1 Master Thesis
Finance
Did the determinants for the initial returns, and abnormal returns
at the end of the lock-up agreement change overtime?
By Thomas Holtslag
Abstract
This paper analyses whether the determinants between 1980 and 2000, still predict the initial returns and abnormal returns at the ending of the lock-up agreement for the period 2000 and 2017. I find average initial returns of 20.24% and abnormal returns for the three-day window of -1.17%, and seven-day window of -2.07%. I find that VC involvement, tech firms, hot issue markets are determinants for the initial returns, while VC involvement and the run-up are determinants for the abnormal returns. I conclude that the determinants from the past still do a good job in predicting the anomalies in the current time.
Student number: S3273660
Program: MSc Finance
Supervisor: Dr. A. de Ridder
Date: 12-06-2019
Email: t.j.holtslag@student.rug.nl
Field Key Words: IPO, abnormal return, initial return, lock-up agreement
JEL: G12, G14
2
1. Introduction
On March 2, 2017, Snap Inc. offered its shares for the first time to the market. Snap Inc. hired Morgan Stanley as the underwriter to manage the process in becoming a publicly traded firm through an initial public offering (IPO). The offer price of each share was $17, which was above the pricing range of $14-$16. When the first trading day ended, the price of the shares soared to $24.48, resulting in an initial return of 44%. The total proceeds of the offering were $3.4 billion. Six months later, the stock price fell by 5% surrounding the ending of the lock-up agreement, a lock-up agreement is an agreement that prohibits initial
shareholders to sell their shares in the period following the IPO. The case of Snap Inc. describes a situation that is common in the period following the IPO.
An IPO provides the initial shareholders of the firm with the opportunity to sell their stakes to the public for the first time. The initial shareholders are early investors in the firm, such as the founders, family, or venture capitalists (VCs). Several situations can be the reason why the initial shareholders decide to go public. Ritter and Welch (2002) argue that a firm could be in need of new capital for future growth, or it has an upcoming debt payment to one of the senior debt holders in the firm. Another reason could be that the initial
shareholders can cash out for the first time and get a value for their stake in the firm. Finally, a reason could be that the initial shareholders are no longer willing to bear the risk by
themselves and want to spread the risk over a diverse group of owners.
The process of an IPO is managed by an underwriter, which is hired by the issuer. The underwriter provides potential investors with a prospectus of the firm in which information about the firm is disclosed. The prospectus is mandatory according to the Securities Act of 19331 and contains information about the issuer, such as specific company information,
detailed ownership specification, and information about the offerings of the firm. The underwriter presents the prospectus to potential investors to get them interested in
subscribing to the IPO. The issuer and underwriter determine the offer price the day before the IPO. The offer price can be determined based on two ways. First, the company can fix the price with help of the underwriter, which is called the fixed price method. Second, the underwriters can determine the price based on the demand of the investors, which is called book building.
Subject to the prospectus of the issuer is the so-called lock-up agreement. Although the Securities and Exchange Commission (SEC) does not regulate such an agreement2, it
typically last for 180 days currently. A lock-up agreement is an agreement between the underwriter and the initial shareholders of the firm that prohibits insiders from offering their
1 SEC regulates prospectus: https://www.sec.gov/answers/about-lawsshtml.html
3 shares following the IPO. Specific information about the agreement is disclosed in the
prospectus of the firm. Bartlett (1995) argues that two factors are important in the existence of the lock-up agreement. First, it prevents a sudden increase in the supply of shares in the market. This would result in a surplus, which would influence the share price. Second, the underwriter wants to signal to potential investors that the initial shareholders will not cash out at the first moment in the case that they expect bad news after the IPO. This signals that the initial shareholders are confident about the future performance of the firm.
During the first 180 days after the IPO of Snap Inc., two anomalies occurred, which occur almost with every IPO with a lock-up agreement. The first anomaly dealt with the initial return on the first trading day. Ritter and Welch (2002) define the initial return as: ‘’the percentage change between the offer price and the closing price on the first trading day’’. The underwriter aims to price the offering such that the demand for the shares is high, which results in a higher profit for the underwriter. The second anomaly occurred
surrounding the ending of the lock-up agreement. During this period, the share price
showed different returns compared to the expected return in the market, hence the returns are abnormal. An explanation for this might be that investors anticipate the ending of the lock-up agreement, which increases the float of shares, resulting in a negative pressure on the stock price.
Two important studies that are the foundation of this paper are the studies by Loughran and Ritter (2004), and Field and Hanka (2001). First the study by Loughran and Ritter (2004), introduce several determinants for the initial returns like tech firms, venture capital (VC) involvement, and hot issue markets. This study tested multiple cycles between 1980 and 2000 and found that earlier mentioned determinants are of influence on the initial returns. Furthermore, this study finds initial returns of 18.9% over the whole period. The second study is the paper by Field and Hanka (2001) investigates a sample between 1988 and 1997 and found abnormal returns of -1.5% at the ending of the lock-up agreement for the three-day window starting one day prior to the ending of the agreement. The study test similar determinants for the abnormal returns occurring at the ending of the lock-up
4 The aim of this study is to test whether the determinants between 1980 and 2000 for the initial returns and abnormal returns at the end of the lock-up agreement, are still of influence between 2000 and 2017. The study will follow the models developed in Loughran and Ritter (2004), and Field and Hanka (2001) and therefore four cycles are included that measures the dotcom bubble in 2000, the period between the dotcom bubble, 2001 till 2007 and the financial crisis occurring in 2008 and 2009. The last cycle is after the financial crisis and last from 2010 until 2017. The full period is between January 1, 2000 and December 31, 2017. Specifically, this paper will examine the following research question:
‘’Are the determinants for the initial returns and abnormal returns at the ending of the lock-up agreement, between 1980 and 2000, still an influence, between 2000 and 2017?’’
To answer the research question, I examine a sample of 1,745 firms that conducted IPOs in the period between January 1, 2000 and 31 December 2017 in the US, on either the Nasdaq or the NYSE. For each IPO, deal and firm specific information is found using multiple databases. The contribution of this study is to test whether the determinants are still of influence in the current time. Furthermore, I will include a variable for firms operating in the bio pharma industry, because between 2000 and 2017 an increasing number of firms in this industry went public3, a similar trend was seen in the period before the dotcom bubble
where an increasing number of tech firms went public. A subsample analysis and OLS regression models for the initial returns and the cumulative average abnormal returns for the three-day window starting one day prior to the ending of the lock-up agreement, and the seven-day window starting five days prior to the ending of the lock-up agreement, are used in answering the research question.
The sample shows initial returns of 20.24% for the full period. During this period, it appears that firms operating in the technology industry, with a greater size, backed by VCs, with a higher ranked underwriter or listed on the Nasdaq show higher initial returns than their counterparts list. I find significant results that firms, operating in the tech, with VC involvement, with a large market capitalization, listed on the Nasdaq, and IPOs conducted in a hot issue market, are still good determinants for higher initial returns throughout all cycles.
Surrounding the ending of the lock-up agreement, I find significant abnormal returns of -2.07% for the seven-day window, and -1.17% for the three-day window. The abnormal returns are larger for technology firms, VC backed firms and firms listed on the Nasdaq
3 Information extracted from:
5 compared to their counterparts. The abnormal returns are not the result of a few outliers, because for both event windows roughly 55% are negative abnormal returns. Furthermore, for both windows the median is negative. The multivariate regression reports limited support for the determinants on the abnormal returns, and find that only VC backed firms, the run-up are significant determinants for the level of the abnormal returns.
