Post-IPO exits in the United States : evidence of insider trading prior to the expiration of lockup provisions

(1)

Master Thesis

Post-IPO Exits in the United States: Evidence of Insider Trading Prior to the Expiration of Lockup Provisions 

Marianna Sampol

Student Number: 11400889 Supervisor: Dr. J.K. Martin

Master in International Finance

University of Amsterdam, Amsterdam Business School

(2)

Abstract

Using a sample of 2,872 IPOs, I extend the Brav and Gompers (2003) analysis on why underwriters suspend the lockup periods of prepublic investors. More specifically I test (i) the effectiveness of lockup periods as a commitment device and (ii) the effect of the IPO cycle on these early releases. My results are in line with those of Brav and Gompers (2003). I find evidence for their commitment hypothesis in that early releases are more likely in firms that have experienced larger excess returns and which are backed by venture capitalists. On the other hand, I do not find evidence of underwriters’ quality influencing on early releases. Regarding the extensions of their approach I find that early releases occur even if the information asymmetries are not fully resolved, indicating limited effectiveness of suspendable lockup periods in alleviating moral hazard problems. Moreover, hot IPO markets negatively influence early releases by underwriters.

(3)

1

1. Introduction

When taking a company public, the original shareholders can decide to agree on “locking themselves up”. This lockup, which Van Twisk (2015) explains as a voluntary agreement between pre-IPO shareholders and the underwriters, prohibits the pre-IPO shareholders to sell their shares for a pre-specified time window after the IPO (Initial Public Offering). Typically, the lockup lasts around 180 days and covers most of the shares that are not sold in the IPO (Field and Hanka, 2001).

Brav and Gompers (2003) show that these lockup agreements exist as a way to alleviate the moral hazard problem associated with information asymmetries regarding the actions of management. They argue that the quality of a firm taken public can be observed ex-ante. However, managers might not act in the best interest of shareholders after the IPO has taken place. Therefore, firms in which the probability of managers taking advantage of shareholders is greater will need to have (longer) lockup periods1_{in order to make} shareholders feel comfortable of buying into the IPO. Hence, lockup agreements can be seen as a commitment device.

Since lockup agreements are not legally binding, they can be broken. In this case, underwriters allow insiders to sell their equity share before the lockup expiration date. Breaking a lockup agreement is commonly known as an early release. Brav and Gompers (2003) show that as information asymmetries are reduced, the probability of early releases goes up. This is in line with the role of lockup agreements serving as a commitment device. The reason is that less information asymmetry makes underwriters more willing to suspend the lockup agreement.

This thesis extends Brav and Gompers’ 2003 analysis on early releases during lockup periods in two ways.

1_{As time passes, more information about the firm’s future prospects will be made public} thereby reducing information asymmetries, which, in turn, alleviates moral hazard and the need for a commitment device.

(5)

3 Firstly, it looks at the effectiveness of lockup periods as a commitment device. While Brav and Gompers (2003) show a clear relationship between the probability of an early release and the several proxies for the information asymmetry between in-and-outsiders, they leave unanswered whether insiders actually manage to take advantage of shareholders when they receive permission to sell their own locked up equity shares. When insiders sell shares, they can do so because they might need liquidity or want to rebalance their portfolio. In addition, they might want to sell shares because they have an informational advantage over other market participants (insider information) and take advantage of this by selling their equity share in anticipation of a decrease in market prices. If information asymmetries are completely resolved, and therefore the lockup mechanism is effective, insiders would no longer have an information advantage and would not be able to generate an excess return2_{. In line with Brav and Gompers, I find that firms with less information} asymmetry are more likely to have an early release of lockups. I also find that after the early release, the market prices of the shares are outperformed significantly by an index. The excess returns made by insiders (namely cutting their losses by selling early) indicate insiders have an informational advantage over outsiders. This action suggests informational asymmetries are not fully resolved and moral hazard is not fully alleviated.

Secondly, this thesis controls for IPO cycle effects. Brav and Gompers (2003) analysis of early releases during the lockup period controls for year effects. The authors find a statistically significant negative trend in the probability of early releases. This result can be interpreted as early releases becoming less likely over time. Brav and Gompers, however, leave unanswered what drives this trend. At the same time, a well-documented phenomenon is the so-called ‘hot’ and ‘cold’ IPO markets. These markets make up a cyclical pattern in which both the volume and degree of underpricing of IPO’s vary over time. This cyclicality tends to be driven by investor’s sentiment (Servaes & Rajan, 2002). Investor’s sentiment can also affect the probability of an early release, since during a period with widespread optimism, the window of opportunity for opportunistic behaviour by insiders might be larger. I find that “hot” markets have a negative impact on early releases.

2_{Insiders can also get an exposure by using derivatives. In line with Brav and Gompers} (2003), derivatives is not accounted.

(6)

4 The rest of this thesis is structured as follows. Chapter 2 presents an overview of the relevant literature, chapter 3 explains the hypotheses and methodology used after which chapter 4 describes the data and the selection process of the sample. The results of the study are described in chapter 5, and chapter 6 is the conclusion.

(7)

5

2. Literature Review

This section will present and review the most relevant literature available regarding insider sales before the lockup agreement expiration. Firstly, it will explain Brav and Gompers (2003) study, lockup agreements, inside trades and IPO cycles. Secondly, it will clarify as firm’s IPO motivation, the Underpricing theory, IPOs’ long term underperformance and the role of the underwriter.

2.1 Brav and Gompers (2003)

Since this thesis is an extension of Brav and Gompers (2003), I devote a specific section of this chapter to elaborating on their findings.

In their 2003 paper titled, “The Role of Lockups in Initial Public Offerings”, Brav and Gompers examine why prepublic investors (insiders) agree to so called lockup agreements with their underwriters in which they restrict the sale of their equity share after a firm goes public. Brav and Gompers test 3 hypotheses: (i) lockups as a signal of quality, (ii) as a commitment device to alleviate moral hazard problems, and (iii) as means for investment banks to extract additional compensation from issuing firms. Their analysis finds support for the abovementioned commitment hypothesis such that the length of the lockup period is longer when moral hazard in the aftermarket is higher. Moreover, they test one particular prediction of the commitment hypothesis on the suspension of the lockup period. In their paper, they state that “because the lead underwriter has the ability to release locked-up

shares early, we examine insider equity sales prior to the expiration of the lockup. Under the commitment hypothesis, only insiders of firms that have greatly diminished moral hazard risk should be allowed to sell equity prior to the expiration of the lockup.’’

