The length of the lockup versus long term performance : an examination of the role of the lockup.

(1)

The Length of the Lockup versus Long Term Performance;

an Examination of the Role of the Lockup

Aaron Renkers 6081819

University of Amsterdam Faculty of Economics and Business

Master Thesis February 2014

Abstract

This paper examines the role of the lockup agreement by relating lockup length to long term performance, price reaction around expiration and underpricing using a sample of 1046 initial public offerings in the United States for the period 1996-2005. This study is unique because it measures buy-and-hold abnormal returns as a proxy for firm quality to determine the role of the lockup. I find that issuing firms that agree to longer lockups have a significantly worse one and three year long term performance measured by the buy-and-hold abnormal returns. I also find a significant negative relationship between lockup length and the price reaction around expiry, measured by three and five-day cumulative abnormal returns. A significant positive relationship is found between lockup length and underpricing. Furthermore, I analyze the subgroup of issuing firms with multiple lockups. However, I do not find statistical evidence for this group which might be due to a limited dataset. The significant results support the commitment theory of lockups and thereby show that the lockup agreement serves as a bonding device to convince investors of aligned interest.

MSc Business Economics, Finance track Supervisor: Jens Martin

(2)

1

1. Introduction

A lockup agreement prohibits insiders from selling their shares for a particular period after the IPO. The length and terms of a lockup agreement may vary. Moreover, a lockup agreement is not mandated by the Securities and Exchange Commission (SEC). Keasler (2001) mentions that the period of locking up shares declined over the years to a standard level of 180 days for IPOs in the United States. Still, not all issuing firms establish a lockup period equal to 180 days. The explanation for variety in terms of the length of the lockup is something that has been widely studied.

According to existing literature, there are two main theories that explain the divergence of lockup periods and thereby the role of the lockup: the signalling and commitment theory. The first theory suggests that lockups serve as a signalling device for issuing firms to signal their true quality. Courteau (1995) and Brau et al. (2005) argue that the probability of negative information to be revealed is less for firms of high quality. Therefore, these high quality firms can afford to accept longer lockups and thereby signal their quality to investors. In contradiction, the commitment theory suggests that lockup agreements serve as a bonding device to convince investors of aligned interest. Brav and Gompers (2003) mention that according to the commitment theory high quality firms are associated with less information asymmetry and therefore less agency costs. These firms of high quality therefore need a shorter bonding commitment to convince investors, which means that these high quality firms have shorter lockups.

As mentioned, both of the above discussed theories predict a complete opposite relationship between lockup length and the quality of the issuing firm. A limitation of all the existing literature regarding the role of the lockup is that these studies do not relate the length of the lockup period to a real measure of firm quality. This study is unique since it measures firm quality as long term stock performance calculated by one and three-year BHARs. For additional support, I relate lockup length to the market reaction around expiry and underpricing. I also examine the market reaction for different types of insiders and “earlier” and “later” lockups. Ordinary Least Squares regression models are used to test the above mentioned relationships. The tested hypotheses are developed based on the commitment role of lockup agreements.

The results support three out of the four developed hypotheses based on the commitment theory regarding the role of lockups. First of all, I find a significant negative relationship between lockup length and both the one and three-year BHARs. This finding is consistent with the idea that issuing firms of high quality on average have shorter lockup periods and is supporting evidence for the commitment theory. Second, I show that there is a significant negative relationship between lockup length and the price reaction around expiration of the lockup period which again supports the commitment theory. As predicted by the hypothesis, the results show that the CARs around lockup

(4)

3

expiry are more negative for issuing firms with a high probability for agency problems to occur. In other words, longer lockups are associated with worse CARs. Third, I show that there is a significant positive relationship between lockup length and underpricing. This finding provides additional support for the bonding role of lockups. Finally, I do not find significant nor robust results for the comparison between different types of insiders and “earlier” and “later” lockups regarding the price reaction around lockup expiry. The fact that I do not find a significant relationship might be due to a limited dataset. I therefore suggest that further research should try to obtain more data in order to perform a robust analysis on this part. Summarized, the results give insight in the exact role of the lockup. Knowing the motive to establish a lockup period can be valuable to investors. Moreover, it could be dangerous for investors to think that issuing firms signal their quality by agreeing to a longer lockup period, while the opposed relationship is true.

The remainder of this paper is organized as follows. Section two provides an overview of existing literature regarding the topic. Furthermore, section three develops the hypotheses based on the commitment theory of lockups and describes the methodology to test each of these hypotheses. The results are presented in section four of this paper. I conclude with section five, in which I summarize and discuss the main results.

(5)

4

2. Background Literature

2.1 Introduction

This section will first discuss in general the process of going public. Second, it will describe existing literature regarding the role and use of the lockup period. Finally, post-IPO performance and the market reaction around expiration of the lockup will be discussed.

2.2 Going Public

An initial public offering (IPO) refers to the first time that a company issues common stock. In other words, the company goes public. There are various reasons to go public for companies. Berk and DeMarzo (2011) list these possible reasons. First of all, the need for funding could be a reason to go public. The proceeds of issuing shares can be used for investment purposes and help firms to meet their growth strategy. Thereby, some firms prefer to raise capital through an IPO instead of taking more debt. The reason for this is that equity does not have to be repaid. Public ownership also helps for better accessibility of future funds. Second, going public provides liquidity for the shareholders. Public ownership enables shareholders to easily buy and sell their shares in the stock market. Third, the better accessibility of capital will reduce capital costs. A firm that is publicly owned can raise capital in several ways. This means that it is not only depending on the bank’s willingness to lend. Moreover, the costs of bank credit loans will be lower due to a better negotiation position. Fourth, public ownership will put a market price on the firm. This facilitates merger and acquisition transactions and can be desirable for firms that want to grow using M&A. In this way, the stock can be used as a currency to acquire other companies. The firm itself also becomes a more attractive suitor for potential acquisitions. Finally, Berk and DeMarzo (2011) mention that public ownership provides a way to cash out, makes it able to use stock-based incentives and that it increases the visibility of the firm.

