing compensates investors for taking risk.
12
(2) Internet IPOs are riskier as the internet market is uncertain
The internet market is a relatively new and uncertain market (Loughran & Ritter, 2004). Internet firms are difficult to value, as the ‘original’ drivers used to value firms cannot be (fully) applied to internet firms (Hand, 2000).
This was confirmed by Trueman et al (2000), who investigated the association between bottom-line net income and the matching sample firms’ market prices and did not find a significant relation. As valuing internet firms remains ambiguous, investing in internet firms carries more risk.
(3) Internet IPOs are riskier as the issuing firms are younger
Another characteristic of internet firms at an IPO is the significantly lower age at which internet companies go public (Demers and Joos, p. 340). Demers and Joos showed that age is negatively correlated with the change of an IPO failure (2006, p. 394). As internet-firms have a significant lower age, they also have a higher risk of failure.
The results of Demers and Joos (2006) are confirmed by the results in table 2. Table 2a shows that internet IPOs are significantly younger (1% significant level) than non-internet IPOs. And although internet IPOs tend to be older after the Dotcom bubble period than during the bubble (1% significance level), they still remain significantly younger than non-internet firms (1% significance level).The age of non-internet IPOs did not significantly change after the Dotcom bubble (10% level).
The higher perceived risk, caused by the characteristics of internet market and the lower average age of internet firms entails higher levels of uncertainty for internet IPOs. Consequently, this higher level of uncertainty leads to a higher level of underpricing. As indicated in table 1, internet firms indeed show significantly higher levels of underpricing for the Dotcom period (t-stat. of diff. = -6.2841) as well as for the period after the Dotcom period (t-stat. of diff. = -4.0016). This confirms that internet IPOs are on average more underpriced than non-internet IPOs.
13
Premise 2 IPOs during a bubble have higher levels of underpricing than IPOs after that bubble
(1) A ‘hot’ market has higher underpricing
Ljungqvist, Nanda and Singh (2006) modelled the IPO process and concluded that the existence of ‘irrationally exuberant’ investors at an IPO leads to high first-day returns (short run) and eventually underperforming (long run). This is very applicable to the Dotcom bubble, which can be characterized as a ‘hot’ market. This can be seen in figure 1, which shows that the number of internet IPOs during the period 1999-2000 was extremely high. This result was also confirmed by Loughran and Ritter (2002) who showed that there is more ‘money left on the table’ after a market rise than after a market fall. In other words, the level of underpricing is higher after a market rise, which is the case in a ‘hot’ market. This concept is confirmed by the data in table 1, which shows significant differences in underpricing, regarding both internet and non-internet IPOs, between the bubble period and the period after the bubble.
Premise 3 Prestigious underwriters have higher incentives to underprice than non-prestigious underwriters
Different underwriters may price the stock of an IPO differently. Prior research shows that underwriter prestige might play a role in this.
Cooney, Singh, Carter, and Dark (2001, p. 294) compared three methods for
estimating the prestige of an underwriter. They compared the ranking of Johnson and Miller (1988), Megginson and Weiss (1991) and Carter and Manaster (1998) and found that the ranking based on Carter and Manaster yielded the most significant results. Carter and Manaster ranked underwriters according to the position of the underwriter at the tombstone announcement2. In this thesis the underwriters are ranked according to Ritter (2011), who
2 A tombstone announcement is a listing of a pending public security offering where the underwriters which are part of the syndicate are mentioned. Carter and Manaster created the list by comparing tombstone
14
based his ranking on the ranking of Carter and Manaster3 Premise 3 is built upon two arguments:
(1) Underwriters have an incentive to underprice
If the stock is heavily underpriced and the underwriter can decide whom to sell the stock to, the underwriter can knowingly sell underpriced stock to the clients of the underwriter. These clients receive underpriced stock that can be most likely sold at a higher price on the stock market and thus make a profit.
Ritter and Loughran (2002) confirm that the clients (investors) are willing to offer ‘quid pro quos’ to the underwriters, when the underwriter allocates ‘cheap’ IPO stock to them.
Thus, because of the advantage that the underwriter receives from the clients in return for allocating underpriced stock, the underwriter has an incentive to underprice the stock (Loughran and Ritter, 2002, p. 5).
(2) Prestige of the underwriter influences underpricing
There are differences between the underwriters. Johnson and Miller (1988 p. 28) concluded, after the replication of five studies made earlier, that prestigious underwriters underprice less than non-prestigious underwriters. As the underwriter has a good reputation (enjoys a high prestige), the prospectus is perceived as more credible. Carter and Manaster (1990) found that there indeed was a significant negative relation between the level of prestige and the first-day return (the level of ‘the IPO price run-up’) (1990, p.1062).
This relation, however, changed in 1990s. Beatty and Welch (1996, p. 584) and Cooney, Singh, Carter, and Dark (2001, p. 295), found that the significantly negative relation reversed from 1980s to the 1990s. After the 1990s the relation between the prestige of the underwriter and the level of underpricing has been significantly positive. According to Ritter and Loughran (2004), the reasoning that especially prestigious firms underpriced stocks in 1990 is twofold. More specifically:
3
Instead of announcements in newspapers, Ritter ranked the underwriters according to the announcement in prospectuses.