Moral hazard risk can be reduced through alleviating the information asymmetry between inside and outside investors. Brav and Gompers use a series of proxies to estimate the degree of information asymmetry. In their paper, they describe their findings and the intuition behind the various proxies:

“As predicted by the commitment hypothesis, firms that have reduced asymmetry problems

are more likely to have early insider sales. The abnormal return3_{over the preceding 30 day}

(8)

6

period is positively related to the probability of early sales. Firms with higher returns are likely to have received a series of good news events and investors would be less concerned about management “cashing out”. Similarly, venture capital backing and larger firm size are all related a greater probability of early lockup release. Large firms likely have more information available about the firm, and venture capital-financed firms have the reputation of their venture capital investors to certify the firms’ quality. Finally, firms with a greater fraction of their post-IPO insider shares locked up are less likely to have insiders selling shares prior to the lockup expiration, consistent with the greater need for insiders in these firms to commit to not selling equity”

Brav and Gompers also investigate changes in the market price at the time of the lockup

expiration. They argue that because “the parameters of the lockup are well specified (in terms of length and number of shares locked) and known at the time of the IPO, if markets perfectly anticipated the release there should not be an abnormal price reaction at the time of the expiration. Indeed, even if demand curves for stocks slope downward, investors should correctly forecast on average the number of shares that insiders will sell at lockup expiration, and hence the average abnormal return should be zero’’. However, contrary to this

argument, they find a significant drop of 2% around the lockup expiration. Their evidence is consistent with downward sloping demand curves and costly arbitrage. In addition, they examine cross-sectional differences in the magnitude of the price decline and find that these are also consistent with the predictions of the commitment hypothesis.

2.2 Lockup Agreements

The lockup period is an agreement between pre-IPO shareholders and the underwriters, under which corporate insiders are not allowed to sell their shares for a specified period after the IPO, which is on average 180 days (Van Twisk (2015), Field and Hanka (2001), Bradly et al (2001)). As Bartlet (1995) says, “it prevents a surplus of stock hitting the market all at once”.

Field and Hanka (2001) report a trend towards this 180-day lockup agreement in their sample, which consists of IPOs from 1988 through 1997 in the US. Once the lockup period is expired there is an effect surrounding the lockup expiration date. They find that the price

(9)

7 movement after the expiration date is consistently worse than expected, violating the market efficiency theory (Fama, 1970). Moreover, this movement can be even bigger if it is a venture-backed IPO, since venture capitalists are more aggressive. Bradley et al (2001) also confirmed this hypothesis.

According to Field and Hanka (2001), the lockup terms and conditions are disclosed in the IPO prospectus, including the unlock date. The lockups serve several purposes: they secure the market by ensuring that key employees will not leave the company at least for the next months; they signal that insiders are not attempting to cash out in advance of bad news; and they may aid the underwriter’s price support efforts by temporarily constraining the supply of shares.

Brav and Gompers (2003) also find that the lockups serve as the commitment device to alleviate moral hazard problems. Firms with moral hazard issues will have even longer lockup periods, which is denoted by the Commitment Hypothesis.

Furthermore, Brav and Gompers conclude that insiders sell prior to lockup expiration, with a waiver from the underwriter, in firms that are associated with less moral hazard. These are larger firms, firms with higher turnover, firms backed by venture capitalists, and firms with higher abnormal returns in the preceding 30-day period. The period analysed by Brav and Gompers (2003) is between 1988 and 1996.

Still, it does occur that a single company has more than one lockup expiration date. In that case, the insider shares are gradually released. This is known as a staggered lockup agreement. Hoque (2011) finds that single lockups are associated with a higher level of information asymmetry, which is also associated with a greater price drop on the unlock date. This theory is consistent with the Commitment Hypothesis of Brav and Gompers (2003). Hence, firms with higher information asymmetry are likely to adopt staggered lockups in order to avoid the price movement on the lockup expiration date.

2.3 Inside Trades

(10)

8 the expiration of the lockup, which contradicts the Commitment Hypothesis. In theory, the agreement is not supposed to be “broken”, but underwriters waive the agreement for some of the insiders.

According to the SEC (Securities and Exchange Commission), corporate insiders are classified as corporate officers, directors, employees who have company shares (stock options) and any beneficial owners of more than 10% of a company’s voting shares. Those insiders need to file forms 144, 3,4 and 5 with the SEC.

Form 3 - Initial Statement of Beneficial Ownership for all Officers: An insider must file it to initially disclose his or her ownership of the company’s securities. It must be filed within 10 days after the person becomes an insider (SEC and Wharton WRDS).

Form 4 - Change in an Insider's Ownership Position: This form is required for a purchase, sale, option grant, option exercise, gift, or any other transaction that causes a change in ownership position. It must be filed within two business days following the transaction date. Transactions in a company’s common stock as well as derivative securities--such as options, warrants, and convertible securities--are reported on the form (SEC and Wharton WRDS).

Form 5 - Annual Statement of Change in Beneficial Ownership: This form covers activity for exempt transactions not required on Form 4. Exempt transactions may include small transactions or small transfers within company plans. It usually must be filed with the SEC no later than 45 days after the company’s fiscal year ends and is required from an insider only when at least one transaction, because of an exemption or failure to earlier report, was not reported during the year (SEC and Wharton WRDS).

Form 144 - The Sale of Restricted Securities: This form is for stock purchased in a private placement directly from an issuer before the company goes public. This type of sale is governed by SEC Rule 144. Rule 144 allows for the sale of restricted securities in limited quantities in the aftermarket. Specifically, a person who has beneficially owned shares of common stock for at least one year is entitled to sell, within any three-month period, a number of shares that do not exceed the greater of 1% of the number of shares of common

(11)

9 stock then outstanding or the average weekly trading volume during the four calendar weeks preceding the filing of a notice on Form 144 with respect to the sale (Brav and Gompers, 2003).

2.4 IPO cycle

Since 1970 there have been 22.027 IPOs in the United States. In the chart below we can clearly observe that the volume of IPO’s varies over time. This phenomenon is known as the IPOs cycle.

Figure 1. Number of Initial Public Offerings (1970 – 2017)

Source: Thomson One (From 01/01/1970 until 03/05/2017)

Ibbotson and Jaffe (1975) show that there are cycles in the number of IPO’s and the average initial return (underpricing). More specifically they show the former follows the latter. They divide IPO cycles into hot and cold by classifying the market as hot if the IPO underpricing is high and correspondingly as cold if IPO underpricing is low. Lowry and Schwert (2002) also find there to be a positive relationship between initial returns and future IPO volume.

0 200 400 600 800 1000 1200 19 70 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 Number of IPOs

(12)

10 Moreover, they find that this relationship is predominantly driven by information learned by investment bankers during the so-called registration period4_{. They state that “the}

information about the value of an IPO firm that becomes available during the registration period has an effect on the prices and offering decisions for other firms. Since the book-building period averages two months, but often lasts as long as four months, IPOs in subsequent months have overlapping registration periods. Investment bankers’ learning processes throughout this registration period cause monthly aggregate initial returns to be autocorrelated and to be positively related to future levels of IPO activity.”

As to what drives the IPO cycle, Rajan and Servaes (2002) find investor sentiment and the propensity of investors to chase trends explains (amongst other things) the characteristics of the IPO cycle. Lowry (2003) also finds investor sentiment is one of the drivers of variation in IPO volume.