The paragraph above lists some advantages of public ownership. However, it is not always favourable. Berk and DeMarzo (2011) suggest that one of the major advantages of going public also is the major disadvantage. Although the IPO provides liquidity and enables shareholders to diversify, it also means that shareholders become more widely spread. Then it will be more difficult for investors to monitor the management of the firm which could increase information asymmetries. Going public therefore denotes a loss of control. A second disadvantage is the fact that public companies are highly regulated. By going public a firm needs to satisfy reporting requirements, which

(6)

5

will increase accounting costs. It also has to disclose its operations and policies and with that provides information to its competitors. Finally, going public is a costly and time consuming process. Firms may spend 25 percent of the raised equity on expenses that apply to the IPO.

Then the question arises: how to go public? Each IPO is managed by an underwriter which in fact is an investment banking firm. Most IPOs are managed by multiple underwriters, the lead underwriter and the syndicate. The primary investment banking firm that is responsible is called the lead underwriter. The syndicate are other underwriters that help to sell the issue. Berk and DeMarzo (2011) distinguish three types of offering structures. First, the underwriter accepts to lead the IPO along a best-efforts basis. This design is commonly used for smaller IPOs. The underwriter tries to sell the shares at the best possible price, but it will not guarantee that those shares will be sold. It often means that either all shares will be sold or the deal will not be executed at all. Another design is to issue shares using an auction IPO. Over some particular time period investors will place their bids for some number of shares. Then the underwriter sets the price in such a way that all shares to be issued can be sold. This means that each winning bidder will pay the same price for a share. Finally, the underwriter and the issuing firm could sign a firm commitment contract. In this case, the underwriter does guarantee to sell all shares. It will purchase all issued shares at a price slightly under the offer price, which is set by the underwriter itself. The underwriter therefore is exposed to the risk of making a loss that will incur if it for instance will not sell all shares.

For most IPOs the first trading day results in an abnormal increase in the firm’s stock price, which means that the offer price initially undervalued the issuing firm. This phenomenon is called IPO underpricing and has been studied a lot. Ibbotson (1975) was one of the first who studied the price performance of newly listed firms. He finds an average first day return of 11.4 percent for IPOs in the US for the period 1960 till 1969. Other studies of Carter et. al. (1998) and Mohan and Chen (2001)

likewise find positive first day market adjusted returns of respectively 8.08 and 10.63 percent. Loughran and Ritter (2004) mention that IPO underpricing has changed over time. On

average, issuing firms around the 1980s experienced returns of seven percent after the first trading day. These initial returns doubled over the period 1990-1998 and heavily increased to 65 percent in the years 1999 and 2000, when the internet bubble took place. After that peak, the average first day returns declined to 12 percent during the period from 2001 till 2003. Loughran and Ritter try to examine the reason for this underpricing pattern. They argue that the increase in 1999-2000 partially can be explained by different firm characteristics, since most IPOs during the internet bubble concern very young companies which are more difficult to value correctly. The overall increase over the whole period can be explained by the fact that issuers had the incentive to choose to work with underwriters that have highly ranked analysts and thereby did not consider whether the underwriter had the reputation to heavily under price IPOs. This desire is called the analyst lust hypothesis.

(7)

6

Why are IPOs involved with underpricing? Baron and Holmström (1980) argue that there is information asymmetry between the issuer and underwriter. This is created during the registration period, when the underwriter gains more information regarding the demand for the issuing stock. A later study of Baron (1982) shows that there is an incentive to underprice more since underwriters want to minimize costs and risks. Setting a lower price will help to sell shares faster and attract more potential buyers without a huge marketing effort. Moreover, the risk of losing money for underwriters increases for larger IPOs. Which indicates that larger IPOs will be more underpriced. Likewise, there is information asymmetry between the issuer and investors. Beatty and Ritter (1986) show that there is a positive relationship between the investor’s uncertainty about the real value of the issuing firm and initial returns. Investors need to be compensated for this missing information which results in a lower offer price and thereby creates underpricing.

According to Schultz (1993) the initial returns are less for larger issuing firms since it is a measure of information asymmetry. They correct for IPO size using a variable that divides the proceeds by the assets of the issuing firm and find that underpricing decreases with 0.35 percent if variable increases with one percent. A better measure for size and thereby uncertainty is pre-IPO total assets, like Leone et al. (2003) use in their model to explain underpricing. Besides taking into account for IPO size, Schultz (1993) shows that there is a negative relationship between the age of the firm at the IPO and underpricing. Investors are better able to correctly price older firms because there is more information available about them, which results in lower underpricing.

Lee and Wahal (2003) study the role of venture capitalists in IPO underpricing. They show that issuing firms backed by VC experience higher initial returns. The average difference in returns compared to non-VC backed group of issuing firms lies between 5.01 and 10.32 percent for the period 1980 until 2000. They argue that their results are consistent with the theory of Brav and Gompers (1996) that the cost of experiencing more underpricing will be compensated by the benefit from grandstanding.

2.3 The Lockup Period and its Role

Another important and intensively studied aspect of IPOs is the lockup period. A lockup agreement prohibits insiders from selling their shares for a particular period after the IPO. Insiders can be founders, managers but also venture capitalists that funded the firm when it was private. The length and terms of a lockup agreement may vary. Moreover, a lockup agreement is not mandated by the Securities and Exchange Commission (SEC). Information about the lockup agreement is disclosed in an IPO prospectus. The “shares eligible for future sale” section of the prospectus states among others how much of the shares are subjected to the lockup, for which shareholders the lockup

(8)

7

agreement applies to and as mentioned above the length of the lockup period. The most common length of lockup for US firms is 180 days (SEC, 2013). For almost 90 percent of the IPOs in the sample that Bradley et al. (2001) used in their study a lockup of exactly 180 days was in place. They tried to examine why most lockup periods equal 180 days, but were not able to answer this question. Keasler (2001) mentions that the period of locking up shares declined over the years to a standard level of 180 days, due to the increased competition in the investment banking industry.