15
1. As prestigious firms wanted to gain more market share, they chose not to charge higher direct fees to the issuing firm, but instead left more ‘money on the table’ and focused more on advantages from allocating their other clients underpriced (‘cheap’) stocks. These advantages usually included their clients shifting (a larger share of) their business to the underwriter. As the underwriter is prestigious, it is fair to expect that this underwriter will have more underpriced IPO stock in the future.
Non-prestigious underwriters, however, valued the relation with the issuing form more and did not leave much ‘money on the table’; they underpriced less. As they are non-prestigious, potential clients are less inclined to shift business to the underwriter, as non-prestigious underwriters are not expected to handle many other IPOs in the future
2. In the 1990s the more prestigious underwriters shifted into smaller and younger firms. These IPOs were ‘riskier’ and also more difficult to value. Due to the higher risk, investors require a higher compensation (first-day return). Additionally the underwriter can more easily underprice more as the firm is difficult to value.
This argument was confirmed by the findings of Ritter and Loughran (2004) who found that the difference in underpricing between a prestigious and a non-prestigious underwriter was 19.41%.
Premise 4 Prestigious underwriters underprice internet IPOs more
Premise 4 is the key premise for this thesis and will include reasoning used to answer the main research question.
In Premise 1 it can be seen that internet IPOs have a higher uncertainty than non-internet IPOs. The reasoning is twofold:
1. Internet IPOs are riskier as the internet market is uncertain 2. Internet IPOs are riskier as the issuing firms are younger
Pricing an internet IPO is difficult due to the uncertainty caused by a risky market and a lower average age. This leads to underwriters being in a position of liberty when choosing a price, since it is less obvious what a ‘fair’ price would be. Therefore internet firms are an easy target for underpricing.
16 0 50 100 150 200 250 300 1 9 9 0 1 9 9 2 1 9 9 3 1 9 9 4 1 9 9 5 1 9 9 6 1 9 9 7 1 9 9 8 1 9 9 9 2 0 0 0 2 0 0 1 2 0 0 2 2 0 0 3 2 0 0 4 2 0 0 5 2 0 0 6 2 0 0 7 2 0 0 8 2 0 0 9 2 0 1 0 2 0 1 1 2 0 1 2
Number of Internet IPOs
In Premise 3 we argue that prestigious underwriters on average underprice more. The reasoning is twofold:
1. Prestigious underwriters have shifted into ‘riskier’ IPOs
2. Prestigious underwriters value the relation with the issuing firm less than non-prestigious underwriters do
With internet IPOs being easy targets for underpricing and prestigious underwriters intending to underprice more, we argue that if a prestigious underwriter is present, the level of
underpricing of an internet IPO is higher.
Figure 1 Number of Internet IPOs per year, 1990-2012
The sample includes all internet IPOs which were listed on “A List of Internet IPOs” (Ritter, 2012). This figure exhibits the total number of internet IPOs on the y-axis and the years on the x-axis.
17
Chapter 3 Data
From the Thomson ONE database (formerly known as SDC Platinum), we obtained a list of all IPOs that have traded stocks between 01-01-1999 and 01-04-2012 and are classified as U.S. firms.
To classify firms as an ‘Internet-firm’ we used the updated list from Ritter (2012) and manually matched all the names with the names from the Thomson ONE database to correct for the spelling differences.
We excluded all stocks for which a complete dataset4 in suiting our model was not available. This resulted in a sample of 1367 firms. See table 3 for a detailed survey of the sample construction.
After constructing the sample, we matched the name of the firm with the name stated in “a ranking of underwriters” (Ritter). To do this, we matched all exact matches
automatically and the rest we matched by hand, based on the name and IPO date. The prestige of an underwriter is ranked from 1 to 9, the lowest being 1 and the highest being 9. In line with Ritter and Loughran (2004), in this thesis, ‘prestigious underwriters’ are defined as those underwriters with a rank of 8 or higher.
The age (in years) is calculated as the difference between the year of the IPO and the foundation year of the firm +1, with a minimum of 1 year. If the firm is founded in the same year as the IPO, than the ‘age’ of the firm is 1 year. If the firm is founded in the year before the IPO, the ‘age’ of the firm is two years, etcetera.
As a benchmark, in this thesis the S&P 400 is chosen. This benchmark’s feature of not including dividend is not relevant for this thesis (only first-day return) and therefore it serves as a good benchmark.
4
A complete database has to include founding date, IPO date, change of stock price, change of benchmark & at least one underwriter
18
Chapter 4 Assumptions and Hypotheses
This chapter explains the assumptions made and formalizes the premises and research questions in hypotheses.
Assumptions
The assumptions presented below are based on prior literature and each assumption is
explained. Assumptions were made for the sake of simplification and because further research on those was outside the scope of this thesis.
i. Although often a syndicate of underwriters is formed, the IPOs are ranked after the ranking of the lead underwriter
Many issuing firms have not one, but a syndicate of underwriters underwriting the IPO. Schultz and Corwin (2005) suggested that the number of underwriters might influence the level of underpricing, but found no large and/or significant effect (p. 40). Partly due to their results but mostly due to the scope of this thesis, we rank the IPOs after the ranking of the lead underwriter (the ´bookrunner´). This underwriter takes the lead in the syndicate and usually plays the largest role in the IPO process and therefore we argue that this assumption is justified.