2.5 IPO motivations

An IPO is often an exit for private equity funds that want to take advantage of the “window of opportunity” created for insiders to obtain their returns from the investment made. Pagano, Panetta & Zingales (1998) discussed that Companies appear to go public not to finance future investments and growth, but to rebalance leverage and investments after a long period of high investment and growth. IPOs are also followed by a lower cost of borrowing, as a sign of transparency, and consequently a reduction of cost of capital (Weighted Average Cost of Capital – WACC).

𝑊𝐴𝐶𝐶 = 𝐸𝑞𝑢𝑖𝑡𝑦

𝐷𝑒𝑏𝑡 + 𝐸𝑞𝑢𝑖𝑡𝑦∗ 𝐶𝑜𝑠𝑡 𝑜𝑓 𝐸𝑞𝑢𝑖𝑡𝑦 +

𝐷𝑒𝑏𝑡

𝐷𝑒𝑏𝑡 + 𝐸𝑞𝑢𝑖𝑡𝑦∗ 𝐶𝑜𝑠𝑡 𝑜𝑓 𝐸𝑞𝑢𝑖𝑡𝑦 ∗ (1 − 𝑇𝑎𝑥 𝑅𝑎𝑡𝑒)

Furthermore, firms are motivated by getting better credibility, since the publicity of listing and the prestige of being listed improve their reputation, all of which attract and facilitate obtaining the right competence (Brau & Fawcett, 2006 and Pagano, Panetta & Zingales, 1998). Brau & Fawcett (2006) add that the primary motivation for going public is to facilitate acquisitions. However, Brau & Fawcett (2006) as well as Pagano, Panetta & Zingales (1998)

4_{The registration period is the time between filing for an IPO with the SEC and the offering} at which trade starts.

(13)

11 claim that the main reason for remaining private is to preserve decision-making and ownership.

Rydqvist and Högholm (1995) have found that according to the firms, the reasons to go public can be divided into two categories: Financial and Productivity related.

2.6 Underpricing Theory

Underpricing is a phenomenon that usually happens after the IPO, on the stock’s first day of trade. Sophisticated investors who are not insiders and who have bought shares in the IPO are allowed to sell their stake on the first day of trade. The literature argues that there is a significant spike on the first day of trade due to the low offer price in the IPO prospectus. Brau & Fawcett (2006) argue that on average IPOs are priced lower than their first-day market closing price. (Between the period of 1960 and 2003, IPOs have averaged 18% underpricing)5_{. This phenomenon is known in the financial market as Underpricing. The} CFOs are relatively well informed regarding the expected level of underpricing and they feel that it exists primarily to compensate investors for taking the risk of investing in the IPO. Furthermore, the authors indicate that the second-most important reason for underpricing is the desire of underwriters to obtain the favour of institutional investors.

These are the three best-known theories that explain Underpricing: a) The Asymmetric Information Hypothesis;

b) The Signalling Theory; c) The Impresario Hypothesis;

a) The Asymmetric Information Hypothesis:

Rock (1986) classifies two types of investors: Informed and Uninformed. The insiders are the ones who have access to privileged information, while the new investors are the ones uninformed. Hence, in order to reward the uninformed investors for the higher risk they are taking, the issuing company should offer its shares with a discount. Otherwise, uninformed investors would not have the incentive to invest,

(14)

12 since they do not know if it is a good issue or if they are investing in overvalued stocks. New investors would be victims of the Winner’s Curse by investing in overvalued assets and would not get positive returns.

Baron (1982) discusses information asymmetry between underwriters and issuers. Bankers are on average better informed about capital markets than issuers, and they set an offering price based on its superior information. The issuer must compensate the banker for the use of his information. On the other hand, if both parties had the same level of information about the capital markets, the offer price will be at first-best level.

Carter and Manaster (1990) follow but extend Rock’s theory (1986) by suggesting that the greater the proportion of informed capital participating in an IPO is, the greater the equilibrium price run-up6_{. Good quality firms will signal by selecting a} prestigious underwriter that their risk is lower and that there is less asymmetry information. Prestigious underwriters will also pick good firms because they want to keep their reputation for future IPOs.

b) The Signalling Theory:

Myers and Majluf (1984) assume that the information asymmetry is a fact and managers know more about a firm’s value than new investors. Given information asymmetry, they suggest that the issuance of public shares might signal a last resort to raise capital, since firms would prefer debt, which is a cheaper cost of capital than equity, if they need external funds.

Allen and Faulhaber (1989) add to the theory that good firms would like to signal to investors their high value and they can signal that by setting a low IPO price. By doing that, high value firms expect to get a good pricing on subsequent underwritings. Good firms find it worthwhile to underprice their IPOs because by doing so, they condition investors to more favourably interpret subsequent dividend results. The owners of bad firms also know their expected performance and

6_{The period defined by Carter and Manaster (1990) for the price run up is a two-week} window, including the first day of underpricing.

(15)

13 subsequent market valuation. They know they cannot recoup the initial loss from underpricing, and so they cannot afford to signal. Alternatively, a bad firm can signal a strategic move to acquire a competent competitor to increase reputation (Brau & Fawcett, 2006; Pagano, Panetta & Zingales, 1998).

c) The Impresario Hypothesis:

Combining the work of Carter and Manaster (1990) with that of Allen and Faulhaber (1989), Shiller (1990) defines the Impresario Hypothesis by explaining that underwriters set low prices during IPOs on purpose to produce a high return on the very first day of trade, as compensation to investors for entering the IPO. The high initial return helps investors maintain appearances, which creates an increased demand for subsequent issuances. Additionally, the underwriter’s reputation is at risk if the offering turns out to be unsuccessful, due to its commitment to certify that IPO shares are not overpriced. Otherwise, they would lose market share in the IPO market (Derrien, 2005 p.489-490).

Another interesting finding about Underpricing is that, according to Francis and Hasan (2001), on average, the initial day returns of venture-backed IPOs are higher than non-venture-backed ones. They attribute such phenomena to the greater pre-market deliberate underpricing that characterizes IPOs brought to the market by venture capitalists7_.

2.7 IPO Long Term Performance

Once the IPO is concluded and the stock trade has started, Draho (2004) argues that there is academic evidence proving that going public when the shares are overvalued implies that the issuer will have poor post-IPO stock returns, as investors adjust the price to the appropriate level. Time-varying underperformance further supports the notion that

7_{These are managing partners or associates who work for a venture capital fund and seek a} high rate of return by investing in promising startups or young businesses that have a high growth potential but are also risky. Venture capitalists usually provide equity financing, funding in exchange for an equity stake in the company. Startups are non-established businesses that are very often unable to secure loans within financial institutions due to insufficient cash flows, lack of collateral or high risk profile. Thus, equity funding is their last resort for obtaining investments.

(16)

14 investors’ sentiment affects IPO timing. Firms that go public during a period of heavy IPO volume generate the worst long-run abnormal returns, whereas IPOs that occur in cold markets show no sign of underperformance (Loughran and Ritter, 1995; 2000; Helweg and Liang, 2004).