Most studies that examine the lockup period are done with data from IPOs in the United States. US studies like those of Brau et al. (2004), Field and Hanka (2001) and Mohan and Chen (2001) all report standardised lockups of 180 days. However, evidence from studies that focus on other markets is not in line with these US studies. Especially UK studies show lockup lengths that deviate from the standard of 180 days. Espenlaub et al. (2001) mention that lockup agreements of LSE listed firms are not only more diverse, but also more complex in contrast to US firms. They report an average length of lockup of 561 days, which is significantly higher than the standard US lockup period. In the same way as Espenlaub et al., a recent study of Hoque (2011) also reports more diversity with respect to the lockup agreement. He analyzes lockup agreements from the LSE between the years 1999-2006 and finds that the average length of locking up shares lies in the range of 383 and 714 days. In sum, empirical evidence shows that US firms that go public have significantly lower lockup periods than firms in the UK.

Although US firms are not committed by law to sign a lockup agreement, most firms that go public do have a lockup agreement established with the lead underwriter. Field and Hanka (2001) and Keasler (2001) both argue that a longer lockup period is beneficial for the underwriter because it makes sure that the IPO is not destabilized by insider sales. Insiders themselves would prefer a shorter lockup period since it comes with a cost of illiquidity.

The major motivation to establish a lockup is to solve information asymmetries between insiders and investors. Existing literature shows mixed results regarding the exact role of the lockup. In fact, there are two theories with respect to the role of the lockup: The signalling theory and the commitment theory.

The firm’s insiders are informed and able to correctly value the firm, while other uninformed investors are unable to do this. Leland and Pyle (1977) argue that insiders can signal firm quality at the IPO by retaining a large equity stake. This signal of willingness to invest in their own company does not hold if insiders are able to sell their shares directly after the IPO. As Gale and Stiglitz (1989) mention, the insiders of a low quality firm are able to mimic the signal of quality by holding a great portion of the shares up to the IPO and sell them as quickly as they can after the IPO. Establishing a lockup period will increase the cost of mimicking of insiders of low quality firms since the real value of the firm could be disclosed during the lockup period.

(9)

8

The study of Leland and Pyle has been extended by Courteau (1995). Besides the retained insider ownership, she introduces the length of the lockup period as a device to signal firm value. By including both the retained ownership and length of lockup into a model to explain a firm’s risk and information characteristics, she finds that firms of high quality are more likely to have a longer lockup period to be able to convince uninformed investors of their value.

The signalling theory is tested by Brav and Gompers (2003) using three different measures of firm quality. First, they mention that information regarding the lockup period is available to investors via the prospectus a long time before the company goes public. Hence, investors are able to separate good quality firms from the bad ones. Then high quality firms should be able to revise the offering price upwards since investors are willing to pay more for those firms. However, Brav and Gompers find that the group of firms that revise their price upwards have shorter lockup periods which contradicts the signalling theory. In addition they use the probability of having a seasoned equity offering and dividend initiations as measures of firm quality based on studies of respectively Welch (1989) and Michaely and Shaw (1994). A comparison using both measures of firm quality reveals once more that firms of high quality have shorter lockups. In conclusion, Brav and Gompers reject the signalling theory of lockups.

In a reaction on Brav and Gompers, Brau et. al. (2005) argue that the measures of firm quality are not correct. They mention that firms with high information asymmetry and therefore have a greater need to signal their quality, on average have longer lockups. They also argue that having a longer lockup will lower the cost of mimicking. However, this contradicts the theory of Courteau and is definitely not logic. Brau et. al. acknowledge that their findings regarding the cost of mimicking is neither significant or robust.

While Brav and Gompers (2003) reject the signaling theory, they find evidence that supports the commitment theory which indicates that the lockup exists to reduce moral hazard problems. Agency costs arise when the firm’s insiders do not act in the interest of outside investors. Establishing a lockup period reduces the ability for insiders to take advantage of other investors. Therefore, potential investors would be more likely to invest in the firm. The agency problem applies more to firms that experience high information asymmetry. The reason for this is that investors are not able to monitor the actions that are taken by managers. Brav and Gompers argue that the probability for high agency costs is less for firms of high quality. Therefore high quality firms are able to incorporate a shorter lockup period since investors need a relative small bonding commitment from these firm’s insiders.

Their results show that the lockup period of larger firms and venture capital backed firms are shorter on average. These firms are associated with less information asymmetry. Brav and Gompers argue that VC backed firms don’t need to show their commitment by establishing a longer lockup

(10)

9

period since those firms probably will not take advantage of information asymmetries. In addition, they show that (profitable) firms with higher cash flow margins have shorter lockups, which again indicates that firms of high quality have shorter lockup. On the other hand, they show that riskier firms (indicated by lower book-to-market ratios) and firms with high growth perspectives have longer lockup periods in place.

As mentioned above, following the commitment theory there is a positive relationship between the length of the lockup period and the degree information asymmetry. Then one could expect that also underpricing is related to the length of the lockup period. An earlier study of Brav and Gompers (2000) shows that there is indeed a strong positive relationship between the length of the lockup period and underpricing, which gives the commitment theory additional support.

The above mentioned existing literature shows mixed results regarding the role of the lockup period. The problem with these studies is that they never related the length of the lockup with a direct measure of long term performance. Which in fact is crucial to test the relationship between the lockup period and the quality of a firm and thereby examine whether the lockup period functions as a signalling or commitment resource. High quality firms have longer lockups following the signalling theory. This would mean that issuing firms with longer lockup periods in place should have better long term performances. However, if the commitment theory is correct, then firms with longer lockup periods will have a worse long term performance caused by higher agency costs.