The remainder of this paper is structured as follows. Section 2 gives insight into the existing theoretical framework of the study. Section 3 presents the descriptive statistics of the data and the methodology used in this study. Section 4 reports the findings. Section 5 concludes the research.
2. Previous research
2.1. Research on lock-up agreements
A wide range factors can make firms decide to go public. A firm can decide that it needs more capital for future investments and growth, or insiders might decide that an IPO is an exit strategy in which they can get a good market value for their shares. If the latter is the case, the underwriter and the issuer typically sign a lock-up agreement in which the initial shareholders agree to hold their shares for a pre-defined number of days. The underwriter and issuer determine the duration of the lock-up agreement. Typically, such a period lasts for 180 days starting from the first trading day. Although lock-up agreements are not a mandatory part of the process of an IPO, it is common that the parties involved in the process sign such a covenant as noted in Section 1. Bartlett (1995) argues that lock-up agreements exist to prevent an increase in the supply of shares entering public market at once. The most effective way in doing this, is signing an agreement with the initial
shareholders that prohibits them to sell their shares for a pre-defined period.
6 McQueen (2005) discuss how lock-up agreements can overcome the information asymmetry between initial and potential shareholders by committing to longer lock-up agreements.
Brav and Gompers (2003) introduce another explanation for the existence of lock-up agreements and argue that they serve as a commitment device. They argue that a lock-up agreement overcomes the moral hazard problem that might occur in the period after the IPO. In this period, new information through earning reports and media items enters the market that reveals the true quality of the firm. If the initial investors sign a lock-up
agreement, the probability that they will undertake actions that harm the firm and thus their own shares is less likely.
A study by Yung and Zender (2008) argues that information asymmetry and moral hazard can occur simultaneously. They argue that one of theories will be dominant over the other, depending on whether the issuer hired a top rank underwriter. Top rank underwriters overcome the information asymmetry, and thus the moral hazard will be the dominant theory. Conversely, information asymmetry will be the dominant factor if the issuer hired a low-ranked underwriter.
2.2. Research on initial returns
Initial returns on the first trading day is a topic that has attracted much attention among academics. In the period between 1970 and 2001, studies such as Ritter and Welch (2002), Rock (1986), McDonald and Fisher (1972), and Ritter (1982), all find initial returns for US firms on the first trading day varying between 11% and 49%. Several studies discuss theories and determinants in trying to explain the initial returns.
A study by Rock (1986) argues that information asymmetry is an important
determinant for the initial returns and introduces the “winner’s curse”. The theory assumes that informed investors have a better capability in predicting the future performance of a firm than uninformed investors. The result of this is that the informed investor only subscribes to successful offerings where the offer price is below the intrinsic value.
Uninformed investors, on the other hand, are not able to predict future performance, and therefore subscribe to all IPOs. Therefore, to remain interesting for uniformed investors, all IPOs are underpriced on average, resulting in higher initial returns.
7 explanation for the hot issue market by stating that the high initial returns is linked to high-risk offerings.
A study by Baron and Holmström (1980) discusses a form of information asymmetry by looking at the relationship between issuers and underwriters. They argue that
underwriters are better informed regarding the IPO process and that the issuers are unable to monitor the underwriters perfectly. Because finding potential investors is a costly process for the underwriter, he will not put much effort into finding investors unless the return is sufficient. The result of this behavior is that the offer price needs to be below the intrinsic value of the share. A study by Baron (1982) introduces a model describing the issuer and underwriter. The model describes a situation where the issuer gives the underwriter a set of contracts in which they try to optimize the issuers objective. The optimal contract is the one where the issuer gets the superior information from the underwriter. In exchange for this information, the issuer needs to compensate the underwriter. The underwriter determines the offer price based on the superior information and the underwriter shares in the gains of the IPO. Therefore, the underwriter prefers a contract consisting of a mix between the offer price of the IPO and the underwriter’s spread. In a case where the demand for the IPO is low, the spread will be smaller, and the reverse will be true in a case where demand is high. This concludes that greater the uncertainty around an IPO results in lower initial returns. Biais, Bossart, and Rochet (2002) enhance the model introduced by Baron (1982) in their study with informed investors. This leads to a new problem between investors and underwriters because the offering price will be higher when the allocation to informed investors is higher due to superior information. In a situation where the information the informed investors have turns out to be negative, the advantage will disappear and mitigate the winner’s curse.
Studies by Benveniste and Spindt (1989), and Sherman (2000) discuss another
explanation for the initial returns and notes that the book building process provides a signal. This study argues that underwriters get information from their large and informed clients about the interest they have in subscribing to an IPO. In return, the clients demand an attractive offering price. These studies conclude that an explanation for the initial returns is that underwriters prefer to distribute their shares to known clients.
8
2.3. Research about ending of lock-up agreement
At the ending of the lock-up agreement Field and Hanka (2001) find that abnormal returns of -1.5% appear for the three-day window. This study discusses several determinants for the abnormal returns. An important finding is that the abnormal returns for firms with VC involvement is three times larger than without this involvement. It appears that they sell their shares more aggressively in the period 1988 and 1997. Another study from Brav and Gompers (2003) between 1988 and 1996 found that abnormal returns of -0.86% appear for the three-day window, although less than in the study by Field and Hanka (2001) they are significant. Both studies discuss several determinants such as the run-up, starting from the first trading day until the day prior to the ending of the agreement, shares locked-up, and the exchange of the listing are of influence on the abnormal returns.
2.3.1. Downward sloping demand curve
A theory that gained much attention in explaining the abnormal returns is the downward sloping demand curve theory. This theory has its origin in a study by Scholes (1972) that predicts a decrease in share price due to a temporarily increased supply. The downward sloping demand curve theory is similar to the theory introduced by Scholes (1972) and predicts that the share price decreases surrounding the ending of the lock-up agreement. The reasoning behind this theory is that initial shareholders are entering the public market for the first time, something that the lock-up agreement prohibits in the period between the IPO and the ending of the agreement. This causes an increase in the supply of shares, which results in downward pressure on the share price. Ofek and Richardson (2000) discuss several studies such as Mikkelson and Partch (1985), and Keim and Madhaven (1996), which discuss whether demand curves even slope downward, but that is beyond the scope of this study and assume they do. Ofek and Richardson (2000) test in their study the downward sloping demand curve and find evidence that it exists, resulting in a price drop of -3% on the expiration day.
Field and Hanka (2001) test in their study whether firms with VCs as initial shareholders show higher negative abnormal returns than firms without such initial
shareholders. The reasoning behind this is that VCs aim to invest in firms in the early phase and exit the firm via an IPO. The exit strategy gives the VCs the opportunity to distribute the wealth gained from the IPO to their limited partners4. The expiration of the lock-up
agreement is the first opportunity to do this, resulting in a sudden increase of the supply of shares.
4 Limited partners are the investors that provide the venture capitalist with funding, which the venture
9 In the same study by Field and Hanka (2001), they test whether the exchange – either the Nasdaq or the NYSE – has explanatory power for the abnormal returns. The reasoning behind this argument is that the Nasdaq index consists of more technology and younger startup firms compared to the NYSE. These kinds of firms are difficult to value, resulting in a share price that is more volatile. The study further argues that there is a relation between the negative abnormal returns and the shares retained by the initial shareholders. They find that the number of shares locked up has a significant impact on the float of shares on the expiration day of the agreement. Moreover, the stake held by insiders after the IPO influences the abnormal returns on the lock-up expiration day.