According to Draho (2004), positive “sentiments could result in overvaluation, thereby

inducing more firms to go public. Poor long-run returns would follow as the market corrects the initial pricing error. Fewer firms would find it attractive to go public when valuations are in line with fundamentals. Those firms that do are properly priced and produce normal returns” (Draho, 2004, p.17).

Loughran et al. (1994) found that in the long-run, the lower the returns the bigger the risk of firms going public and the higher the level of the market at the time of going public the “hotter” the capital market condition. The literature tends to view hot markets as the result of wild bullishness on the part of irrational investors (Loughran and Ritter (1995), Lerner (1994)), and hence as an opportunistic move for managers to take advantage of a “window of opportunity” to go public.

2.8 The role of the underwriter

Berk & DeMarzzo 2011 explain that ‘’after deciding to go public, managers of the company

work with an underwriter, an investment banking firm that manages the offering and designs its structure. More commonly, an underwriter and an issuing firm agree to a firm commitment IPO, in which the underwriter guarantees that it will sell all of the stock at the offer price. The underwriter purchases the entire issue (at a slightly lower price than the offer price, in the book building process) and then resells it at the offer price. If the entire issue does not sell out, the underwriter is on.’’

Berk & DeMarzzo 2011 also defined the IPO mechanics. Larger offerings are managed by a group of underwriters. The lead underwriter is the primary banking firm responsible for managing the deal. The lead underwriter provides most of the advice and arranges for a group of other underwriters, called the syndicate, to help market and sell the issue. Underwriters market the IPO, and they work together with the issuing company to help with

(17)

15 all the necessary filings with the SEC, such as the prospectus. Before the offer price is set, the underwriters work closely with the company to come up with a price range that they believe provides a reasonable valuation for the firm. Once an initial price range is established, the underwriters try to determine what the market thinks of the valuation. They begin by arranging a road show, in which senior management and the lead underwriters travel around the world promoting the company and explaining their rationale for the offer price to the underwriters’ largest customers—mainly institutional investors such as mutual funds and pension funds.

Once investors decide to purchase the stocks in the IPO, the underwriter starts the book building process by pricing the deal and managing its risks. Underwriters appear to use the information they acquire during the book-building stage to intentionally underprice the IPO, thereby reducing their exposure to losses. Furthermore, once the issue price is set, underwriters may invoke another mechanism to protect themselves against a loss—the overallotment allocation. This option allows the underwriter to issue more stock, amounting to 15% of the original offer size, at the IPO offer price. If the issue is a success, the Underwriter exercises the option, covering its short position. If the issue is not a success and its price falls, the underwriter covers the short position by repurchasing the overallotment in the aftermarket, thereby supporting the price.

Once the IPO process is complete, the company’s shares trade publicly on an exchange. The lead underwriter usually makes a market in the stock and assigns an analyst to cover it. By doing so, the underwriter increases the liquidity of the stock in the secondary market. A liquid market ensures that investors who purchased shares via the IPO are able to easily trade those shares. If the stock is actively traded, the issuer will have continued access to the equity markets in the event that the company decides to issue more shares in a new offering.

Complementing Berk & DeMarzzo 2011, and as mentioned in the Underpricing section, Carter and Manaster (1990) argue that prestigious underwriters, in order to maintain their reputation, market only IPOs of low risk firms, which, in turn, select only prestigious investment bankers. By using this strategy, underwriters signal their reputation to the market. They find a negative relation between the underwriter reputation and the level of

(18)

16 the IPO underpricing. Therefore, reputable underwriters are associated with IPOs that suffer from a lower level of information asymmetry.

Yung and Zender (2010) find that firms that select reputable underwriter are more likely to see moral hazards, such as commitment, as an issue that motivates the use of a lockup period. On the other hand, for firms that do not select a prestigious underwriter, it is more likely that asymmetric information, such as signalling, motivates the use of lockup periods. Hence, they suggest that the role of lockup agreements is determined by the reputation of the underwriter.

(19)

17

3. Hypotheses and Methodology

This chapter is structured as follows: I first present the hypotheses used. Secondly, I explain the analysis as conducted by Brav and Gompers (2003) and how this thesis extends their analysis.

3.1 Hypotheses

In this thesis, I extend Brav and Gompers’ analysis (2003) in two ways: firstly, by evaluating the effectiveness of lockup periods as a commitment device and secondly, by controlling for business cycle effects.

Effectiveness of lockup as a commitment device

Brav and Gompers (2003) show that the probability of early releases decreases as the information asymmetry between in-and-outside investors goes down. This is in line with the commitment hypothesis as a rationale for lockups, which suggests lockups mitigate the moral hazard problem associated with IPO’s. That is, if outside investors become more knowledgeable about the firm, they are less worried that managers try to cash out at their expense, thereby making it more likely for underwriters to suspend the lockup period. Therefore, early releases might happen if insiders have a sudden need for liquidity or a rebalancing of their personal portfolio but not if they want to take advantage of privileged information. After all, market participants would anticipate the change in market price by trading on it as soon as the information becomes available thereby, eliminating the information arbitrage opportunity for insiders. Therefore, if lockup periods are only suspended if information asymmetries are fully resolved, such that lockups are truly effective ways to prevent moral hazard, insiders will not be able to profit from insider information.

𝐻₀: 𝑖𝑛𝑠𝑖𝑑𝑒𝑟𝑠 𝑑𝑜 𝑛𝑜𝑡 𝑚𝑎𝑛𝑎𝑔𝑒 𝑡𝑜 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒 𝑒𝑥𝑐𝑒𝑠𝑠 𝑟𝑒𝑡𝑢𝑟𝑛𝑠 𝑡ℎ𝑟𝑜𝑢𝑔ℎ 𝑒𝑎𝑟𝑙𝑦 𝑟𝑒𝑙𝑒𝑎𝑠𝑒𝑠

In the case of early releases, insiders reduce their exposure by selling their equity share. As such, an excess return would mean a drop in the price on secondary markets from which they cut their losses.

(20)

18

Controlling for the IPO cycle

In their analysis Brav and Gompers (2003) allow for year effects and show there is a trend in the probability of early releases from the lockup period. The authors do not elaborate on what drives this time-trend. As mentioned in the literature review, IPO markets tend to be cyclical, such that the level of underpricing and IPO volume fluctuates. Considering that this cyclicality is, in part, driven by investor’s sentiment, I investigate whether the IPO cycle is an omitted variable in the analysis by Brav and Gompers (2003).

𝐻₀: 𝑡ℎ𝑒 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑒𝑎𝑟𝑙𝑦 𝑟𝑒𝑙𝑒𝑎𝑠𝑒𝑠 𝑖𝑠 𝑛𝑜𝑡 𝑎𝑓𝑓𝑒𝑐𝑡𝑒𝑑 𝑏𝑦 𝑖𝑛𝑣𝑒𝑠𝑡𝑜𝑟′𝑠 𝑠𝑒𝑛𝑡𝑖𝑚𝑒𝑛𝑡𝑠

3.2 Methodology

Replicating the Brav and Gompers approach

This thesis’ analysis builds on the original logit regression used by Brav and Gompers (2003). I recreate this regression as best as possible given the limitations on available data and then include additional variables and run a series of robustness checks.