ii. Underwriters set the price and can estimate the demand through the bookbuilding method; underpricing is done on purpose (knowingly)
As explained in ‘The underwriter and determining the pricing’, bookbuilding is the method most commonly used to determine the price. Although more methods for determining the price exist, we assume that all IPOs use the bookbuilding method. As a result of this
assumption, at every IPO the underwriter knows the demand for the stock and can determine a ‘fair’ price. This means that if underpricing occurs, the IPO is underpriced on purpose by the underwriter.
iii. From a higher level of underpricing one can infer that the underwriter values advantages from (potential) clients more than the relation with the issuing firm (see ‘Prestige of the underwriter influences underpricing’)
19
From higher levels of underpricing we infer that the underwriter knowingly chooses to underprice more. This higher level of underpricing/higher first-day return will benefit the initial investors, i.e. the investors who got the initial IPO stock allocated from the underwriter, but will put the issuing firm at a disadvantage as more money is ‘left on the table’ (or actually goes to the initial investors). Therefore, we assume that if a higher level of underpricing arises, the underwriter values the relation with potential/other clients above the relation with the issuing firm.
iv. The relation between underwriter prestige and IPO underpricing did not change after the Dotcom bubble, as no literature has reported such change
As highlighted in ‘Prestige of the underwriter influences underpricing’, the relation between a prestigious underwriter and the level of IPO underpricing has changed from a significantly negative relation in the 1980s to a significantly positive relation in the 1990s. As nowhere in literature a reverse has been described, we assume the relation still to be significantly positive, meaning that prestigious underwriters induce higher levels of underpricing. This assumption is checked by testing hypothesis 3 and persisted to create hypothesis 4.
Premises
Based on prior literature, while using the assumptions, the following premises which already discussed in Chapter 2 are formed:
1. Premise 1 Internet IPOs have higher levels of underpricing than non-internet IPOs 2. Premise 2 IPOs during a bubble have higher levels of underpricing than IPOs after
that bubble
3. Premise 3 Prestigious underwriters have higher incentives to underprice than non-prestigious underwriters
4. Premise 4 Prestigious underwriters underprice internet IPOs These premises will be tested formally by testing hypotheses.
20
Hypotheses
Based on the premises above, the following hypotheses have been formed.
H1: Internet IPOs are significantly more underpriced than non-internet IPOs
H2: IPOs in general and additionally Internet IPOs are significantly more underpriced during a
bubble
H3: IPOs with prestigious underwriters are significantly more underpriced than IPOs with
non-prestigious underwriters
H4: The presence of a prestigious underwriter has a significant positive effect on underpricing
internet IPOs
For each hypothesis, a null hypothesis has been formed as well.
null H1: Internet IPOs are not significantly more underpriced than non-internet IPOs
null H2: IPOs in general and additionally Internet IPOs are not significantly more underpriced
during a bubble
null H3: IPOs with prestigious underwriters are not significantly more underpriced than IPOs
with non-prestigious underwriters
null H4: The presence of a prestigious underwriter does not have a significant positive effect
on underpricing internet IPOs
All of these hypotheses are tested, but emphasis is laid on the fourth hypothesis, which is the key-hypothesis of this thesis.
21
Chapter 5 Research Methodology
5.1 Variables
In this chapter the variables used for the regressions to test the hypotheses are presented. For more information on construction of the data, see ‘Chapter 3 Data’.
FIRST-DAY RETURN (NETCHANGE)
The level of underpricing is determined by the first-day return (the change of the stock price on the first day), according to the following formula:
( ) 5
INTERNET DUMMY
If a firm is listed on Ritter’s list of internet IPOs, its internet dummy has value 1, otherwise 0. BUBBLE DUMMY
If a firm has its IPO from 1-1-1999 until 1-1-2001, its bubble dummy has value 1, otherwise 0.
PRESTIGIOUS UNDERWRITER DUMMY
If the lead underwriter of a firm has a ranking of 8 or higher in Jay Ritter´s rank (2011), its dummy variable has value 1, otherwise 0.
INTERACTION TERMS
If the IPO is an internet-IPO with a prestigious underwriter, the interaction term INTERNET DUMMY * PRESTIGIOUS DUMMY has value 1, otherwise 0.
The same goes for the interaction term INTERNET DUMMY *BUBBLE DUMMY,which has value 1 if an IPO is an internet-IPO during the bubble, otherwise value 0.
NATURAL LOG OF AGE (LNAGE)
In line with Beatty and Welch (1996), Cooney, Singh, Carter and Dark (2001) and Ritter and
22
Loughran (2004) we use (the natural log of) age as a control variable. The variable LNage takes the natural log of age, i.e:
( ) 5.2 Multicollinearity
If two or more explanatory variables are highly correlated, multicollinearity may arise. This reduces the precision at which the individual coefficients can be determined, as one is
correlated with another. Two methods are commonly used in checking multicollinearity: The Correlation of Variables and the Variance Inflation Factor - VIF.
The Correlation of Variables method
‘Table 4 Correlation matrix’ exhibits the correlation between variables. There is only one
variable that has a (absolute) correlation of more than 0.756. The variable INTERNET DUMMY correlates with the interaction term INTERNET DUMMY * PRESTIGIOUS DUMMY. Before concluding correlation, the measure of Variance Inflation Factor is used.
The Variance Inflation Factor method
If we look at VIF values (‘Table 5 Variance Inflation Factor’) we see that there are no values which have a VIF of above ten7 (which would indicate possible multicollinearity). The VIF values in table 5 exhibits the VIF values of the full regression (testing the key-hypothesis, derived from premise 4).