2.3.2. Information asymmetry
A theory that gained much attention in the academic literature explaining the
abnormal returns surrounding the ending of a lock-up agreement is information asymmetry. Fama (1970) introduced the ‘’efficient market hypothesis (EMH)’’, meaning that share prices reflect all information available in the market at all times. Any violation of this hypothesis indicates an inefficient market with undervalued and overvalued stocks.
Brau, Lambson, and McQueen (2005) argue that the aim of lock-up agreements is to prevent initial shareholders from selling their shares for a pre-defined period, and as mentioned earlier as noted in Section 2.1., the typical lock-up period lasts 180 days. The researchers assume that initial shareholders have superior information compared to
potential investors, resulting in a violation of the EMH. Moreover, the two types of investors are not aligned. Although lock-up agreements exist to mitigate information asymmetries between initial and potential shareholders, practice shows that information asymmetries still appear. Brau, Lambson, and McQueen (2005) provide an explanation for this and argue that during the lock-up period, only a limited amount of information will become available. For example, during the 180 days, normally only two earning reports will be published.
Several studies discuss determinants for the asymmetric information problem between the two types of investors. A study by Lowry and Schwert (2002) argue that firms operating in the technology industry are more sensitive to information asymmetry
10 Altiase (1985) argues that the size of a firm determines the level of information asymmetry. Due to the fact the larger firms attract more analyst coverage, more information is available to enter the market. The result of this is that information asymmetry for larger firms is less common compared to smaller firms. Bradley et al. (2001) finds similar results for this relationship. This implies that larger firms show smaller abnormal returns.
A study by Yung and Zender (2008) argues that the underwriter that the issuer hires is a predictor for the level of information asymmetry. This study discuss that firms with top-ranked underwriters are able to overcome information asymmetry problem and thus show smaller abnormal returns.
2.4. Summary of previous research
Sections 2.1 discusses multiple theories and explanations for the anomalies
investigated in this study. Several studies argue that lock-up agreements typically last 180 days and exist because underwriters restrict initial shareholders from selling all their shares at the first possible moment –the IPO. They prevent initial shareholders from doing this because a sudden increase in the supply of shares results in a negative effect on the share price. Furthermore, initial shareholders signal by retaining their shares that they are confident on the future performance of the firm.
Section 2.2 discusses the theories explaining initial returns. An important explanation for initial returns is the existence of information asymmetry between investors that have superior information and investors without superior information. This results in IPOs with higher initial returns to attract both type of investors. Furthermore, the literature argues that the timing of the IPO – whether the IPO takes place in a hot or cold issue market – is of influence on the initial returns. A predictor for the initial return is the underwriter that that the issuer hires; if it is a top-ranked underwriter, the information asymmetry is less likely to appear.
11
2.5. Hypothesis development
The theoretical framework introduces several determinants for the initial and the abnormal returns at the end of the lock-up agreement. These findings are developed in the following hypotheses and are based on the studies by Loughran and Ritter (2004), and Field and Hanka (2001). The hypotheses will answer the research question whether the same determinants are still of influence in the period between 2000 and 2017.
Hypothesis 1: The same determinants between 1980 and 2000 influence the initial returns between 2000 and 2017.
Hypothesis 2: The same determinants between 1980 and 2000 influence the abnormal returns at the end of the lock-up agreement between 2000 and 2017.
3. Data and methods
3.1. Sample selection
The initial sample is created using two primary sources: the search engine Zephyr of Bureau van Dijk and the Thomson Reuters database. I include only firms in the sample that conducted IPOs between January 1, 2000 and December 31, 2017 on either the Nasdaq or the NYSE. This period measures different cycles between the chosen period. Furthermore, I follow the procedure in Field and Hanka (2001) for collecting data, and therefore IPOs with a price below $5.00 per share (penny stocks), non-common stock offers (REITS, ADRs, and unit offers), closed-end and mutual funds, and firms that reported earnings within three days before the expiration of the lock-up agreement are excluded. After this selection process, the final sample consists of 1,745 firms that went public between January 1, 2000 and December 31, 2017 in the US.
Several sources are used in extracting specific information about each IPO relevant for this study. The following deal information is found by consulting the Nasdaq 5 website:
offer price, closing price, lock-up expiration date, duration of the lock-up agreement, number of shares offered, and the exchange where the IPO took place. The Thomson Reuters database is the starting point for every firm and deals with specific information like total earnings, and total assets under management. The final book year prior to the IPO is chosen to prevent that information is not fully available. I manually collect information that was not available at the time. Stock returns are obtained using the in-built function in Thomson Reuters called Datastream. Information about insider ownership is found in each
12 prospectus handed in at the SEC, and can be found in the section on principal shareholders. Finally, information about the founding dates of the firms and company classification codes, which determine the industry firms are operating in, is found on the website of Jay Ritter6.
This website reports specific information about IPOs since 1980.
3.2. Descriptive statistics 3.2.1. Sample statistics
Table 1 presents an overview of the number of IPOs per year and the cycles, along with specific information about the number of observations, initial returns, and the money left on the table. The data show that in the years following the crises in 2000 and 2008– 2009, significantly fewer IPOs were conducted than in previous years. An explanation for this is that firms time their IPOs for a period with high market valuations. Furthermore, the years that show a decline in the number of IPOs are those that show smaller initial returns. The year 2000 show statistics that deviate significantly from the others. An explanation for this is that in 2000 the dotcom bubble was at its prime, resulting in overvaluations for almost all IPOs. The column money left on the table indicates a considerable outlier for 2008 compared to the other years. An explanation for this is the IPO of Visa Inc., which, at that time, was the largest offering ever conducted in the US, raising over $17 billion. Finally, the financial crisis occurred shortly after this offering, resulting in only 21 IPOs in that year that could
compensate for this outlier.
3.2.2. Summary of statistics about the firm, IPO, and lock-up agreements
Table 2 displays an overview of important characteristics of the firms that conducted IPOs based on the mean, median, standard deviation, minimum, and maximum. Additionally, the table reports the means of the different cycles in this study. The table is segmented in two panels.
Panel A reports general information about the issuing firms such as the ages, assets under management, market capitalizations, total incomes, VC involvement, and the
industries the firms are operating in. The difference between the founding date and the IPO date determines the age of the firm. The sample reports a mean age of 19.57 and a median age of 9. The study by Loughran and Ritter (2004) reports a median age of seven throughout the whole study over different cycles between 1980 and 2000. My data clearly shows that the mean age of the firms increased significantly after the dotcom bubble, with a mean age of 11.59 in 2000 and a mean age of 18.49 between 2010 and 2017. The oldest firm in my sample is 175 years. The data shows a wide dispersion of the sizes of the assets. A reason for
13 this might be that technology firms typically do not have many tangible assets under their management. An implication is that these kind of firms are difficult to value. Looking at the market capitalization of the firms, the median capitalization is $103 million. The largest firm in my sample has a capitalization of $81 billion. The firms in my sample have a mean of $3.91 million for their net income. My data show that VCs back 52% of the firms. This is similar to the data in the study by Field and Hanka (2001) that found VC backing in 48% of their sample between 1988 and 1997. The presence of VCs is still significant in the market. The cycles show that after the 2000s, the presence of VCs was declining. However, in the last cycle, VCs are involved in 51% of the IPOs. The sample reports that 37% of the firms are labeled as tech firms, which is similar to the 34% that Ritter and Welch (2002) find in their study. Analysis of the data indicates that after the dotcom bubble, a smaller percentage of the firms is labeled as tech firms. Finally, IPOs were listed on the Nasdaq 68% of the time. In the year 2000, the data shows that 92% of the firms was listed on the Nasdaq. The index of the Nasdaq
comprises mostly tech firms, which explains this.