Brav and Gompers’s 2003 paper shows that the probability of insider trading is negatively related to the information asymmetry between in-and-outside investors. They estimate a logistic model where they use several proxies to estimate the information asymmetry. Section 2.1 describes which proxies they use and the corresponding intuition. The interpretation of the results is simply the sign and significance of the estimated coefficients.

I estimate a model similar to Brav and Gompers. Due to limitations on the available data, it wasn’t possible to get exactly the same specification. The data and variables used are described in section 4.

Extension 1: Effectiveness of lockup as a commitment device

In order to test whether lockup is an effective commitment device I check if there is any abnormal return generated by insiders after the early release. I define abnormal returns as the return that would have been generated by a buy-and-hold strategy after the early

(21)

19 release up until the expiration of the lockup period relative to a benchmark. In this case the NASDAQ index.

The rationale here is that when there are information asymmetries, insiders who know about some future adverse idiosyncratic shock that will cause the market price of their equity share to go down, such as when the firm presents disappointing quarterly earnings, would be better off selling their equity share and reinvesting the proceedings in a diversified index. In this case, they would generate a negative abnormal return.

In contrast, if there are no informational asymmetries, and therefore insiders don’t have an informational advantage over the market, the future idiosyncratic shock will be priced at the moment the information is revealed8_{. In this case the expected value of the abnormal return} would be greater than or equal to zero. As such the share performance vis-à-vis the index serve as a proxy of the informational advantage of insiders.

The abnormal returns after early release is in line with the abnormal return metric used by Brav and Gompers (2003). They use the abnormal returns leading up to the early release as a proxy of the information asymmetry9_{. They state that firms “with higher returns are likely}

to have received a series of good news events and investors would be less concerned about management “cashing out”.

Figure 2 states how the two abnormal return indicators are related.

8_{This reasoning also holds if the information does not become public and is priced in. In that} case, it is a necessary condition that the underwriter has the same level of knowledge as the insider such that he will not allow the insider to cash out in the prospect of future lower market prices and therefore block the early release.

(22)

20 Figure 2. Timeline

Extension 2: controlling for the IPO cycle

In order to test whether the IPO cycle might affect the probability of an early release, I include a dummy that takes the value one for hot markets and zero for cold markets. This dummy variable is based on Figure 1, which depicts the periods of 1996-2000, 2004-2007, 2010-2016 as hot markets and 2001-2003, 2008-2009 as cold markets. The inference is that during hot IPO markets, investors are optimistic about share prices going up and, therefore, less concerned with managers not acting in their best interest. To identify whether there might have been an omitted variable bias that the year effects have picked up upon, I estimate the model both with and without the hot dummy.

The logistic regression is performed in order to avoid any dummy variable trap, or in other words to avoid perfect multi-collinearity and getting a biased estimator.

According to Brooks (2014) the logistic function F, which is a function of any random variable, z, is: 𝐹(𝑍𝑖) = 1 1 + 𝑒−𝑍_𝑖 U nlo ck D ate IPO D at e Fir st D ay o f Tr ade/ un de rp ric in g Abnormal Return prior to transaction date Tr an sac tio n D ate Abnormal Return after transaction date

(23)

21 where e is the exponential under the logit approach. The model is so called because the function F is, in fact, the cumulative logistic distribution. Thus, the logistic model estimated is:

𝑃𝑖 =

1

1 + 𝑒−(𝛼+ 𝛽1𝑥1+𝛽2𝑥2+⋯+𝑢𝑖)

where 𝑃𝑖 is the probability that the dependent variable is one, or that insider trading occurred. In the logistic model, zero and one are asymptotes to the function and thus the probabilities will never actually fall to exactly zero or rise to one, although they may come infinitesimally close. Clearly, this model is not linear (and cannot be made linear by a transformation) and thus is not estimable using Ordinary Least Squares:

𝑌_0,1 = 𝛼 + 𝛽₁𝑥₁+ 𝛽₂𝑥₂+ 𝜀

Each β parameter indicates a linear change in the probability of Y being one. This non-linear change is associated with a unit change in X, while controlling for the other explanatory variables in the regression model.

Hence, to identify the changes in the odds of the dependent variable being one as the independent variable changes, the following logit regression is performed:

𝐼𝑛𝑠𝑖𝑑𝑒𝑟 𝑆𝑎𝑙𝑒𝑠_0,1= 𝛼 + 𝛽₁𝑈𝑛𝑑𝑒𝑟𝑝𝑟𝑖𝑐𝑖𝑛𝑔 + 𝛽₂𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑅𝑒𝑡𝑢𝑟𝑛 𝐵𝑒𝑓𝑜𝑟𝑒 𝑇𝑟𝑎𝑛𝑠𝑐𝑡𝑖𝑜𝑛 + 𝛽₃𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑅𝑒𝑡𝑢𝑟𝑛 𝐴𝑓𝑡𝑒𝑟 𝑇𝑟𝑎𝑛𝑐𝑡𝑖𝑜𝑛 + 𝛽₅𝐻𝑜𝑡 𝑎𝑛𝑑 𝐶𝑜𝑙𝑑_0,1+ 𝛽₅𝑃𝐸𝐵𝑎𝑐𝑘𝑒𝑑_0,1+ 𝛽₆𝐿𝑜𝑐𝑘𝑢𝑝 𝐿𝑒𝑛𝑔𝑡ℎ + 𝛽₇𝐹𝑖𝑟𝑚 𝑄𝑢𝑎𝑙𝑖𝑡𝑦 + 𝛽₈𝐹𝑖𝑟𝑚 𝑆𝑖𝑧𝑒 + 𝛽₉𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐵𝑜𝑜𝑘𝑟𝑢𝑛𝑛𝑒𝑟𝑠 + 𝛽₁₀𝑈𝑛𝑑𝑒𝑟𝑤𝑟𝑖𝑡𝑒𝑟 𝑅𝑎𝑛𝑘 + 𝛽₁₁𝐹𝑟𝑒𝑒 𝐹𝑙𝑜𝑎𝑡 + 𝛽₁₂𝑇𝑒𝑐ℎ 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦_0,1+ 𝛽₁₂𝑇𝑖𝑚𝑒 𝐸𝑓𝑓𝑒𝑐𝑡𝑠 +

ℇ𝑖

The null hypothesis is that all the coefficients in the regression are not statistically significant and equal to zero. The logit regression will test the null hypothesis against the alternate hypothesis that all the coefficients and variables in the model predict correctly the changes in the probability of insider sales. In other words, the test identifies if the regression coefficients are significantly different from the null of zero.

(24)

22 In order to test the statistics’ impact on these hypotheses, the dependent variable must be an insider sales dummy variable. This variable is one if there is an insider selling shares during the lockup in the Company I, or zero otherwise. All the remaining variables are independent, and control ones are defined below:

• Underpricing is Company I’s return on the first day of trade relative to the IPO offer price.