Combining the results of the correlation matrix and the VIF values, we see no problems with multicollinearity in our model.
6 As no fixed common cutoff point for determining possible multicollinearity exists, we take 0.75. This is on the lower end (the ‘safe side’) of the band from 0.8 to 0.9, suggested by Mason & Perreault (1991, p.270).
23
5.3 Regression model
The four hypotheses formed in this thesis will be tested by an Ordinary Least Squares - OLS - regression.
Testing hypothesis 1
To test hypothesis 1, we regress NET CHANGE on an INTERNET DUMMY. The natural log of age (LNAGE) is included to smooth the risk as a result of a lower age, which most internet IPOs have (‘Table 1 (a, b & c) Underpricing of Internet IPOs and other IPOs, 1999-2012)’). The model used in this regression is
( )
From prior literature and from the result of the simple comparison of means (see table 1) we expect to see that INTERNET DUMMY is significantly positive, and LNAGE significantly negative.
Testing hypothesis 2
To test hypothesis 2, we regress NET CHANGE on a BUBBLE DUMMY, the interaction term INTERNET DUMMY *BUBBLE DUMMY and the natural log of age (LNAGE). The interaction term INTERNET DUMMY *BUBBLE DUMMY is included as the Dotcom bubble might be a bubble which primarily affects internet IPOs (and thus might not have affected non-internet IPOs). To test this hypothesis, the model used in this regression is
( )
Based on prior literature and Table 2 (a, b & c) Average age of Internet IPOs and other IPOs, 1999-2012, we expect to find that the Bubble Dummy and/or the interaction term (depending on if the Dotcom bubble affected all IPOs and internet IPOs extra) is/are positively
significant. As with the model for testing hypothesis 1, we expect LNage to be significantly negative.
Testing hypothesis 3
To test hypothesis 3, we regress NET CHANGE on the PRESTIGIOUS DUMMY and the natural log of age (LNAGE). To test this hypothesis, the model used in this regression is
24
( ) Based on prior literature and Table 2 (a, b & c) Average age of Internet IPOs and other IPOs, 1999-2012, we expect to find that the Prestigious Dummy is positively significant. As with the model for testing hypothesis 1, we expect LNage to be significantly negative.
Testing hypothesis 4
Hypothesis 4 is the key hypothesis of this model. To test this hypothesis, we regress NET CHANGE on the Internet Dummy, PRESTIGIOUS DUMMY, the interaction termPRESTIGIOUS * INTERNET DUMMY, the BUBBLE DUMMY and the natural log of age (LNAGE). To test this hypothesis, the model used in this regression is
( )
Whereas with the first three hypotheses we could make clear expectations, based on previous literature and the results of the comparison of means, for hypothesis 4 no previous literature has been published. Therefore expectations are purely based on our own reasoning, as already explained in Premise 4.
25
Chapter 6 Results
6.1 Descriptive statistics
The data contain 1367 observations of IPOs in the US during the period from 1-1-1999 to 01-04-2012. The descriptive statistics can be found in Table 6. In addition to LNAGE, Age is given (between brackets).
The average first-day return (NETCHANGE) of the complete sample (during and after Dotcom bubble) is 26.60%. This means that the average level of underpricing is 26.60% and this result has low significance, for two reasons. Firstly, we know that during the bubble the average underpricing was significantly higher than after the bubble, as shown in Table 1. Secondly, the standard deviation is relatively high (0.59), causing a less significant value to the ‘simple’ average.
The dummy Internet Firm has a mean of 0.16, meaning that in our sample 16% of the IPOs are classified as internet IPOs. This can be also seen in Table 3, which shows that 219 out of 1367 firms are internet firms (16%). The standard deviation has low significance here, as this variable is a dummy variable.
78% of the IPOs in our sample is underwritten by a prestigious underwriter, and 14% of the IPOs is an internet IPO and underwritten by a prestigious underwriter as well. From the data can be derived that 84.93% of the internet IPOs are underwritten by a prestigious
underwriter.
28% of the IPOS in our sample are issued during the Dotcom bubble. The mean LNage of our sample is 2,05, meaning that the average age of an IPO is 14.61 years. The standard deviation is relatively quite large (18.26), meaning that taking a simple mean has low significance.
6.2 Results from regressions
The results of the regressions with the models as given in 5.3 Regression model are exhibited in Table 7. The standard errors mentioned are Robust errors, to account for possible
26
Results from regression 1
The coefficient of the intercept is highly significant (1% significance) and positive (.184677) The coefficient of the Internet dummy is high and also highly positive (coefficient of .6319014 with a t-value of 8.78; 1% significance). This is in line with the prior research (see reasoning at ‘Premise 1 Internet IPOs have higher levels of underpricing than non-internet
IPOs’).
As expected, the coefficient of age is negative (-.0098794). Although this coefficient is not very significant, this result is in line with previous findings stating that IPOs of younger firms are associated with more risk (yielding a negative coefficient).
This regression as a whole has an F-statistic of 39.16 (highly significant), and an R2 of 0.1536. The value of R2 is not extremely high, so the presence of an internet dummy does not explain the level of underpricing to a high extent.