Panel B of Table 2 reports general information about the IPOs such as the rank of the underwriter, the proceeds, the free float of shares, and number of shares offered.
14 Finally, panel B reports information about the lock-up agreements. In my sample lock-up agreements are part of 90.31% of the IPOs and last 180 days 86% of the time. The study by Field and Hanka (2001) reports that on average, 80% of IPOs between 1988 and 1997 have agreements that last for 180 days. Looking at the cycles, the results indicate that after the dotcom bubble, lock-up agreements were part of almost every IPO, typically with a length of 180 days. In this sample, the shortest lock-up period lasted 45 days and the longest lasted 540 days. Finally, the data shows that initial shareholders hold, on average, 38% of the shares after the IPO.
Table 1
Segmented overview IPOs per year
The total sample consist of 1745 firms who conducted an IPO between January 1, 2000 and December 31, 2017. Information about the opening price, closing price, initial return, and the amount of money left on the table is given and are equally weighted averages. The four cycles are presented below.
Year: Number of IPOS % total Opening Price ($) Closing price ($) Initial Return (%) Money left on the table ($ millions)
15 Table 2
Summary information of issuing firm, IPO, and lock-up agreements
Sample is 1,745 IPOs between January 1, 2000 and December 31, 2017 in the US, on either the Nasdaq or the NYSE. Mean, medians, standard deviations
and the minimum and maximum are given for the full sample. The mean of the cycles is given. Panel A provides general information about the issuing firm. Panel B reports specific information about the IPO, and lock-up agreement itself. Information about how the variables are measured is presented in parentheses.
Mean of Cycles: Variables:
Mean Median Standard
deviation Minimum Maximum 2000 2001-2007 2008-2009 2010-2017 Panel A: Issuing firm
information
Age (years) 19.02 9.00 26.35 0 175.00 13.31 21.10 27.75 18.49
Assets under management ($ millions)
1,384.06 112.00 8,903.70 0.70 255,567 1,638.56 856.30 997.18 1,621.60
Market capitalization after
($ millions) 975.911 103.40 43.06 20.51 81,247 749.45 719.56 1,587.55 1,181.68
Total income ($ millions) 3.91 -2.10 259.15 -3,426.00 8,988.00 183.75 421.87 793.81 747.10
Dummy variable for VC (%) 51.69 68.33 48.00 35.25 51.43
Dummy variable for tech (%) 37.53 67.41 36.00 29.41 28.00
Dummy variable for Nasdaq (%) 68.90 92.5 71.82 55.22 61.11
Panel B: IPO and lock-up agreement information
Mean Median Standard
deviation
Minimum Maximum
Prestigious underwriter 7.99 8.50 1.61 0 9.00 8.20 7.80 8.20 8.05
Proceeds ($ millions) 218.15 89.88 743.24 1.71 16,006 47.70 182.85 543.54 242.52
Free float (%) 28.06 35.00 25.00 0 100.00 22.98 30.00 32.36 26.91
Shares offered (millions) 13.94 8.84 23.86 0.17 478.00 9.42 11.06 21.47 13.81
Lock-up agreement (%) 90.31 42.08 96.21 98.41 99.16
180 days agreement (%) 94.88 94.05 94.69 93.54 97.22
Lock-up length (days) 181.56 180.00 30.46 45.00 540.00 183.66 182.12 178.52 181.13
16
3.3. Dependent variables
The aim of this study is to test whether the determinants of the initial returns and the abnormal returns at the end of the lock-up agreement still do a good job in predicting the two anomalies in a different time. Two dependent variables are tested to answer the research question. First, the initial returns of the IPOs. Second, the abnormal returns surrounding the lock-up agreement for the three-day window, starting one day before the end of the lock-up agreement, and the seven-day window, starting five days before the end of the lock-up agreement. The dependent variables are explained in the subsections that follow.
3.3.1 Initial return
The first dependent variable in this study is the initial return, based on the study by Loughran and Ritter (2004). The initial return is the return on the first trading day. For example, if a firm offers their shares for $20 and the stock closes at $30 on the first day, the initial return is 50%. To test whether differences occur during the cycles, five regression models are included to test the full period and the different cycles.
3.3.2 Abnormal return
For the second dependent variable, I follow the same procedure used in Field and Hanka (2001). The windows that will be measured are the three-day window and seven-day window. The windows will be calculated as the cumulative average abnormal return (CAAR). A more comprehensive explanation is given in the following.
The methodology in calculating the CAAR is a three-step procedure developed by Fama, Fisher, Jensen, and Roll (1969). First, the abnormal returns for the firms in the sample are calculated. The abnormal return, see Eq. (1), is derived by subtracting the expected return, based on the CRSP value weighted index (including the dividend), from the realized return of the firm:
𝐴𝑅𝑖,𝑡 = 𝑅𝑖,𝑡− 𝐸(𝑅𝑖,𝑡) (1)
in which AR stands for abnormal return for stock i at time t. Ri,t stands for the return of stock
stock i at time t. E(Ri,t) stands for the expected return for stock i at time t. The second step is
17 in which AARt represents the average abnormal return at time t, N represents the number of
firms, and ARi,t is the sum of the abnormal returns for the stocks at time t (e.g., t = -5
means five trading days before the lock-up expiration day at t = 0). The third step in this procedure is to calculate the CAAR, see Eq. (3). Fama, Fisher, Jensen, and Roll (1969) argue that the CAAR is a better statistical figure than the average abnormal return (AAR) if an aggregate effect is to be measured. The expectation is that the abnormal returns have already occurred in the days prior to the lock-up expiration day because investors anticipate the end of the lock-up agreement. The formula is as follows:
𝐶𝐴𝐴𝑅𝑡= ∑ 𝐴𝐴𝑅𝑡 𝑇 𝑡=1
(3)
in which the CAAR at time t is the sum of all average abnormal returns until time t.
3.4 Independent variables
The independent variables for answering the research question are based on determinants similar to those in the studies of Field and Hanka (2001), and Loughran and Ritter (2004). For the variables age, market capitalization, and run-up, the natural logarithm is used to deal with non-normality. See a description of the variables in Appendix A.
Loughran and Ritter (2004) include a variable for the ‘’age’’ of the company. They argue that the amount of information available for the firm increases linearly with the age of the firm, thus resulting in less information asymmetry. Measured as the number of years between the IPO and founding date between the founding date and the date of the IPO. Information about the founding dates is extracted from the website of Jay Ritter. For both variables, I expect smaller initial returns and abnormal returns the older the firm is.
Both studies insert a ‘’dummy variable for tech firms’’. The studies argue that tech firms show higher initial returns and abnormal returns because they are difficult to value due to their business model. This results in greater stock volatility. Tech firms are labeled as such based on the SIC codes from the website of Jay Ritter. I expect higher initial returns and abnormal returns for tech firms.
18 In a manner similar to both of the abovementioned studies, I include a ‘’dummy
variable for venture backing’’. Field and Hanka (2001) argue that firms with VC involvement
show abnormal returns three times higher than firms without VC involvement. A reason for this is that venture capitalists sell their shares at the end of the lock-up agreement and therefore supply increases at around that time. I expect higher initial returns and abnormal returns for firms with VC involvement.
Both studies follow Carter, Dark, and Singh (1998) by including a ‘’dummy variable for
prestigious underwriter’’. An underwriter with a rank of eight or better is seen as prestigious.