• Abnormal Return Before Transaction (ARBT) is the average return obtained from the IPO date until the transaction date if the dependent variable is one. Otherwise, ARBT is the average return of the stock from the IPO date until the unlock date.

• Abnormal Return After Transaction (ARAT) is the average return obtained from the transaction date until the unlock date if the dependent variable is one. Otherwise ARAT is the average return of the stock from the IPO date until the unlock date. • Hot and Cold Market dummy, which is zero for cold markets or one for hot market. • PE Backed is a dummy variable, which is one if the IPO is backed by a Private Equity

or Venture Capital firm, or otherwise this variable is zero.

• Lockup Length is Company I’s time, in days, to lockup expiration.

• Firm Quality is Company I’s book of equity to market capitalization ratio that represents firm quality and identifies overvalued (ratio < 1) and undervalued (ratio > 1) firms.

• Firm Size is Company I’s market capitalization in the IPO, which is offer price times the number of shares outstanding after the IPO. Firm size is a control variable since larger companies have more information available, as suggested by Brav and Gompers (2003).

• Number of Bookrunners is the total number of investment bankers that marketed the IPO.

• Underwriter Rank is the Company I’s underwriter grade. According to Corwin and Schultz (2004), rank10_{represents the quality and reputation of the investment bank} that marketed the IPO.

10_{The sample is from Corwin and Schultz (2004) and includes 669 underwriters involved in} at least one IPO syndicate from January 1997 through June 2002. See Corwin and Schultz (2004) for a description of the IPO sample. The Ritter Rank is an adjusted Carter-Manaster rank based on rankings from Jay Ritter's web site. The Megginson-Weiss rank is a ranking

(25)

23 • Free Float is Company i’s percentage of shares that went public and a control variable since the firms with more shares sold on the market have more information available.

• Tech Industry is a dummy variable, which is one if Company i is a high-tech company or zero otherwise. Tech Industry is a control variable, as suggested by Brav and Gompers (2003), since technology companies have more information available. • Time effect is a control variable that takes value from 1996 to 2016 and controls for

potential time effects and thereby mitigates the omitted variable bias.

The control variables are included in the regression to avoid spurious relationship and omitted variable bias to the estimators.

Moreover, to argue if insiders traded at an assertive moment in time, and potentially traded on privileged information, the subsample means the difference of abnormal returns before and after transaction date is tested.

The outcomes obtained from the event study will be used in order to test whether the averages of the abnormal returns in the two subsamples significantly differ from each other. The means of the two subsamples with unequal variances will be compared using the following t-statistic: 𝑡 =(𝑥1− 𝑥2) − (𝜇1− 𝜇2) √𝑠12 𝑛₁+ 𝑠₂2 𝑛₂ ~ 𝑡[𝑑𝑓] With 𝑑𝑓 = √𝑠12 𝑛1+𝑠2 2 𝑛2 2 (𝑠1 2 𝑛1)2 𝑛1−1+ (𝑠2 2 𝑛2)2 𝑛2−1 based on market share of offer proceeds, with full credit given to the lead underwriter. The Underwriting Rank is a new ranking based on the proportion of offer proceeds underwritten by each underwriter, as listed in the final prospectus. Both Megginson-Weiss ranks and Underwriting ranks are adjusted to account for underwriters that enter or leave the sample as a result of mergers and acquisitions. Underwriters involved in a merger are treated as a new entity after the merger.

(26)

24 The μ1 denotes the insiders’ abnormal returns prior to transaction date subsample and μ2 denotes insiders abnormal returns after the transaction date until the unlock date subsample. The difference between the two averages will be tested using the following hypothesis:

𝐻₀: 𝜇₁− 𝜇₂ = 0 𝐻₁: 𝜇₁− 𝜇₂ > 0

The null hypothesis is that the difference between the subsamples means of the ARBT and ARAT is not statistically significant. The t-statistic tests the null hypothesis against the alternate hypothesis that μ1 is significantly larger than μ2. The interpretation if the null hypothesis is rejected is that the insiders’ decision to sell their stake beforehand and to “break” the lockup agreement was indeed an assertive decision.

(27)

25

4. Data Collection and Descriptive Analysis

4.1 Data Collection

The data collected for the insider trading research is based on Thomson One, Centre of Research in Securities Prices (CRSP) and Thomson Reuters. The last two sources can be accessed on the website Wharton Research Data Service (WRDS).

The initial sample includes companies that performed an IPO and issued ordinary shares in the years from 1996 to 2016, according to the Thomson One database. The companies are required to be listed on the NASDAQ, New York Stock Exchange (NYSE) or the American Stock Exchange (AMEX) after the offering. Foremost, the companies must have a lockup agreement. After the restrictions, there are 2,886 remaining companies conducting an IPO with lockup agreements and common shares trading in exchanges in the United States. Furthermore, 14 IPOs were excluded from the sample due to missing lockup days, which directly impacts the insider sales variable calculation.

The Thomson One database contains the ticker symbol for each company that went public, the date of shares issuance, the private equity and venture capital backed IPO flag. It includes the amount of days that the shares were locked, the IPO offer price in dollars and the percent change from the first day of trade and the IPO offer price. It also contains the shares’ amount outstanding after the offering, the number of shares sold in the IPO and the book value of equity before the offering. It also includes the main SIC code (Standard Industrial Classification), the number of bookrunners and which investment banks were involved in the offering.

Additionally, other variables are calculated based on the Thomson One sample:

• The private equity backed IPO is transformed into a dummy variable, which is one if the IPO is backed by a private equity or venture capital firm or zero otherwise; • The year dummy variable is calculated based on the date of issuance;

• The hot and cold dummy variable is calculated based on Figure 1, which depicts IPOs’ cycles and is one for years with a lot of common shares issuances or zero otherwise. It is zero for the period of 2001-2003 and 2008-2009 and it is one for the period of 1996-2000, 2004-2007 and 2010-2016.

• The total lockup length is calculated based on the sum of the days that the shares are locked in 1.217 cases of staggered lockup agreements in the sample, while the

(28)

26 remaining sample has a normal lockup with a fixed number of days to the lockup expiration;

• The unlock date is calculated based on the IPO issuance date plus the number of days of lockup, but for the staggered lockups, only the first lockup expiration date is taken in consideration;

• The free float is calculated by dividing the number of shares sold in the IPO by the outstanding amount of shares after the offering;

• The market capitalization is calculated by multiplying the IPO offer price by the outstanding shares after the offering;

• The book-to-market ratio is calculated based on the book value of equity before the offering divided by the market capitalization calculated as abovementioned;

• As suggested by Brav and Gompers (2003), the high-tech industry dummy accounts for one if the SIC code belongs to the following industries: computer manufacturing (SIC codes: 3570-3579); electronic equipment (SIC codes: 3660-3669, 367-3679, 3610-3659, 3680-3699); computer and data processing services (SIC codes: 7370-7379); optical, medical and scientific equipment (SIC codes: 3810-3849); and communications (SIC codes: 4800-4899); or zero otherwise.