Results from regression 2
The coefficient of the intercept is highly significant (1% significance) and positive
(.1045456). The coefficient of the intercept corresponds to a 10.45% level of underpricing. The coefficient of the bubble dummy is .3162238, with a t-statistic of 7.13 (a 1% significance level), and the coefficient of the interaction term (internet dummy * bubble dummy) is also significantly positive (1% significance), with a coefficient of .6858316. These coefficients show that in the Dotcom bubble all IPOs had higher levels of underpricing; internet IPOs especially exhibited higher levels of underpricing. This is in line with previous results that ‘hot markets’ (which are bubble periods) show higher levels of underpricing. The coefficient for LNage has almost no influence (.000161) and is not significant. This regression as a whole has an F-statistic of 52.37 (highly significant) and an R2 of 0.2738. Looking at the value of R2, one can see that the chosen variables explain more of the level of underpricing than the variables at regression 2.
Results from regression 3
The coefficient of the intercept is again significant (1% significance) and positive (.1939937). The coefficient of the intercept corresponds to a 19,34% level of underpricing
The prestigious underwriter dummy has a coefficient of .1799152 (1% significance). This means that the presence of a prestigious underwriter has a positive effect on the level of
27
underpricing.
As expected, the variable LNage has a small negative effect on the level of underpricing (1% significance).
This regression as a whole has an F-statistic of 20.63 (highly significant) and an R2 of 0.0187. The extremely low value of R2 shows that this model is a non-appropriate model in explaining the level of underpricing.
Provisional conclusions from regressions 1, 2 & 3
In this subchapter, we combine the results of regressions 1, 2 and 3 and relate them to the underlying hypotheses and premises. This has already been explained in order to assess the validity of regression 4.
Regression 1 confirms the previous research findings explained in premise 1, and therefore it rejects the null-hypothesis of hypothesis 1. Moreover, the result coincides with the data in table 1: internet IPOs are more overpriced than non-internet IPOs. The result from this regression confirms the validity of premise 1.
Regression 2 confirms the prior findings explained in premise 1, and therefore rejects the null-hypothesis of hypothesis 2. Notable is the fact that the Dotcom bubble significantly influenced underpricing of all IPOs, and in addition increased levels of underpricing for internet IPOs. The result from this regression confirms the validity of premise 2.
Regression 3 confirms previous findings and therefore rejects the null-hypothesis of hypothesis 3. As expected, the presence of a prestigious underwriter during an IPO
significantly increases the level of underpricing. Therefore we conclude that the relation between the presence of a prestigious and the level of underpricing did not change after the 1990s. This result confirms the validity of assumption 4 and premise 3.
Premise 4 and hypothesis 4 are the key-premise and key-hypothesis of this thesis. Premise 4 is based on the first 3 premises that have all resulted in being valid premises. The fact that the results of the first three premises corroborate previous findings assures the validity and consistency of our dataset.
28
Results from regression 4
The intercept is significantly positive (5% significance). This is in line with the previous research, according to which underpricing compensates the investors for taking risk. The intercept could thus be expected to be positive.
The variable INTERNET DUMMY is significantly negative. This contradicts previous research which suggested that internet firms have on average higher levels of underpricing . Also, this result contradicts the findings in regression 1.
Two possible explanations for this flaw are:
- Most internet companies are underwritten by prestigious underwriters and the effect of an IPO being an internet IPO is captured in the interaction term: INTERNET DUMMY * PRESTIGIOUS DUMMY. The correlation between Internet Dummy and Prestigious Dummy is more than 0.9 (Table 4).
- As many internet companies went public during the Dotcom bubble, the effect of Internet Dummy is captured in the interaction term INTERNET DUMMY * BUBBLE DUMMY
The variablePRESTIGIOUS UNDERWRITER DUMMY is significantly positive (1% significance). This is in line with previous research, which claimed prestigious underwriters to underprice more (see Premise 3). The effect of a prestigious underwriter is quite small, (± 5.78%).
The Interaction term INTERNET DUMMY * PRESTIGIOUS DUMMY is significantly positive (1% significance). This confirms the expectations based on the argumentation as shown in premise 4.
The dummy variable BUBBLE DUMMY is very positive and highly significant (1% significance). This is in line with previous literature and data in table 1 and with the results from regression 2.
The interaction term INTERNET DUMMY * BUBBLE dummy is significantly positive (1% significance level). This is in line with previous literature and data in table 1, and with the results from regression 2.
The control variable, LNage, is negative, but not significant. Previous literature suggests that as younger firms carry more risk, investors require a higher level of (first-day) return, and therefore we could haveexpected a negative coefficient.
29
Chapter 7 Conclusions
In this chapter we present the results from our regressions and discuss possible shortcomings of the model.
We investigated the research questions as presented in chapter 1 while thereafter the sub-research questions and the main sub-research questions are examined. The regression results are discussed in Chapter 6 R and summarized in table 7.
Sub question (i) “Is the level of underpricing for internet IPOs in the years after the bubble significantly different from the level of underpricing during the bubble?” broached the subject of the difference in levels of underpricing for internet IPOs and non-internet IPOs. According to literature regarding the difference in levels of underpricing for tech-firms, there seems to be a significant difference between the levels of underpricing for internet IPOs and non-internet IPOs. In Premise 1, three arguments are given for the difference. Additionally, table 1a
exhibits a significant difference between both types of IPOs. We have formally tested this sub question and premise through hypothesis 1. The null-hypothesis 1 stated “Internet IPOs are not significantly more underpriced than non-internet IPOs”.