The literature discusses that fact that firms with a prestigious underwriter show mixed results for initial returns and abnormal returns. An explanation that is given for this is that underwriters sometimes prefer IPOs with higher initial returns because that increases their income, while other studies discuss the fact that prestigious underwriters overcome the information asymmetry resulting in smaller initial returns and abnormal returns. I expect that a firm that hires a prestigious underwriter will show smaller initial returns and abnormal returns.
Both studies include a variable for the size of the firm. The ‘’market capitalization’’ of the firm is measured as the issue price multiplied by shares outstanding following the IPO. The literature argues that larger firms show smaller initial returns and abnormal returns. This is due to the information that is available on larger firms Therefore, I expect that larger firms show smaller initial returns and abnormal returns.
The studies include a ‘’dummy variable for Nasdaq listing’’. The Nasdaq index
comprises many tech firms and it is therefore argued that, due to this composition, it shows higher initial returns and abnormal returns.
A ‘’dummy variable for hot issue market’’ is included to measure the effect of offerings during such periods. It is expected that IPOs in a hot issue market show higher initial returns. An explanation for this is that firms time their issue for a period prior to such a market.
In the regression model for the abnormal returns, similarly to the study of Field and Hanka (2001), I include a variable that measures the ‘’run-up’’. A run-up refers to the return of the stock that is different from the benchmark, starting at the IPO date until the day prior to the end of the lock-up agreement. I expect that a positive run-up results in higher
abnormal returns because investors are willing to sell their shares if they have received a positive return.
19 in higher initial returns and abnormal returns at the end of the lock-up agreement. This is measured by the percentage of the shares retained relative to the shares issued.
3.5 OLS regression models
The OLS regressions for the dependent variables are presented below. Several
variables are sensitive to skewness, and therefore are not normally distributed. The data will use White’s robust standard errors to overcome the non-normality. Furthermore, the data is winsorized at the 99th percentile to deal with extreme outliers.
The OLS regression for the initial returns is presented in Eq. (4):
𝐼𝑅 = 𝛽0+ 𝛽1∗ 𝑎𝑔𝑒𝑙𝑜𝑔+ 𝛽2∗ 𝑡𝑒𝑐ℎ𝑑𝑢𝑚𝑚𝑦+ 𝛽3∗ 𝑏𝑖𝑜𝑝ℎ𝑎𝑟𝑚𝑎𝑑𝑢𝑚𝑚𝑦+ 𝛽4∗ 𝑉𝐶𝑑𝑢𝑚𝑚𝑦+ 𝛽5∗ 𝑝𝑟𝑒𝑠𝑡𝑖𝑔𝑖𝑜𝑢𝑠 𝑢𝑛𝑑𝑒𝑟𝑤𝑟𝑖𝑡𝑒𝑟𝑑𝑢𝑚𝑚𝑦+ 𝛽6∗ 𝑆𝑖𝑧𝑒𝑓𝑙𝑜𝑎𝑡 + 𝛽7∗
𝑁𝐴𝑆𝐷𝐴𝑄𝑑𝑢𝑚𝑚𝑦+ 𝛽8∗ 𝐹𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑠ℎ𝑎𝑟𝑒𝑠 𝑙𝑜𝑐𝑘𝑒𝑑 − 𝑢𝑝𝑓𝑙𝑜𝑎𝑡+ 𝛽9∗ 𝐻𝑜𝑡 𝑖𝑠𝑠𝑢𝑒 𝑚𝑎𝑟𝑘𝑒𝑡𝑑𝑢𝑚𝑚𝑦 + 𝑒𝑖
(4)
The OLS regression on the three-day window and seven-day window is presented in Eq. (5), the CAAR is taken for the results:
𝐶𝐴𝐴𝑅𝑤𝑖𝑛𝑑𝑜𝑤 = 𝛽0+ 𝛽1∗ 𝑎𝑔𝑒𝑙𝑜𝑔+ 𝛽2∗ 𝑡𝑒𝑐ℎ𝑑𝑢𝑚𝑚𝑦+ 𝛽3∗ 𝑏𝑖𝑜𝑝ℎ𝑎𝑟𝑚𝑎𝑑𝑢𝑚𝑚𝑦+ 𝛽4∗ 𝑉𝐶𝑑𝑢𝑚𝑚𝑦+ 𝛽5∗ 𝑝𝑟𝑒𝑠𝑡𝑖𝑔𝑖𝑜𝑢𝑠 𝑢𝑛𝑑𝑒𝑟𝑤𝑟𝑖𝑡𝑒𝑟𝑑𝑢𝑚𝑚𝑦+ 𝛽6∗ 𝑆𝑖𝑧𝑒𝑓𝑙𝑜𝑎𝑡 + 𝛽7∗ 𝑁𝐴𝑆𝐷𝐴𝑄𝑑𝑢𝑚𝑚𝑦 + 𝛽8 ∗ 𝑟𝑢𝑛 − 𝑢𝑝𝑙𝑜𝑔+ 𝛽9∗ 𝐹𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑠ℎ𝑎𝑟𝑒𝑠 𝑙𝑜𝑐𝑘𝑒𝑑 − 𝑢𝑝𝑓𝑙𝑜𝑎𝑡+ 𝑒𝑖
20
4. Empirical results
4.1. Results initial returns 4.1.1. Analysis initial returns
Table 3 presents an analysis of the initial returns based on the categories of age, proceeds, market capitalization, income, industry, VC involvement, offer price range, stake insiders, prestigious underwriter, and the exchanges the firms are listed. Each category is split into subcategories in order to analyze the differences between them.
The data shows that the initial returns during the dotcom bubble outperformed those of all the other cycles. An explanation for this is that technology firms were overvalued during this period because of the difficulty in valuing their business models. Looking at the cycles of periods following the dotcom bubble, initial returns show an upward trend again but are nowhere near the initial returns seen during the 2000s.
The split in the age category is made between firms younger than eight years and firms older than eight years. The data shows that in the first two cycles, younger firms had higher initial returns than did older firms. In the last two periods this difference is reversed. Overall, younger firms show higher initial returns in my sample, which is in line with the findings in the literature, specifically those of Loughran and Ritter (2004).
The literature argues that firms with larger proceeds show smaller initial returns because more information is available, which results in a lower risk. The split is made based on the median value of the category. However, the data reports that firms with larger proceeds show higher initial returns.
The data in Table 4 shows that firms with a larger market capitalization show higher initial returns. As with the proceeds category, the median value determines whether the firm is labeled large or small. For all the cycles, firms with larger market capitalization show higher initial returns. This means that the pattern persists overtime.
An inspection of the income category shows that firms with a negative income report higher initial returns than do firms with a positive income. An explanation for this is that firms that report a negative income are often firms in the tech or bio pharma industry, which, as discussed above, show higher initial returns.
21 The category based on VC involvement shows that firms backed by a VCs show higher initial returns. On average, firms backed by VC show initial returns of 30.06%, while firms without such backing show initial returns of 16.12%.
Furthermore, the offer price category reports results that are expected based on the literature. Firms that are priced above their range show higher initial returns. The
percentage of IPOs in the sample that are priced within their range is 55%.
Inspection of the fraction of shares locked up shows that IPOs in which a larger stake is retained by insiders report higher initial returns. As mentioned above, the larger the stake retained by insiders, the more volatility there is concerning the share price. This results in higher initial returns.
Considering the underwriter rank category, Table 4 indicates that firms with a top-rank underwriter show higher initial returns than do firms with a low-top-rank underwriter. However, the 2000s have an influence on the results. Reviewing the other three periods, the results are similar for both of these types. The literature, on the other hand, produces mixed arguments as to whether having a prestigious underwriter shows higher initial returns. The results in Table 4 are in line with the argument that underwriters price IPOs such that they increase their own income.