In order to complement the dataset with additional variables, some other data sources are used. Firstly, the Corwin and Schultz (2004) rank is used to calculate the underwriter prestige variable, which is the average grade of the bookrunners that have underwritten the IPO. In the cases that the investment bank is not included in the rank, the grade attributed is zero. Secondly, CRSP data is assessed on the WRDS website to calculate the abnormal return variable before and after insiders’ sales transaction date. The Abnormal Return Before Transaction variable is the difference, on average, between the return obtained from the first day of trade until the transaction date for the companies that had insider sales before the lockup expiration date and the NASDAQ return for the same period. For companies that did not have any insider trading after the IPO, this research uses the difference between return in the whole period and the NASDAQ return in the same period, from the first day of trade until lockup date expiration. The Abnormal Return After Transaction variable is the difference, on average, between the return obtained from the transaction date until the unlock date for the companies that had insider sales before the lockup expiration date and

(29)

27 the NASDAQ return on the same period. For companies that did not have any insider trading after the IPO, the difference between return in the whole period and the NASDAQ return in the same period, from the first day of trade until lockup date expiration has been used. Additionally, to test if the average abnormal returns prior and post transaction date are statistically different from each other, the subsample of Abnormal Return Before Transaction and Abnormal Return After Transaction variables of 571 observations have been selected. Finally, information regarding the insider sales is sourced from the Thomson Reuters database, provided by the WRDS website. This database reports all the US insider trading activity as disclosed in Forms 3, 4, 5 and 144 by the SEC, already discussed in section 2.8. The dependent dummy variable, inside trading, is calculated based on Table 1 from the SEC Form 4 collected from Thomson Reuters in the Wharton WRDS database. Form 4 is considered the most important insider document and it forms the basis of the main insiders’ dataset. Table 1 contains conventional stock or non-derivative transaction information. Most of the total transactions (purchases, sales, and gifts) fall under Table 1. Table 1 reports the date of the transaction, number of shares bought/sold, and price paid/received. It also tells the number of shares remaining after the trade. Insiders are required to report both Direct and Indirect holdings. It is important to note that shares purchased by insiders were excluded from the sample and also any missing information about these purchases. Hence, the results and impact can be even larger than what is found in the research. The dummy variable is one if insiders filed that they sold shares on a transaction date between the issuance date and the unlock date or zero otherwise.

At last, note that Thomson One has a lot of missing data for certain companies, which can impact positively or negatively the results of the research. In order words, the relationship found in this study could be even larger or smaller. Moreover, it would be interesting to go further in the study and look into the prospectuses to fill in the missing data from Thomson One and look closely at the SEC and to fill in the missing data about insiders’ filings that are neither clear acquisitions nor clear disposals. The CRSP database also has problems with finding specific stocks prices on a specific date, so the assumption made is that if there is no price from the event date plus 5 days, the return is considered as missing. This occurs in 100 observations of the subsample. Hence, the abnormal returns effect could also be larger or smaller due to the missing data. Overall, the main regression contains 1.324 observations for the whole period between 1996 and 2016.

(30)

28 4.2 Descriptive Analysis

This section provides 3 tables in which the sample is summarized. The total sample consists of 2,886 companies with lockup agreements, but only 2,872 companies have the lockup period provided by Thomson One. The subsample of abnormal returns contains 571 companies in which insider trading occurred.

The summary statistics of the full sample can be found in Table 1. This table depicts that the insider sales sample is a dummy variable that has no missing values. Also, it can be assumed that for most of the time there is no insider selling shares beforehand, since 25% of the observations are zero or more. The underpricing sample has 293 missing data. On 25% of the observations the underpricing is 26.67% or more, and on average it is 20.63%. Thus, this phenomenon can be assumed to be recurring on the IPOs in the sample. The minimum and maximum ranges are both very large [-99.61%;966.67%, respectively]. The abnormal return before transaction has 407 missing values in the sample and on 25% of the observations, it is 12.29% or more. On average, the abnormal return before transaction is -4.23%. The minimum and maximum ranges are also very large [-193.67%;468.46%, respectively]. The abnormal return after transaction has 407 missing values in the sample and on 25% of the observations it is 13.55% or more. On average the abnormal return after transaction is -6.59%. The minimum and maximum range are -193.67%;468.46%, respectively. The hot and cold market dummy shows that 93% of the insiders’ sales occur in hot markets. The PE backed variable is also a dummy variable that has no missing data. It can be assumed that most of the time (58% of the cases), the IPOs are an exit for private equity or venture capital funds. The lockup length sample is on average 334 days, which is high compared to average reported by Brav and Gompers, which is 180 days. The average is skewed however. The median lockup period is 180 days, but in 25% of the observations, it is over 540 days. This can be explained by the way staggered lockup periods are accounted for. The book to market ratio, which is a proxy for firm quality, has 497 missing observations, and 25% of it is over 0.14. The average ratio is 0.05. Hence, the majority of the IPOs are overvalued on the IPO pricing. This analysis is also in line with the literature about IPOs being associated with underperformance in the long run. The minimum and maximum ranges are [-6.91;59.58, respectively]. The market capitalization is a proxy for the firm size, and since the values may have a large range, a logarithmic distribution is used to normalize it. Hence, statistics

(31)

29 analysis does not show large variations. On average, the IPOs in the sample have 1.51 bookrunners, and in 25% of the observations, they have 2 or more investment banks working on the IPO. The range of bookrunners of the IPO in the sample may vary from 1 to 15 investment banks. The underwriter rank average grade is on average 2.66 and in 25% of the cases, 4.05 or more, based on the Corwin and Schultz (2004) reputation evaluation. The minimum and maximum grades are 0 and 9.25. The free float for companies in the sample that went public is on average 32.01%. For 50% of the observations the free float is 27.2% or more and for 25% of the observations, it is 37% or more. The minimum and maximum ranges are [0.31%;200%, respectively]. The high-tech industry variable is a dummy variable that for 25% of the observations what is 1 or more, and on 50% of the observations it is 0 or more. Forty percent of companies that went public are in the technology sector.

This tables provides the summary statistics for the full sample.

Insider sale is a dummy variable which is one is any insider sold shares before the end of the lockup agreement; Underpricing is the return on the first day of trade; Abnormal Return BT is the difference between the average return from the first day of

trade until the transaction when an insider trade occurred and the NASDAQ return on the same period or the difference between the return from the first day of trade until lockup expiration date and the NASDAQ return on the same period; Abnormal Return (AT) is the difference between the average return from the transaction date to the unlock date when an insider trade occurred and the NASDAQ return on the same period or the difference between the return from the first day of trade until lockup expiration date and the NASDAQ return on the same period. The hot and cold market dummy accounts for 0 in cold markets and 1 otherwise; PE Backed is a dummy variable which is one if the IPO is backed by a private equity or venture capital fund, or zero otherwise. Lockup Length is total amount of days that insiders cannot sell their stake under the

lockup agreement. The values attributed to the firm size variable is the logarithmic distribution of the market capitalization distribution that is not normalized. The values attributed to the firm quality variable is the firm’s book value of equity divided

by the market capitalization; that is, the value is smaller than one if the firm is overvalued and larger than one if the firm is undervalued in the IPO valuation. The number of bookrunners is the total amount of investment banks that market the IPO. The underwriter rank is the average grade of bank’s quality that market the IPO. The grade given is based on the Corwin and

Schultz (2004) rank study. The free float is the amount of shares sold in the IPO divided by the total amount of shares outstanding after the offering.