2.3 Long Term Performance of IPOs

Existing literature shows that firms perform well immediately after the IPO. As mentioned before, this phenomenon is called underpricing. However, empirical evidence shows a contrary performance of IPOs for the long run. Ritter (1991) was the first who examined the long term performance of newly listed firms. In his sample of 1,526 IPOs during 1975-1998 he finds an average underperformance of 17 percent measured over a three year period after the IPO. The paper shows that the long term performance depends on the year in which the company goes public and the industry in which it operates. As mentioned before, Loughran and Ritter (2004) show that there are years with a relatively high number of IPOs and relatively high underpricing. Demers and Joos (2007) argue that in these periods the market conditions are favorable for issuing firms since they are able to raise capital at a relative low cost. This attracts lower quality firms that want to take advantage of these favorable conditions, which perform worse in the long run. For instance, Coakley et al. (2007) show that this is the case for the years 1998 until 2000 when the internet bubble took place. Regarding the differences between industries, Ritter (1991) uses a wealth relative to show industry

(11)

10

specific long term performances and documents an outperformance for financials, drugs and airlines. Furthermore, Ritter (1991) includes age of the firm at the IPO as a measure for risk and level of information asymmetry in his model. His results suggest a strong positive relationship between age and post-IPO performance, but Ritter acknowledges that this result is somewhat biased. Most of the oldest firms in his sample concern financial institutions, which heavily outperformed during the studied period. Other studies of Schultz (1993), Hensler et al. (2004) and Demers and Joos (2007) confirm the positive relationship between age and post-IPO performance. These studies show that older issuing firms are more likely to survive.

Later studies of Brav et al. (2000) and Ritter and Welch (2002) both find the same results as Ritter regarding post-IPO performance. Ritter and Welch find an average underperformance of 23.4 percent for IPOs in the US from 1980 to 2001. The results of Brav et al. (2000) show that IPOs in the US for the period 1975 to 1992 underperform the S&P 500 and Nasdaq respectively by 44 and 31 percent over a five year period. They mention that the underperformance is mainly driven by issuing firms that are small and have low book-to-market ratios. Earlier studies of Schultz (1993) and Hensler et al. (2004) highlight the relationship between size and post-IPO performance as well. Both studies show that larger issuing firms are more likely to survive.

The earlier mentioned study of Coakley et al. (2007) measures long term performance of IPOs using the five year operating cash flow and in line with other studies finds a post-IPO underperformance. Besides taking into account for size and age, they include a ratio which divides EBIT by total assets. By incorporating this variable into the model, they control for earnings and thereby measure how well the issuing firm is performing.

The long term underperformance does not hold for all type of firms that go public. Brav and Gompers (1997) replicate the results of Loughran and Ritter (1995) regarding post-IPO performance in the US and show that the VC backed IPOs outperform their counterparts when equal weighting the returns. Kraus and Burghof (2003) give a more detailed analysis of the performance of VC and non-VC backed IPOs in Germany. They suggest that firms which have venture capitalists as insiders might outperform the market due to a better corporate control, a faster decision making process and third party certification. Nevertheless, the above does not hold when venture capitalists try to cash out and therefore exit the firm. The results of Kraus and Burghof show that the expiration of the lockup period is a breakpoint in time regarding the performance of VC backed IPOs. In concrete terms, VC backed IPOs perform better before expiration of the lockup period and worse after expiration of the lockup compared to non VC backed IPOs.

Some studies directly relate underpricing to long term performance and do not all find the same result. Studies of Schultz (1993) and Hensler et al. (2004) that model post-IPO survival both find a positive relationship between underpricing and survival of IPOs. Contrary to these studies, a study

(12)

11

of Hamza and Kooli (2010) regarding IPO long term survival in the US for the period 1985-2005 shows that issuing firms with higher underpricing have a higher probability to fail.

Finally, not all studies confirm the long term underperformance of IPOs. The before mentioned study of Coakley et al. (2007) did find long term underperformance on average over the whole period for their sample of UK IPOs from 1985-2003. But they suggest that these statistically significant results may not be robust. They argue that the results are mainly driven by the total market underperformance during the internet bubble. By analyzing only the ‘normal’ period after the internet bubble, they find that IPOs in the UK perform normally during 2001-2003. Likewise, the before mentioned study of Brav and Gompers (1997) regarding VC backed IPOs doesn’t support the long term underperformance of IPOs. The authors show that the results of Loughran and Ritter (1995) are different when they are corrected for size and book-to-market ratios. They mention that performance of small IPOs with a low book-to-market ratio is not significantly different from similar non issuing firms.

2.4 Market Reaction around Expiration Lockup

Besides IPO underpricing, long term underperformance and the use and role of lockup periods, a lot of studies examined the market reaction around expiration of the lockup period. All of these studies find negative abnormal returns around expiration of the lockup period. The negative abnormal returns that occur around expiration of the lockup period are a violation of the Efficient Markets Hypothesis of Fama (1965). The characteristics of the lockup agreement are basically known by everybody. But still, empirical evidence shows that the share price does not reflect a fully anticipated event. For a sample of 1662 IPOs in the US during the period 1996-1998, Ofek et al. (2000) show a drop in the share price at expiration of the lockup period in the range of 1.15-3.29 percent. This price drop goes hand in hand with an increase in trading volume of 38 percent.

Some studies not only examine the abnormal returns on the lockup expiration day itself, but also examine a longer event period. The results of the study of Bradley et al. (2001) also show significant share price declines around the expiration date of the lockup period. At expiration they find an average price decline of 0.74 percent and for a five-day surrounding period they find an average negative cumulative abnormal return of 1.61 percent. Field and Hanka (2001) and Brav and Gompers (2003) find similar results regarding the market reaction to the expiration of lockups. In addition, Field and Hanka show that the trading volume for the whole sample increases with 80 percent one day after the expiration of the lockup. The increase in sales volume at expiration of the lockup period is most likely boosted by insider sales. Field and Hanka (2002) suggest that insider sales tend to carry bad news, because insider sales show a lack of insider confidence. Additionally, the

(13)

12

increase in supply of shares could cause the share price to decrease. Stephens and Dussold (2007) show that the negative abnormal returns start to occur multiple days before the expiration of the lockup. They therefore suggests that investors anticipate on the coming expiration date and moreover the potential wealth loss by selling some of their shares before the end of the lockup. Still, the negative abnormal returns around expiry show that it is not a fully anticipated event.