Finally, the literature expects that firms listed on the Nasdaq will show higher initial returns. This is explained by the composition of the index, which mostly comprises
22 Table 3
Average initial returns categorized by age, proceeds, market capitalization, income, industry, VC involvement, offer price range, stake insiders, prestigious underwriter, and
exchange
Sample is 1,745 IPOs between January 1, 2000 and December 31, 2017 in the US, on either the
Nasdaq or the NYSE. Equally weighted averages of the initial returns and observations are reported for the four cycles and the full sample. Proceeds, market capitalization, and prestigious underwriter, are segmented based on the median value of the categories.
2000 2001 - 2007 2008 - 2009 2010 - 2017 2000-2017 Categories: Return N Return N Return N Return N Return N
Age (years) Young (<8) 58.58% 139 14.20% 222 14.20% 9 15.05% 283 25.33% 654 Old (8) 41.18% 100 13.95% 385 17.19% 53 19.82% 553 22.72% 1091 Proceeds Below 43.55% 156 11.83% 295 13.09% 31 16.16% 418 21.24% 866 Above 65.80% 84 16.13% 312 10.21% 31 20.28% 418 29.74% 879 Market capitalization Below 28.29% 114 9.98% 326 11.90% 31 14.06% 418 15.81% 855 Above 72.18% 126 18.75% 281 11.31% 31 22.42% 418 32.88% 890 Income Negative 56.59% 182 12.70% 275 16.09% 47 18.42% 496 25.68% 968 Positive 34.83% 58 15.15% 332 18.84% 15 17.77% 340 20.85% 777 Industry Tech 66.21% 157 18.08% 205 15.28% 17 23.31% 234 32.25% 613 Bio Pharma 22.33% 28 4.33% 70 4.48% 3 17.08% 209 12.55% 310 Other 23.62% 55 13.59% 332 16.77% 42 18.29% 393 17.29% 822 VC involvement Involvement 63.12% 164 16.88% 280 15.98% 20 21.21% 430 30.06% 894 No involvement 25.89% 76 11.61% 327 17.13% 42 14.91% 406 16.12% 851
Offer price range
23
4.1.2. OLS regressions on initial returns
Table 4 presents the results of the OLS regression with initial returns as the
dependent variable. The variables included are based on the previous research: log variables for age, and market capitalization, dummy variables for firms operating in the technology and bio pharma industries, those with VC involvement, those listed on the Nasdaq, those with a prestigious underwriter, and IPOs conducted in a hot issue market. Finally, a variable is included for the fraction of shares retained by the initial shareholders and those sold to the public. A regression model is included for the full sample and, for the four cycles
measured in this study, whether the determinants changed during the cycles. The results in Table 4 show patterns that are also observable in Table 3.
The log variable for age shows a significant negative relation to initial returns. This relation is in line with the literature, which expected a negative relation with initial returns the longer a firm has been operating. Looking at the different cycles, the results show that during the dotcom bubble, age had a positive relation to initial returns. The cycle after the dotcom bubble shows that older firms indeed show smaller initial returns. During the last cycle, the sign is reversed again. Although the cycles show mixed evidence, the full sample shows a significant negative relation between age and initial returns.
The dummy variable for tech firms shows the expected positive relation with initial returns at the highest significance level, for the full sample. During the dotcom cycle in 2000, the coefficient for tech firms was 12%, while the following cycle shows a coefficient of -11%. An explanation for this is that after the dotcom bubble, tech firms were valued critically and issuers and underwriters were no longer willing to undervalue the stock. However, the full sample shows a significant positive relation between tech firms and initial returns.
A similar relation to that for tech firms is expected for the dummy variable for bio pharma firms. However, the full sample shows a significant negative relation for bio pharma and initial returns. This relation is observable through all cycles. An implication of this is that bio pharma firms show smaller initial returns, which is remarkable due to the lack of
transparency in this kind of firm.
As mentioned above, venture capitalists want to signal to the market that they are capable of undertaken a successful IPO. Hence, they want to show higher initial returns. The full sample and cycles show the predicted positive relation between VC involvement and the initial returns. Exceptional levels were found during the dotcom bubble, where the
coefficient was almost 30% for VC involvement. One can conclude that they fully optimized the chance to increase their own personal wealth in this period.
24 show higher initial returns. During the dotcom bubble and the financial crisis of 2008–2009 in particular, the coefficient was more than 10%. However, only during the financial crisis were significant results reported.
The variable for market capitalization shows significant results at the highest level of significance. This relation is contrary to the expectation that higher market capitalization overcomes the information asymmetry problem. As with the other variables, the cycle for the dotcom bubble shows a coefficient of 17% for initial returns. The coefficient diminishes to 4% for all the subsequent cycles but remains significant.
The row that reports the dummy variable for the Nasdaq listing indicates that firms listed on this exchange show higher initial returns. An explanation for this is that the Nasdaq index comprises mostly tech firms. This relation is clearly seen in the dotcom cycle, when the coefficient for firms listed on the NASDAQ was 26%. Other cycles report a positive significant relation with the initial returns that vary in the 5–11% range. The full sample indicates a significant relation between the dummy variable for Nasdaq listing and initial returns.
The proportion that is locked up by insiders shows a small, significant positive
relation with initial returns. The literature discusses the fact that if a larger part of the shares is retained by initial shareholders, this causes greater volatility as regards the stock. For the cycles, I find the expected relation; however, only the last cycle, from 2010 and 2017, shows a significant relation.
Finally, the dummy variable for hot issue market reports the positive relation with initial returns. During the dotcom bubble the hot issue market reported a positive coefficient of 32%. This is logical because, during this cycle, the market valuations of almost every firm were high. The other cycles and the full sample report significant positive relations with initial returns.
25 Table 4
Multivariate regression on initial returns
Sample is 1,745 IPOs between January 1, 2000 and December 31, 2017 in the US, on either the Nasdaq or the NYSE. The dependent variables are the full
sample, and the four cycles. Log variables are taken for age, and market capitalization to deal with non-normality in the sample. Binary variables are
included for firms: operating in the tech or bio pharma industry, with VC involvement, prestigious underwriter, listed on the Nasdaq, and IPOs conducted in
the hot issue market. The fraction shares locked-up is the percentage shares retained compared with the offer shares to the public. Initial returns are
winsorized at the 99th percentile. P-values are presented in parentheses. Significant at 10%, 5% and 1% level is given with asterisks *, ** and ***.