This null hypothesis was tested using regression 1. With 1% significance the null hypothesis can be rejected and it can therefore be concluded that internet IPOs are indeed more underpriced than non-internet IPOs. The difference between the underpricing level of an internet IPO in contrast to a non-internet IPO is 63%, ceteris paribus. This was in line with our expectations, based on previous research. However, certain factors were neglected, and relying only on this regression could be misleading. Some of these factors are included in hypothesis 2, chapter 4.
Sub question (ii) “Is the level of underpricing for internet IPOs in the years after the bubble significantly different from the level of underpricing during the bubble?” highlighted the particular years during the Dotcom bubble. As explained in Previous research, during the period of a bubble the level of underpricing goes up. This is confirmed by the data in table 2. This statement is formalized in the null hypothesis 2: “IPOs in general and additionally Internet IPOs are not significantly more underpriced during a bubble”.
30
a) All IPOs are more underpriced during the Dotcom bubble. This was found to be true: all IPOs were on average almost 37% more underpriced during the Dotcom bubble. b) Internet IPOs are being extra underpriced. This was found to be true as well: if an IPO
was an internet IPO, it was ± 68% extra underpriced, ceteris paribus. For subquestion (ii) we concluded that the null hypothesis is rejected, at a 1% level of significance. Sub question (iii) “What is the effect of the presence of a prestigious underwriter on
underpricing IPOs?” investigated the effect of the presence of a prestigious underwriter. Much literature has been written about the effect of a prestigious underwriter on the level of
underpricing and this has been tested using hypothesis and premise 3. Hypothesis 3 stated: “IPOs with prestigious underwriters were significantly more underpriced than IPOs with non-prestigious underwriters”.
This null hypothesis was tested using regression 3. At a 1% level of significance the null hypothesis 3 is rejected. IPOs with a presitigious underwriter are on average 18% more underpriced than IPOs with non-prestigious underwriters, ceteris paribus. This result is in line with prior research.
The main research question “What is the effect of the presence of a prestigious underwriter on the level of underpricing of internet Initial Public Offerings?” captures the three previous sub questions. This main research question is tested using regression 4, where all the previous mentioned variables and interaction terms are included. The detailed result is presented in “Results” and summarized in table 7. The conclusions drawn are summarized as follows: Internet IPOs (1) not during the Dotcom bubble and (2) without a prestigious underwriter had a significant lower level of underpricing.
The presence of a prestigious underwriter had a significantly positive effect of ± 5% on the level of underpricing.
As expected, IPOs during the bubble and especially internet IPOs during the bubble had a higher level of underpricing.
The key result from regression 4 is that the combination of the presence of a
prestigious underwriter when an IPO is internet IPO had a high, significantly positive effect (± 48%).
31
Therefore, the null hypothesis 4 is rejected.
Summarizing our conclusions, in this work, we show that:
(1) Our dataset is consistent with datasets used in previous literature.
(2) The provisional conclusions that premise 1-3 will result in being valid premises were confirmed. As Premise 4 is partly built upon premise 1-3, we assumed that Premise 4 is valid. (3) The results of regression 4 allowed us to assert that internet IPOs that are not in a bubble and do not have a prestigious underwriter have a significantly lower level of underpricing. (4) IPOs during the Dotcom bubble and especially Internet IPOS during a Dotcom bubble have higher levels of underpricing
(5) The presence of a prestigious underwriter causes higher levels of underpricing (6) Internet IPOs with a prestigious underwriter cause especially higher levels of underpricing.
We rejected the null hypothesis 4 and our answer to the main research question is that the effect of the presence of a prestigious underwriter on the level of underpricing of an Internet IPOs is significantly positive. This is in addition to the separate effects of an IPO being an internet IPO and the presence of a prestigious underwriter for all IPOs.
32
Chapter 8 Discussion and Further Questions
In this chapter we discuss the validity of the assumptions made, possible shortcomings of the model and suggest ideas for subsequent research.
Validity of the assumptions
According to assumption 1 the IPOs can be ranked after the ranking of the lead underwriter. However, often a syndicate is formed and therefore the ranking of IPOs might not have been accurate. A (weighted) ranking of all underwriters might be an option.
According to assumption 2 underpricing is done on purpose. This assumption is very ‘tricky’, as underwriters will not disclose whether they underpriced on purpose or not. This assumption had to be made for this research and did not influence the outcomes or the results, as the result is about the presence of an underwriter and not about their behaviour.
The same holds for assumption 3: this assumption 3 can be made with safety. We concluded with respect to the presence of a prestigious underwriter; we did not make statements about the motives of the underwriter.
Assumption 4 may be further investigated; however that is beyond the scope of this research. The relation between underwriter prestige and IPO underpricing is not fixed over time, per se. For this research we’ve assumed that the relation did not change after 1990s; this assumption can be confirmed by the results of regression 3
Possible shortcomings
This research is limited in scope, as it only focuses on the effect of a presence of a prestigious underwriter on the level of underpricing. From prior literature we know that more factors could influence underpricing. More factors were not taken into account in this work. In this research the Dotcom bubble was considered to be a general “bubble” period. The fact that the Dotcom bubble is generally viewed upon as a clear example of a bubble, cannot exclude the specific characteristics that might make it unfit to be generalized as ‘a general bubble period’. Therefore, the results found and conclusions drawn in this research may not be externally valid (i.e. we may not be able to make the same conclusions for other bubbles).