(32)

30

Table 1. Descriptive statistics of the Sample

Obs 25th Percentile Mean Median 75th Percentile Min Max SD Insider Sale 2,872 0 0.23 0 0 0 1 0.42 Underpricing 2,579 0.00% 20.63% 7.71% 26.67% -99.61% 966.67% 0.50 Abnormal Return BT 2,465 -34.44% -4.23% -6.73% 12.29% -193.67% 468.46% 0.54 Abnormal Return AT 2,465 -36.74% -6.59% -11.62% 13.55% -193.67% 468.46% 0.53 PE Backed 2,872 0 0.58 1 1 0 1 0.49 Lockup Length 2,872 180 334.02 180 540 30 2160 199.80 Firm Quality 2,375 -0.08 0.05 0.02 0.14 -6.91 59.58 1.30 Firm Size 2,575 18.39 19.13 19.09 19.75 15.46 25.12 1.13 #Bookrunners 2,872 1 1.51 1 2 1 15 1.08 Underwriter Rank 2,640 0.40 2.66 1.76 4.05 0.00 9.25 2.58 Free Float 2,575 20.35% 32.01% 27.20% 37.00% 0.31% 200% 0.20 Hot /Cold 2,872 1 0.93 1 1 0 1 0.25 Tech Industry 2,872 0 0.40 0 1 0 1 0.50

The summary statistics of the insider sales sample can be found in Table 2. This table tabulates the dependent variable distribution that is in line with Table 1. It can be inferred that in 25% of the observations the variable can be 0 or more. Thus, very few cases of early releases will occur beforehand and according to Table 2 early releases occur in 671 of companies’ IPOs out of 2,872 companies in the sample; that corresponds to 23.36% of the observations.

Table 2. Frequency of Insider Sales in the Sample

Insider Sale Value Freq. Percent Cum.

0 2,201 76.64 76.64

1 671 23.36 100.00

(33)

31 The summary statistics of the subsample can be found in Table 3. The subsample includes 571 observations out of the 671 companies that experienced inside sales and this difference is due to prices missing data on the CRSP search for a specific time period. The price search is executed using the company ticker and a specific date, such as the IPO date, the transaction date and the unlock date plus 5 days and if no price is found in this window, the observation is excluded from the dataset. This table depicts that the abnormal return until the insiders’ transaction date sales is on average 5.89% and in 25% of the observations is 9.78% or more. The minimum and maximum values have a wide range [-146.77%;436.03%]. The abnormal return from the transaction date until the lockup expiration date is on average -3.87%, and in 25% of the observations, it is 10.92% or more. The minimum and maximum range are not as wide as the abnormal return before the transaction, and this range accounts for -145.19% minimum and 190.41% maximum.

Table 3 shows the summary statistics of the subsample in which company’s IPOs had insider sales before the end of the lockup agreement. The abnormal return before transaction is the difference between the average return obtained from the first day of trade until the transaction date that the insider sold and the NASDAQ’s return in the same period. The abnormal return after the transaction is the difference between the company’s stock return in the whole period from the first day of

trade until the lockup expiration date and the NASDAQ’s return on the same period.

The difference between the number of observations in Table 1 and the Table 2 insider sales observations of 671 companies is due to missing price data on CRSP for specific IPO dates or transaction dates or for unlock dates that impact returns

calculations. Moreover, some NASDAQ pricing information is also missing in the same dates.

Table 3. Descriptive statistics Subsample

Obs 25th Percentile Mean Median 75th Percentile Min Max SD Abnormal Return before transaction 571 -5.46% 5.89% 5.89% 9.78% -146.77% 436.03% 0.37 Abnormal Return after transaction 571 -22.77% -3.87% -5.14% 10.92% -145.19% 190.41% 0.36

(34)

32

5. Results

The hypotheses mentioned in section 3.1 will be tested with the methodology explained in section 3.2 using the data collected and described in section 4.

Brav and Gompers find that “…early sales by insiders are more likely in firms that have lower

incentives for moral hazard. This includes firms that have had higher post-IPO returns, firms that go public with high-quality underwriters, and firms that were financed by venture capitalists.”

Firstly, table 4 replicates, as well as possible given the limitations on available data, the model specification by Brav and Gompers for the period from 1996 to 2016. Similar to Brav and Gompers, I find that the abnormal return prior to the early release and private equity backing are positively related to the probability of an early release. However, I do not find evidence that firm size, as measured by market capitalization, has a statistically significant relationship with the probability of an early release.

Unlike Brav and Gompers, I find that prestigious and non-prestigious underwriters are likely to agree with lockups early releases.

The first novelty of this thesis is introducing abnormal returns after the early release to test the effectiveness of the lockup in mitigating information asymmetries. The new variable included, abnormal returns after transaction, is statistically significant and negatively related to early releases. This means that the probability of underwriters suspending the lockup period and consequently insiders selling their shares beforehand is larger if the consecutive drop in share price relative to the benchmark goes up. It is worth mentioning that the abnormal return post transaction is not observable at the time of the sale. However, it does imply that insider sales are more likely for shares that are about to experience a large drop in share price. This suggests that insiders have an informational advantage over markets and their underwriters.

(35)

33

Table 4. Insider Sales Logit Regression (1) Whole Period VARIABLES 1996-2016 LUNDERPRICING 0.160*** (0.0511) ABN RETURN BT 2.054*** (0.470) ABN RETURN AT -1.769*** (0.492) PE BACKED IPO 0.354** (0.175) LOCKUP LENGTH 0.00214*** (0.000663) BOOK/MARKET -0.0496 (0.0377) LMARKETCAP -0.130 (0.0989) #BOOKRUNNERS 0.126 (0.0808) UNDERWRITER RANK -0.000756 (0.0314) FREE FLOAT -0.744 (0.485) TECH IND -0.0754 (0.149) YEAR DUMMY YES Constant -0.210

(1.865) Observations 1.324 Pseudo R² 0.1844

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Secondly, I test whether the IPO cycle manages to explain variation in early releases. I do this by introducing a dummy variable that takes value one for “hot markets” and zero for “cold markets”. As shown in table 5 this variable is negatively related to the probability of an early release, meaning that it is less likely that insider sales occur during hot markets. It should be noted that the overwhelming majority of early releases occurred during the lockups of IPO’s conducted during hot markets, namely 93%. While an unbalanced sample in and of itself does not introduce a bias to the estimation, there might be an omitted variable bias at play which could affect the results. It could be that opportunistic prepublic investors choose to bring their firms public during hot markets and want to cash out as soon as possible, “opportunism of prepublic investors” and an omitted variable.The rationale is that