The market reaction around expiration of the lockup period is worse for firms that are backed by VCs. Studies of Bradly et al. (2001), Field and Hanka (2002) and Brau et al. (2004) show that this negative relationship is highly significant. Brau et al. argue that VCs contrary to other insiders do not participate in the firm for a long period. They want to exit their investment probably at the first possibly opportunity, which in fact is the end of the lockup period. By looking at the negative five-day abnormal return, the researchers also suggest that the market expects VCs to exit their investment at the end of the lockup.

Brav and Gompers (2003) and Brau et al. (2004) mention that the percentage of primary shares offered can be used as a proxy for the degree of information asymmetry. They find that issuing firms that are able to sell secondary shares have less information asymmetry and therefore experience less negative abnormal returns around expiration of the lockup. This is supported by the results that show a significant negative relationship between percentage of primary shares offered and CARs around the unlock day. Brau et al. (2004) also argue that there is a significant positive relationship between the post-IPO percentage of management ownership and CARs around expiry of the lockup. They suggest that manager with a large stake are less likely to sell shares after the lockup period expires and therefore the market reaction for those firms is less negative.

To explain the CARs around expiration of the lockup, Brau et al. (2004) also use size and percentage of shares subject to the lockup as explanatory variables in their model. Their results suggest the market reaction is less heavy for issuing firms that are bigger and lockup less shares. Like size, percentage of shares subject to lockup is a proxy for the degree of information asymmetry (Goergena et al., 2004). Firms that have a higher probability for agency costs to occur, have to lockup more shares to convince investors that they would not take advantage of them. If there is more information available about bigger firms and firms that lockup less shares, then there is less uncertainty and less market impact around expiration of the lockup period. Therefore, short lockups are expected to be followed by lower negative abnormal returns.

Studies of Espenlaub et al. (2001) and Hoque (2011) also find support for the commitment theory. Both studies show that issuing firms with absolute-date lockups experience higher information asymmetry. These firms are more in need to show their commitment by choosing to accept an expiration date that is fixed. The results of both studies show that absolute-date lockups are associated with more negative CARs around expiration. Again, if the commitment theory holds,

(14)

13

then one could argue that issuing firms with less information asymmetry and therefore have shorter lockups should be followed by lower negative CARs around expiry of the lockup period.

Stephens and Dussold (2007) examine lockup expirations for issuing firms that have multiple lockups in place. Data regarding the exact type of insider to which a lockup applies to is not well documented. By looking at small anecdotal evidence and interviewing investment bankers, they suggest that key insiders like executives and founders are more likely to be locked up in “later lockups”. Stephens and Dussold also compare the CARs of multiple lockups and find that the later lockup expiration is significant negative and most heavy. They argue that this finding is consistent with the theory that true insiders like executives and founders are locked up in “later lockups”.

3. Hypotheses & Methodology

3.1 Introduction

This section formulates the hypotheses and describes the methodology to test these hypotheses. The first part covers the development of expected relationships regarding the length of the lockup period. The second part describes the method to test these particular relationships.

3.2 Hypotheses Development

The aim of this study is to examine the exact role of the lockup agreement. As mentioned in the literature section, the problem with previous studies regarding the role of the lockup period is that the length of the lockup has never been related to a long term performance measure. In fact, this is crucial to test the relationship between the lockup period and the quality of a firm and thereby determine why lockups exist. The development of the hypotheses below is based on the commitment theory, which suggests that the lockup period exists to reduce moral hazard problems. Jensen and Meckling (1976) defined the consequence for firms that experience moral hazard problems as agency costs. These costs are carried by existing shareholders and reduce their wealth. In other words, firms with a higher probability for agency costs to occur will have a worse long term performance compared to firms with less agency problems. According to the commitment theory high quality firms are associated with less information asymmetry and therefore less agency costs. Brav and Gompers (2003) showed that these firms of high quality need a shorter bonding commitment to convince investors, which means that these high quality firms have shorter lockups. Then for the commitment theory to be true there should be a negative relationship between the

(15)

14

length of the lockup and long term performance of issuing firms. This results in the hypothesis stated below:

H1: Issuing firms with shorter lockups will have a better long term performance.

If the regression regarding long term performance shows a contrary result, then this would indicate that the signaling theory of lockups is true. This theory supported by Courteau (1995) and Brau et al. (2005) suggests that the probability of negative information to be revealed is less for firms of high quality. Therefore, these high quality firms can afford to accept longer lockups and thereby signal their quality to investors. Consequently, if the results show that longer lockups have a better long term performance, then we would reject the commitment theory and accept the signaling theory.

Another way to examine the role of the lockup can be performed by testing the relationship between the length of the lockup period and market reaction around expiry. According to the commitment theory, the lockup agreement serves as a bonding device to convince investors that their interest is aligned with that of the firm and that they will not take advantage of other investors. But when the lockup period expires, the bonding commitment is not effective anymore since insiders are able to sell their stake on this particular date. This will strengthen the incentive to take advantage of other investors and consequently increases the probability for investors to suffer from agency costs. One could argue that investors of firms which have a relative high probability for agency problems to occur will sell their stake around the expiration date of the lockup in order to prevent agency costs. The commitment theory suggests that issuing firms with longer lockups in place experience more information asymmetry and therefore on average have more agency problems. Studies of Brav and Gompers (2003), Brau et al. (2004) and Goergena et al. (2004) support this view since they all show that the market reaction is heavier for issuing firms that experience more information asymmetry. Then for the commitment theory to be true, it should hold that issuing firms with shorter lockups and therefore a lower probability for agency costs to occur, will experience a less heavier price reaction around expiration of the lockup period. This expectation is formulated by the second hypothesis below:

H2: Issuing firms with shorter lockups will have a less negative price reaction around expiration of the lockup period.