33
Further Questions
As stated above, many factors were not taken into account because of the scope of this research. Future work could investigate the level of underpricing of internet IPOs after taking more factors into account.
This research has been done on IPOs issued in the United States. Future work could investigate differences between countries and/or continents.
34
List of references
Beatty, R. P., & Welch, I. (1996). Issuer expenses and legal liability in initial public offerings. JL & Econ., 39, 545.
Brau, J. C., & Fawcett, S. E. (2006). Initial public offerings: An analysis of theory and practice. The Journal of Finance, 61(1), 399-436.
Carter, R., & Manaster, S. (1990). Initial public offerings and underwriter reputation. The Journal of Finance, 45(4), 1045-1067.
Cooney, J., Singh, A., Carter, R., & Dark, F. (2001). IPO initial returns and
underwriter reputation: has the inverse relation flipped in the 1990s. Working Paper
(University of Kentucky).
Cornelli, F., & Goldreich, D. (2001). Bookbuilding and strategic allocation. The
Journal of Finance, 56(6), 2337-2369.
Demers, E., & Joos, P. (2007). IPO failure risk. Journal of Accounting
Research, 45(2), 333-371.
Derrien, F. (2005). IPO pricing in “HOT” market conditions: who leaves money on the table?. The Journal of Finance, 60(1), 487-521.
Hand, J. (2000). Profits, losses and the non-linear pricing of Internet stocks. Available
at SSRN 204875.
Johnson, J. M., & Miller, R. E. (1988). Investment banker prestige and the underpricing of initial public offerings. Financial Management, 19-29.
Ljungqvist, A., & Wilhelm, W. J. (2003). IPO pricing in the dot‐com bubble. The
Journal of Finance, 58(2), 723-752.
Ljungqvist, A., Nanda, V., & Singh, R. (2006). Hot Markets, Investor Sentiment, and IPO Pricing. The Journal of Business, 79(4), 1667-1702.
Loughran, T., & Ritter, J. R. (2002). Why don't issuers get upset about leaving money on the table in IPOs?. Review of Financial Studies, 15(2), 413-444.
Loughran, T., & Ritter, J. R. (2004). Why Has IPO Underpricing Changed Over Time?. Financial Management, 33(3).
Mason, C. H., & Perreault Jr, W. D. (1991). Collinearity, power, and interpretation of multiple regression analysis. Journal of Marketing Research, 268-280.
35
Ritter, J. R., & Welch, I. (2002). A review of IPO activity, pricing, and allocations. The Journal of Finance, 57(4), 1795-1828.
Ritter, J. (2011) IPO Underwriter Reputation Rankings (1980 - 2011) [Excel spreadsheet]. (Retrieved from http://bear.warrington.ufl.edu/ritter/FoundingDates.htm)
Ritter, J. (2012) A list of internet IPOs [Excel spreadsheet]. (Retrieved from http://bear.warrington.ufl.edu/ritter/ipodata.htm)
Trueman, B., Wong, M. F., & Zhang, X. J. (2000). The eyeballs have it: Searching for the value in Internet stocks. Journal of Accounting Research, 137-162.
Woo, S., Cowan, L.,Tam, P. (2011, May 20). LinkedIn IPO Soars, Feeding Web Boom. The Wall Street Journal. Retrieved from http://online.wsj.com
Worthen, B. (2012, April 19). Newest Tech IPO Doubles on Day One. The Wall Street
Journal. Retrieved from http://online.wsj.com
Corwin, S. A., & Schultz, P. (2005). The role of IPO underwriting syndicates: Pricing, information production, and underwriter competition. The Journal of Finance, 60(1), 443-486.
36 Tables
Table 1 (a, b & c) Underpricing of Internet IPOs and other IPOs, 1999-2012)
This sample includes all IPOs in the Thomson One, where founding date, IPO date and first-day return are available8. Unit offerings are excluded. First-day return is defined as % change in the stock price minus % change in the benchmark (S&P 400). IPOs are defined as Internet IPOs if they are included on “A List of Internet IPOs” (Ritter, 2012). T-statistics indicated with *** have a 1% significance
level, equal variance not assumed Table 1a Group Number of IPOs Mean First-day Return Standard deviation Non-internet IPOs 1148 0. 1642577 0.3762041 Internet IPOs 219 0.7990629 1.064546 Combined 1367 0.2659566 0.5949386 Difference -0.6348052
t-stat for difference -8.7214***
Table 1b Group Number of IPOs Mean First-day Return Standard deviation
Non-internet IPOs after
bubble 905 0. 095291 0.1778892
Internet IPOs after bubble 76 0.2199159 0.2673393
Combined 981 0.1049459 0.1891331
Difference - 0.1246249
t-stat for difference -3.9904***
37 Table 1c Group Number of IPOs Mean First-day Return Standard deviation
Non-internet IPOs during
bubble 243 0.4211089 0.6845153
Internet IPOs during bubble 143 1.106861 1.194621
Combined 386 0.6751571 0.9648049
Difference - 0.6857524
38
Table 2 (a, b & c) Average age of Internet IPOs and other IPOs, 1999-2012
This sample includes all IPOs in the Thomson One, where founding date, IPO date and first-day return are available9. Unit offerings are excluded. Age is defined years between founding date and IPO date.