Cycles: Sample Full (1) 2000 (2) 2001-2007 (3) 2008-2009 (4) 2010-2017 (5) Intercept -1.40*** (0.00) -3.74*** (0.00) -0.81*** (0.00) -0.90* (0.07) -0.85*** (0.00)
Log age (years) -0.02**
(0.01) (0.20) 0.05 -0.12* (0.07) (0.25) -0.03 (0.74) 0.01
Dummy variable for tech 0.07***
(0.00) (0.14) 0.12 (0.58) -0.11 (0.88) -0.00 (0.26) 0.03
Dummy variable for bio pharma -0.08***
(0.00) -0.19* (0.06) -0.12*** (0.00) -0.13* (0.06) (0.38) -0.03
Dummy variable for VC involvement 0.09***
(0.00) 0.29*** (0.00) 0.05** (0.01) (0.95) 0.01 0.07*** (0.00)
Dummy variable for prestigious underwriter 0.01
(0.61) (0.26) 0.11 (0.91) 0.01 0.15*** (0.00) (0.49) -0.02
Log market capitalization
(share price * shares post IPO) 0.07*** (0.00) 0.17*** (0.00) 0.04*** (0.00) (0.08) 0.04* 0.04*** (0.00)
Dummy variable for Nasdaq 0.12***
(0.00) (0.06) 0.26* 0.05** (0.01) 0.11** (0.02) 0.06*** (0.01)
Fraction shares locked-up 0.07**
(0.01) (0.85) 0.03 (0.17) 0.04 (0.80) 0.02 0.12*** (0.00)
Dummy variable for hot issue market 0.15***
(0.00) 0.32*** (0.00) 0.09*** (0.00) 0.16*** (0.00) 0.12*** (0.00)
26
4.2. Results abnormal returns 4.2.1. Initial analysis of the CAAR
Table 5 presents an overview of the CAARs surrounding the end of the lock-up agreement, from five days prior to the event until five days after it. On the day the lock-up agreement ends, the sample reports a significant abnormal return of -0.43%. The table clearly shows that the abnormal returns already occur five days prior to the end of the lock-up agreement. Shortly after the end of the agreement, for days two and three, the abnormal returns are positive. However, on the following days, negative abnormal returns are again reported. The window starting two days after the end of the agreement and ending on day ten reports a significant CAAR of -0.84% and a median of -0.26%. The seven-day window, which is one of the dependent variables in Section 4.2.3., reports a significant CAAR of -2.07% with a median of -1.46%. The other dependent variable is the three-day window, which shows a significant CAAR of -1.17% and a median of -0.69%. Considering now the column that reports negative abnormal returns during the windows, the results show that the CAARs are not dependent on a few outliers as all the windows report negative CAARs varying between 50% and 60%.
4.2.2. Analysis of CAAR for the subsamples
Table 6 presents an overview of various subsamples for the seven-day and three-day windows at the end of the lock-up agreement. The outcomes for these windows were discussed in the previous section. Similarly to the study by Field and Hanka (2001), the data shows that VC involvement results in abnormal returns of -3.12% for the seven-day window, which is almost three times higher than for firms without VC involvement. For the three-day window, this is almost 12 times higher. As expected, venture capitalists unload their shares at the end of the lock-up agreement. Furthermore, for the three-day window firm without VC involvement show positive abnormal returns. The data show that tech firms show
27 Table 5
CAARs surrounding the ending of lock-up agreement
Sample is 1,745 IPOs between January 1, 2000 and December 31, 2017 in the US, on either the
Nasdaq or the NYSE. The number of lock-up agreements is 1,576. CAARs are calculated using the returns of the firms compared with the CRSP value-weighted index. Multiple windows are presented, in which (t=-5) stands for 5 days prior to the ending of the lock-up agreement. A parametric t-test
calculates the significance of the means (H0: μ=0 against Ha: μ0) A non-parametric Wilcoxon
signed-ranked test is done to test the medians (H0: =0 against Ha: 0). The t-statistics are presented in the
parentheses. Significant at 10%, 5% and 1% level is given with asterisks *, ** and ***.
CAAR (%) Negative abnormal
returns:
Window: Mean Median
28 Table 6
CAAR for subsamples
The CAARs for multiple subsamples are presented below. The total sample exists of 1745
observations with 1,576 lock-up agreements. The event takes place at the lock-up expiration day of the lock-up agreement. The cumulative abnormal returns are calculated for the different event windows. The CAAR is the cumulative of all AAR in the event window. A parametric t-test is done for
the means (H0: =0 against Ha: 0). The t-statistics are presented in the parentheses. Significant at
10%, 5% and 1% level is given with asterisks *, ** and ***.
Subsample: Seven-day (%) CAAR: Negative CAARs: Three-day (%) CAAR: Negative CAARs:
Complete -2.07*** (-6.06) 54.00% -1.17*** (-3.47) 55.42% VC involvement -3.12*** (-6.73) 58.21% -1.20** (-2.39) 61.23% No VC involvement -0.33 (1.02) 54.45% 0.10*** (-4.85) 54.67% Tech sector -3.31*** (5.73) 61.88% -1.20*** (-2.39) 62.28% Bio pharma sector -1.35*
29
4.2.3. Multivariate regression results
Table 7 presents the results of the OLS regression with the CAAR for the seven-day and three-day windows as dependent variables as this study tests whether the same determinants influence the CAARs. I include the following variables in the model: log variables for age, run-up, and market capitalization; and dummy variables for firms operating in the technology and bio pharma industries, those with VC involvement, those listed on the Nasdaq, and those with a prestigious underwriter. Finally, a variable is included for the proportion of shares retained by the initial shareholders and those sold to the public. Two regression models are included to control whether the independent variables influence the seven-day and three-day windows separately. The model for the three-day window is consistent with Field and Hanka (2001); however, I include a regression for the seven-day window.
The log variable for age shows a negative relation to the abnormal returns for the seven-day window and a positive relation to those for the three-day window. The
coefficients for both terms are either -1% or 1% and thus show mixed results. However, neither model is significant.
The dummy variable for tech firms shows the expected relation with the CAARs for both event windows, a pattern which is also seen in Table 6. Firms that are labeled as tech firms show a negative coefficient of -1%. This is in line with the argument that tech firms are expected to experience larger information asymmetry, resulting in greater volatility in the stock. However, neither coefficient is significant.
In a manner similar to that for tech firms, the relation for the bio pharma coefficient shows the expected relation for the dummy variable for bio pharma firms. For both
windows, the models show a negative relation, -1%, with the coefficient. However, neither model is significant.
The dummy variable for VC involvement shows the expected relation with previous studies and therefore firms with VCs as initial shareholders show higher abnormal returns. The two models show coefficients of -4% for the seven-day window and -2% for the three-day window. Hence, I find evidence that VCs sell their shares more aggressively than firms without VC involvement do. This relation is also found in Table 6, where firms with VC involvement report abnormal returns ten times higher than those without. Although the effect diminishes slightly in the three-day window, it is still significant at the 10% level.
30 The variable for market capitalization shows mixed relations with the CAARs for both windows. Information about larger firms is more easily available, which results in less information asymmetry. For the seven-day window, the coefficient is -1%, while for the three-day window it is 1%. It could be that new information about the firm enters the market between the two windows, which would explain the relation. However, neither coefficient is significant.
The row that reports the dummy variable for the Nasdaq listing indicates that firms listed on this exchange show positive CAARs, which is contrary to expectations. An
explanation for this is that the Nasdaq index comprises mostly tech firms, which results in more information asymmetry. The coefficient is almost 1% for both models. As with earlier variables, however, neither model is significant.
The proportion of share that is locked up by insiders shows a small positive effect on the CAARs for both windows, with a coefficient of 1% for both models. The literature discusses the fact that if initial shareholders retain a larger part of the shares, this causes greater volatility in the stock, which negatively influences the CAARs. My findings indicate a contrary view as regards this variable. However, neither model is significant.
Finally, the variable for the run-up reports the expected relation to the abnormal returns. The findings show that IPOs that were followed by a positive run-up show a negative relation with CAARs for the seven-day and three-day windows. Hence, investors are willing to sell their shares when they have received a positive return in the period after the IPO and the end of the lock-up agreement. The row shows a coefficient of -1%, significant at the 5% level.
31 Table 7
Multivariate regression on CAAR
Sample is 1,745 IPOs between January 1, 2000 and December 31, 2017 in the US, on either the
Nasdaq or the NYSE. The dependent variables are seven-day and three-day cumulative average abnormal returns. Log variables are taken for age, and market capitalization, and run-up to deal with non-normality in the sample. Binary variables are included for firms: operating in the tech or bio
pharma industry, with VC involvement, prestigious underwriter, listed on the Nasdaq, and IPOs. The
fraction shares locked-up is the percentage shares retained compared with the offered shares to the
public. Initial returns are winsorized at the 99th percentile. P-values are presented in parentheses.