In order to find additional support for the commitment theory, the market reaction can be further investigated by comparing the market impact for each set of lockups for the group of issuing

(16)

15

firms with multiple lockups. As mentioned above, the probability for agency problems to occur increases on the expiry date of the lockup period since insiders are able to sell their stake. This will strengthen the incentive to take advantage of other investors. The ability to take advantage of other investors differs per type of insider. Insiders like key executives and founders are most likely to be the cause of agency problems since they are responsible for the daily operations and can easily take self interested actions. Consequently, the market should react heavier on the release of those types of insiders because the probability for agency costs to occur is relatively high for this group. Stephens and Dussold (2007) suggest that key insiders like executives and founders are more likely to be locked up in “later lockups”. Assuming that this theory regarding the distribution of insider types for multiple lockups holds, then according to the commitment theory the analysis should also show that later lockup expirations are followed by more negative abnormal returns for the group of issuing firms with multiple lockups. Hypothesis three formulates this expected relationship regarding the difference in market reaction for multiple lockups:

H3: The market reaction around expiry is more negative for lockups that apply to management shareholders and therefore is more negative for later lockups for the group of issuing firms that have multiple lockups in place.

Finally, for the commitment theory to be true there should be a relationship between the length of the lockup period and the degree of underpricing. Existing literature shows that underpricing is mainly caused by information asymmetries that exist between the issuer and investors or underwriters. Brav and Gompers (2003) suggest that lower quality firms experience more information asymmetries. Their results also show that firms of low quality on average have longer lockups. Then based on the commitment role of lockups, issuing firms with longer lockups should have a higher degree of underpricing. The final hypothesis below predicts this positive relationship between the length of the lockup period and underpricing:

H4: There will be a positive relationship between underpricing and the length of the lockup period. Table one below summarizes the developed hypotheses to test whether the commitment role of lockup periods is correct. As the next paragraph will describe, hypothesis one, two and four will be tested using a regression-based analysis. The third hypothesis regarding the price reaction for the subgroup with multiple lockups will be tested using a comparison of means.

(17)

16

3.2 Methodology

Long term performance versus length of lockup

This study will measure the long term performance of issuing firms using a buy-and-hold return approach, which is in line with existing literature. Although early studies regarding post-IPO performance often used cumulative abnormal returns to measure the long term performance, Barber and Lyon (1997) show that this methodology could give biased results and moreover find support for the use of buy-and-hold abnormal returns (BHRs) to correctly measure risk-adjusted long term performance. These biased results are among others caused by the fact that CARs do not take into account for monthly compounding. Consequently, a sample consisting only of firms with 12 month BHARs equal to zero could still have a CAR that is greater if the return of the issuing firm is more volatile than that of the benchmark. Also Kothari and Warner (2006) support the use of BHARs since these give a more representative view of the actual investment experience of investors.

Three and one year BHARs are calculated to test the relationship between lockup length and long term performance. The BHARs are calculated as the buy-and-hold return on the issuing firm minus the buy-and-hold return on an appropriate reference asset or portfolio. A normal index like the S&P 500 is not appropriate since firms within this index are much older and have different firm characteristics than recently publicly offered firms. Similarly to studies of Ritter (1991), Barber and Lyon (1997) and Mitchell and Stafford (2000), reference portfolios will be calculated based on size and book-to-market ratios to determine the appropriate expected return that is needed to calculate

Table 1

Summary of predicted relationships based on the commitment role of lockups

Dependent Variable Variable of Interest Expected Sign Coefficïent Hypothesis

Long Term Performance Lockup Length Negative Issuing firms with shorter lockups have

a better long term performance.

Price Reaction around Expiry Lockup Lockup Length Negative

Issuing firms with shorter lockups have a less negative market reaction around expiry of the lockup.

Price Reaction around Expiry for group

Multiple Lockups Lockup Length Negative

The market reaction is more negative for later lockups for the group of issuing firms that have multiple lockups in place.

Under Pricing Lockup Length Positive

There will be a positive relationship between under pricing and lockup length.

(18)

17 𝐵𝐻𝐴𝑅𝑖 = (1 + 𝑅𝑖,𝑡 𝑇 𝑡=1 ) − (1 + 𝑅𝑏𝑒𝑛𝑐ℎ𝑚𝑎𝑟𝑘,𝑡 𝑇 𝑡=1 )

the BHAR. The reference portfolios will consist of all NYSE, AMEX and NASDAQ firms. First, the reference firms will be matched into five size portfolios. Size will be measured as shares outstanding times the stock price at the end of June each year. The size breakpoints are the quintiles for NYSE stocks, which means that the AMEX and NASDAQ stocks will be matched into reference portfolios based on these breakpoints. The distribution based on size will be skewed since NYSE listed firms on average are bigger than AMEX and NASDAQ listed firms. In other words, most reference firms will be matched into the first en second quintile portfolios. Then within each size based portfolio, reference firms will be matched into five book-to-market based portfolios. The book-to-market ratio breakpoints at the end of June each year will be computed using the book value of equity at fiscal yearend in T-1 and the market value of equity as shares outstanding times the stock price at the end of December in T-1. As a result, there will be 25 different reference portfolios to which a sample firm can be matched according to size and the book-to-market ratio. After matching each sample firm to an appropriate reference portfolio, the BHARs can be computed using the following formula:

(1)

Where Ri,t is the daily return for sample firms and Rbenchmark,t is the daily return for the reference portfolios. Both are adjusted for corporate actions.