IPOs are defined as Internet IPOs if they are included on “A List of Internet IPOs” (Ritter, 2012). T-statistics indicated with ***, **, * have a 1%, 5% and 10% significance level respectively.
Table 2a
Group Number of
IPOs Mean age
Standard deviation Non-internet IPOs 1148 14.6324 18.39565 Internet IPOs 219 7.063927 5.022475 Combined 1367 13.4199 17.20131 Difference 7.568477
t-stat for difference - 11.8206***
Table 2b Group Number of IPOs Mean First-day Return Standard deviation
Non-internet IPOs after bubble 904 15.06969 18.87749
Non-internet IPOs during bubble 243 13.03704 16.45212
Combined 1147 14.63906 18.4023
Difference 2.032653
t-stat for difference 1.5295
39 Table 2c Group Number of IPOs Mean First-day Return Standard deviation
Internet IPOs after bubble 76 9.25 5.86998
Internet IPOs during bubble 143 5.902098 4.075545
Combined 219 7.063927 5.022475
Difference 3.347902
40
Table 3 Construction of the sample of all IPOs, 1999-2012
This sample includes all IPOs in the Thomson One, where founding date, IPO date and first-day return are available10. Unit offerings are excluded. Age is defined years between founding date and IPO date. IPOs are defined as Internet IPOs if they are included on “A List of Internet IPOs” (Ritter, 2012).
T-statistics indicated with ***, **, * have a 1%, 5% and 10% significance level respectively.
Criterion Number of IPOs
List of all IPOs between 01-01-1999 and 01-04-2012
(Thomson ONE, January 2014) 3317
Firms without complete dataset excluded (footnote 4) - 1941
Firms without matching underwriter - 7
Complete list of IPOs 1369
Duplicates in complete list - 2
Complete list of IPOs 1367
List of internet IPOs between 01-02-1990 and 01-04-2012
(Ritter, 2012) 651
Firms before 01-01-1999 -99
Complete list of internet IPOs 552
Internet firms during bubble matched with total list of IPOs 143
Internet firms after bubble matched with total list of IPOs 76
Total number of internet IPOs 219
Total number of non-internet IPOs 1148
41
Table 4 Correlation matrix
This table shows the correlation between the variables defined in Chapter 5 Research Methodology
Netchange Internet
Dummy Bubbly Dummy Prest. Dummy
Internet dummy * Bubble Dummy Internet Dummy * Prest.Dummy LNage Netchange 1 Internet Dummy 0.3915 1 Bubble Dummy 0.4316 0.3595 1
Prest. Und. Dummy 0.1216 0.0693 0.0725 1
Internet dummy *
Bubble Dummy 0.4833 0.7826 0.5448 0.0632 1
Internet Dummy *
Prest. Dummy 0.4152 0.9086 0.3339 0.2084 0.7218 1
42
Table 5 Variance Inflation Factor
This table shows the result of the Variance Inflation Factor for the regression given in ‘Testing hypothesis 4’
Variable VIF
Internet Dummy 7.67
Internet Dummy * Prestigious Dummy 6.54
Internet Dummy * Bubble Dummy 3.27
Bubble Dummy 1.45
Prestigious Underwriter Dummy 1.15
43
Table 6 Descriptive statistics
Table 6 exhibits descriptive statistics for complete dataset
All IPOs from 1999 – 2012, containing 1367 observations (see ‘Table 3 Construction of the sample of
all IPOs, 1999-2012’)
Variable Mean Standard Dev. Min Max
Net change .2659566 .5949386 -.98013 5.68594
Internet firm .1602048 .3669301 0 1
Prestigious underwriter .784199 .4115274 0 1
Internet firm and Prestigious
underwriter .1360644 .3429824 0 1
Bubble .2825769 .4504173 0 1
Internet firms and Bubble .1046852 .3062595 0 1
LNage (Age in years) 2.05245 (14.6106) 1.172988 (18.2614) 0 (1) 4.65396 (105)
44
Table 7 Results from regressions
This table displays the result of the regression as modelled in ‘5.3 Regression model’. The coefficients and corresponding t-statistics are shown, apart from the last column (F statistic for model and corresponding P-value). The reported t-statistics are on the null-hypothesis that the variable has no effect on the dependent variable. T-statistics marked with *, ** or *** represent 10%, 5%, 1% respectively. The results are calculated using robust standard errors.
Regression Intercept Internet Dummy Prest. Dummy Internet Dummy * Prest.Dummy Bubble Bubble Dummy *
Internet Dummy LNage
F-statistic (Prob > F) R-squared 1 .184677 (8.50***) .6319014 (8.78***) - - - - -.0098794 (-1.16) 39.16 (0.0000) 0.1536 2 .1045456 (6.40***) - - - .3162238 (7.13***) .6858316 (6.32***) .000161 (0.02) 52.37 (0.0000) 0.2738 3 .1939937 (6.67**) - .1799152 (6.40***) - - - -.0342754 (-3.15***) 20.63 (0.0000) 0.0187 4 .0540203 (2.53**) -.2816003 (-2.74***) .057798 (2.75***) .4853532 (4.19***) .3228086 (7.31***) .546605 (4.95***) -.001335 (-0.18) 30.69 (0.0000) 0.2945