Significant at 10%, 5% and 1% level is given with asterisks *, ** and ***. Seven-day CAAR:
(1) Three-day CAAR: (2)
Intercept 0.06
(0.53) -0.20* (0.05)
Log age (years) -0.01
(0.98) (0.39) 0.01
Dummy variable for tech -0.01
(0.22)
-0.00 (0.93)
Dummy variable for bio pharma -0.01
(0.68) (0.70) -0.01
Dummy variable for VC involvement -0.04***
(0.00) -0.02* (0.06)
Dummy variable for prestigious
underwriter (0.27) -0.01 (0.60) -0.00
Log market capitalization (share price * shares post IPO)
-0.01 (0.11)
0.01 (0.11)
Dummy variable for Nasdaq 0.01
(0.51) (0.88) 0.00
Fraction shares locked-up 0.01
(0.49) (0.50) 0.01
Log run-up
(cumulative return from IPO to t=-1)
-0.01** (0.01)
-0.01** (0.02)
32
4.3. Robustness checks 4.3.1. Dotcom bubble
The data in the sample measures four cycles between 2000 and 2017. The year 2000 is known for the extreme valuations of technology stocks, which ultimately ended in the bursting of the dotcom bubble. As table 2 reports, the initial returns in 2000 were
significantly higher than the other years. Moreover, the data could influence the regression outcomes. To test whether the dotcom bubble was of influence on the results I used
different cycles in my study. As the tables in Section 4.1 and 4.2 show, the initial returns and abnormal returns appear in every cycle and report overall the same effect. Furthermore, the regression models for the initial returns show that the determinants on the initial returns report the same sign for the different cycles, although the coefficients are not as high compared to the year 2000. A regression with the year 2000 omitted showed no significant difference in the results. Overall, I conclude that the initial returns and abnormal returns indeed are higher in the year 2000, but all the cycles show high initial returns and negative abnormal returns which means that the result are not dependent on the year 2000.
4.4.2. Different benchmarks
33
5. Conclusion
Looking back at the case of Snap Inc., the two anomalies described during the first 180 days (the initial returns on the first trading day and the abnormal returns at the ending of the lock-up agreement) attracted intense academic attention. The studies by Loughran and Ritter (2004), Field, and Hanka (2001) came up with several determinants for the initial returns on the first trading day and the abnormal returns at the end of the lock-up
agreement for the period between 1980 and 2000. This study verifies whether the
determinants that were of influence on the initial returns and the abnormal returns between 1980 and 2000 are still of influence in the period between 2000 and 2017.
To do this, a sample of 1,745 IPOs in the US, on either the Nasdaq or the NYSE, is investigated. The period between 2000 and 2017 measures four cycles, including the dotcom bubble in 2000 and the financial crisis starting in 2008.
Looking at the first anomaly, my sample shows average initial returns of 20.24% on the first trading day, meaning that IPOs still are valued below their intrinsic value. Consistent with Loughran and Ritter (2004), a segmented analysis on the average initial returns shows that younger firms, firms operating in the tech industry, firms with VC involvement, and firms listed on the Nasdaq still show higher initial returns compared to their counterparts. Furthermore, I find significant results that firms operating in the tech, with VC involvement, a large market capitalization, listed on the Nasdaq, and IPOs conducted in a hot-issue market generally have higher initial returns throughout all cycles.
The second anomaly in this study is the abnormal returns occurring at the end of the lock-up agreement. The results indicate a significant CAAR of -2.07% for the seven-day window and a significant CAAR of -1.17% for the three-day window. A subsample analysis shows that the CAAR for firms with VC involvement is almost 10 times higher than that for firms without VC involvement. Looking at the cycles, significant negative abnormal returns appear throughout the different periods. Although the CAARs are diminishing, I find a CAAR in 2000 of -3.78% and a CAAR of -1.69% for the period starting in 2010 and ending in 2017. Various predictors are investigated in explaining the negative abnormal returns. The most compelling determinants in explaining the negative abnormal returns occurring at the seven-day and three-seven-day window are VC involvement and the positive run-up. Field and Hanka (2001) report similar findings in their study.
34 and the ending of the lock-up agreement. Therefore, whether abnormal returns exist after this period is not examined. Moreover, the long-run performance of IPOs is not measured.
An interesting topic for future research is to investigate whether new determinants are of influence on the initial and abnormal returns. Another topic is whether the level of initial and abnormal returns that is seen in the US is also seen in markets outside the US. Furthermore, the data shows that lock-up agreements last for 180 day more of a rule than an expectation. An interesting topic could be which determinants influence the
standardization of the lock-up agreement and why the 180 days length is chosen.
Section 1 introduced the following research question: are the determinants for the
initial and abnormal returns at the ending of the lock-up agreement found between 1980 and 2000 still influences between 2000 and 2017? Analyzing the findings in this study, the answer
35
References
Allen, F., Faulhaber, G., 1989. Signaling by underpricing in the IPO market. Journal of Financial Economics 23, 105-137.
Altiase, R., 1985. Pre-disclosure Information, Firm Capitalization and Security Price Behavior around Earnings Announcements. Journal of Accounting Research, 21-36.
Baron, D. P., Holmström, B., 1980. The investment-banking contract for new issues under asymmetric information: delegation and the incentive problem. Journal of Finance 35(5), 1115-1138.
Baron, David P., 1982. A-model of the demand for investment banking advising and distribution services for new issues. Journal of Finance 37, 955-976.
Bartlett, J., 1995. Equity finance: Venture capital, buyout, restructurings, and reorganizations. Aspen publishers Inc, New York.
Biais, B. Bossaert, P., Rochet, J.C., 2001. An optimal IPO mechanism. Review of Economic Studies 69, 117-146.
Benveniste, L.M., P.A. Spindt., 1984. How investment bankers determine the offer price and allocation of new issues. Journal of Financial Economics 24, 343-361.
Bradley, D. J., Jordan, B. D., Ha-Chin, Y., Roten, I. C., 2001. Venture capital and IPO lockup expiration: An empirical analysis. The Journal of Financial Research 24(4), 465-492.
Brau, J., Lambson, V., McQueen, G., 2005. Lockups revisited. The Journal of Financial and Quantitative Analysis 40(3), 519-530.
Brav, A., Gompers, P.A., 2003. The Role of Lockups in Initial Public Offerings. The review of Financial Studies 16(1), 1-29.
Carter, R.B., Dark, F.H. Singh, A.K., 1998, Underwriter Reputation, Initial Returns and Long Run Performance of IPO Stocks, The Journal of Finance 53(1), 285-311.
Cornelli, F., Goldreich, D. Ljungqvist, A., 2006. Investor sentiment and pre-IPO-markets. Journal of Finance 61(3), 1187-215.
Fama, Eugene F., 1970. Efficient Capital Markets: A Review of Theory and Empirical Work. The Journal of Finance 25 (2), 383-417.
Fama, E., Fisher, L., Jensen, M., Roll, R., 1969. The adjustment of stock prices to new information. International economic review, 10.
Field, L., Hanka, G., 2001. The expiration of IPO share lockups, Journal of Finance 56 (2), 471-500. Hoque, H., 2011. The choice and role of lockups in IPOs: Evidence from heterogeneous lockup agreements. Financial Markets, Institutions and Instruments, 20(5), 191-220.