Subsequently, these BHARs will be used as the dependent variable in the regression to test the relationship between long term performance and lockup length. As mentioned in the literature review, there are some determinants of long term performance that should be incorporated in the model. Studies of Ritter (1991), Schultz (1993), Hensler et al. (2004) and Demers and Joos (2007) include age of the firm at the IPO as a measure for risk and degree of information asymmetry in their models. They all find a positive relationship between age at offering and long term performance.

Additionally, the model should also control for size as Schultz (1993), Brav et al. (2000) and Hensler et al. (2004) show that larger companies are more likely to survive. These studies all measure size as the market capitalization at the offering. Besides market capitalization, I include total assets at the IPO in the regression model as a second measure to control for firm size and expect for both a positive relationship based on existing literature.

Based on Coakley et al. (2007), I control for earnings using a variable that divides earnings before interest and taxes (EBIT) with total assets and thereby measure how well the issuing firm is performing. I follow the same methodology in measuring this ratio as used by Kurshed (1999) and Coakley et al. (2007); for both the EBIT and total assets the average will be determined for the year in which the firm goes public and three years prior to that year.

(19)

18

Another control variable that will be included is underpricing, notwithstanding the fact that existing literature is not clear about the direct relationship between initial returns and long term performance. Hamza and Kooli (2010) find, contrary to Schultz (1993) and Hensler et al. (2004), a negative relationship between underpricing and long term performance. This result is in line with the previous mentioned studies which show that information asymmetries are the reason for underpricing. According to the commitment role of lockups, low quality firms have a relatively high degree of information asymmetry and thereby a higher probability for agency problems to occur. Subsequently, this results in a worse long term performance. Therefore, a negative relationship between initial returns and long term performance is expected.

The model should also correct for venture capitalists as insiders since Brav and Gompers (1997) document an outperformance for VC backed IPOs compared to their non-VC backed counterparts. However, Kraus and Burghof (2003) suggest that the expiration of the lockup period is a breakpoint in time with respect to the relative performance of VC backed IPOs. They show that issuing firms backed by VC perform better before, but worse after lockup expiry. Although it is hard to formulate the expected sign of the relationship, based on Kraus and Burghof (2003) one could argue that the coefficient of the VC dummy should be greater when the one year BHAR rather than the three year BHAR is used as the dependent variable.

Finally, year and industry fixed effects should be incorporated in the regression model. According to studies of Ritter (1991), Loughran and Ritter (2004), Demers and Joos (2007) and Coakley et al. (2007) there is a relative large number of IPOs for periods with excessive underpricing because of favorable market conditions. This attracts relative young firms and firms of low quality, which results in an underperformance for the group of firms that issue in these particular periods. Year dummies are created to control for these effects. In addition, industry dummies will be included to control for industry fixed effects as Ritter (1991) showed that certain industries outperform in the long run. The complete regression to test the first hypothesis regarding the relationship between long term performance and lockup length and thereby determine the role of the lockup, is formulated as follows:

(2) 𝐵𝐻𝐴𝑅 = 𝛼 + 𝛽1𝐿𝑜𝑐𝑘𝑢𝑝 𝐿𝑒𝑛𝑔𝑡ℎ + 𝛽2𝐿𝑛 𝑀𝑘𝑡 𝐶𝑎𝑝 + 𝛽3𝐿𝑛 𝐴𝑠𝑠𝑒𝑡𝑠 + 𝛽4𝐸𝐵𝐼𝑇/𝐴𝑠𝑠𝑒𝑡𝑠

(20)

19

Price reaction versus length of lockup

The price reaction around expiry of the lockup will be measured in an event study using cumulative abnormal returns (CARs), which is in line with existing literature. Subtracting the predicted return from the actual return results in the abnormal return. The calculation of predicted returns is based on the market model stated below:

(3)

Where Ri,t and RMKT respectively denote the daily returns on issuing firm i and the equal weighted benchmark. The market model parameters will be measured using an estimation period that ends 11 days before the expiry date and has a maximum length of 200 days. Although a longer estimation period would give better estimated parameters, the length of the estimation period is limited by the length of the lockup and is therefore relatively short. The ARs are measured over a 21-day period. Three-day and five-day CARs surrounding the expiry date will be used for further cross sectional analysis regarding the relationship between market reaction and lockup length. The abnormal returns within the 21-day event window will be calculated by taking the difference between the actual return and predicted return:

(4)

Then CARs are calculated by simple accumulating the daily abnormal returns for issuing firm i:

(5)

Additionally, to examine the average market reaction for the whole group of issuing firms and check whether this yields similar results as existing literature, average abnormal returns (AARs) for each day within the event period and cumulative average abnormal returns (CAAR) need to be calculated.

An OLS regression model with the CAR as the dependent variable is used to give insight into the relationship between the market reaction around expiry and lockup length. Hypothesis two states that issuing firms with shorter lockups will experience less negative abnormal returns around

𝑅𝑖,𝑡 = 𝛼𝑖+ 𝛽𝑖𝑅𝑀𝐾𝑇,𝑡+ 𝜀𝑖,𝑡

𝐴𝑅_𝑖,𝑡= 𝑅_𝑖,𝑡 − 𝛼_𝑖− 𝛽_𝑖𝑅_{𝑀𝐾𝑇,𝑡}

𝐶𝐴𝑅𝑖,𝑡 = 𝐴𝑅𝑖,𝑡 𝑇

(21)

20

expiry of the lockup period, which as discussed would give support for the commitment role of lockups.

The model should also control for other effects that are discussed in the literature part of this study. Existing literature shows that there is a negative relationship between the degree of information asymmetry and price reaction around expiry of the lockup. Brav and Gompers (2003) and Brau et al. (2004) demonstrate that issuing firms that are able to sell secondary shares during the offering have less information asymmetry and therefore experience less negative abnormal returns around the expiration date.

The latter study also shows that the market reaction is less heavy for issuing firms that are bigger and lockup less shares. Again, these variables are related to the degree of information asymmetry (Goergena et al., 2004). Hence, these effects should be incorporated in the model to explain the CARs at lockup expiry. Similar to the second equation, two measures of size will be incorporated in the model; the market capitalization and total assets at the offering.

Field and Hanka (2002) suggest that insider sales around expiry of the lockup have a negative effect towards the market reaction in two different ways; they carry bad news and increase supply. A dummy variable that equals one if there were insider sales one week before and after expiry will be included to control for this effect. This two week surrounding period is chosen because of the fact that insider fillings could be delayed by a few days.

The last control variable that will be included is a dummy that equals one if the issuing firm is backed by venture capitalists. Multiple studies show that there is a highly significant negative relationship between VC and the market reaction. Brau et al. (2004) argues that these venture capitalists probably exit their investment directly after expiration of the lockup and show that the market expects venture capitalists to exit at that moment.

Finally, I add fixed effects by generating year and industry dummies to control for potential omitted variables. The complete regression model is formulated as follows:

(6)

A simple comparison of means method is used to test hypothesis three by analyzing the differences in market reaction per type of insider and for earlier and later lockups for the group of issuing firms that have multiple lockup agreements and thereby examine whether the commitment theory of lockups holds. Then the following hypothesis will be tested:

H0: µ1 - µ2 = 0 and H1: µ1 - µ2 ≠ 0

+𝛽₅𝑉𝐶 + 𝛽₆𝐼𝑛𝑠𝑖𝑑𝑒𝑟 𝑆𝑎𝑙𝑒𝑠 + 𝐹𝑖𝑥𝑒𝑑 𝐸𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜀_𝑖

(22)

21

Where µ1 and µ2 are the mean CAR for the different subgroups. Then a t-statistic will be calculated to

test whether the results are significant:

(7)

Underpricing versus length of lockup

Finally, hypothesis four will be tested using a regression model with underpricing as the left-hand side variable and lockup length as the variable of interest to find additional support for the commitment theory. Existing literature uses two measures of underpricing; first day return or market-adjusted first day return. The latter is more appropriate since it takes into account the market return and therefore is a measure of excess return that can measure the extent of underpricing:

(8)

Where Pi,t denotes the first closing price of issuing firm i, Oi the offering price of issuing firm i and Mt and Mt,0 the corresponding equal weighted market closing prices. The relationship between lockup length and underpricing will be analyzed using a regression model. Following existing literature there are certain determinants of underpricing for which the model should correct. Age of the issuing firm is a proxy for the degree of information asymmetry. This variable should be incorporated into the model since underpricing is a direct consequence of information asymmetries between the issuer and investors or underwriters.

Another variable that will be included is size. I will use two different measures of size; IPO size measured by proceeds and size of the firm by pre-IPO total assets. Existing literature also shows that venture capitalists experience significantly higher initial returns. Therefore, a dummy variable will be included that equals one if the issuing firm is backed by venture capitalists.

Finally, I create year dummies since Loughran and Ritter (2004) show that IPO underpricing changed over time. Additionally, I add industry fixed effects by creating dummies to control for variables that differ between industries but are constant over time and therefore eliminate a potential omitted variable bias. The complete regression model is formulated below:

(9) 𝑡 = 𝑥 1− 𝑥 2 − (𝜇1− 𝜇2) 𝑆12 𝑛₁+𝑆2 2 𝑛₂ 𝑀𝑘𝑡 𝑎𝑑𝑗 𝐼𝑛𝑖𝑡𝑖𝑎𝑙 𝑅𝑒𝑡𝑢𝑟𝑛𝑖= (𝑃_𝑖,𝑡− 𝑂_𝑖) 𝑂_𝑖 − (𝑀_𝑡− 𝑀_𝑡,0) 𝑀_𝑡,0 ∗ 100 𝐼𝑅 = 𝛼 + 𝛽1𝐿𝑜𝑐𝑘𝑢𝑝 𝐿𝑒𝑛𝑔𝑡ℎ + 𝛽2𝐿𝑛 𝐴𝑠𝑠𝑒𝑡𝑠 + 𝛽3𝐿𝑛 𝑃𝑟𝑜𝑐𝑒𝑒𝑑𝑠 + 𝛽4𝐴𝑔𝑒 +𝛽5𝑉𝐶 + 𝐹𝑖𝑥𝑒𝑑 𝐸𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜀𝑖

(23)

22

4. Data

4.1 Introduction

This section will discuss the data that is used for the examination of the role of the lockup. First, the data collection will be discussed. Subsequently, some descriptive statistics will be provided to describe the dataset.

4.1 Sample Selection

The main source of the dataset is Thompson One, which provides information regarding IPOs. The sample consists of IPOs in the United States for a ten year period starting in 1996. Thompson One reports 4089 IPOs in the period 1996-2005 which have a lockup agreement in place. Excluding observations that are not common stock, concern close-end funds, REITs, acquisition companies, trusts, unit issues or of which a company identifier is missing (which is necessary for further data collection), results in preliminary sample of 2643 IPOs. Table two gives an overview of the sample selection process.

The data from Thompson One is limited since approximately 40-50 percent of the information regarding the lockup agreements within this preliminary sample is missing. I looked up more than 400 prospectuses, available through the EDGAR database maintained by the Securities and Exchange Commission (SEC), to complete the dataset for the years 2001 until 2005. Lockup information is documented in the “Shares Eligible for Future Sale” section of the prospectus and states among others how many shares are restricted and for which period.1 Even if there is information available regarding the length of the lockup, the prospectus has to be looked up when the percentage of shares that are subject to the lockup is missing. This variable is needed to make sure that the dataset contains reliable observations. For a lot of IPOs that have a 180 day lockup period and for which Thompson doesn’t specify how many shares apply to it, the prospectus seems to be in a standard format that contains the standard lockup length of 180 days. Because it cannot be verified that these lockups are correctly documented and additionally because the percentage of shares locked up is an independent variable in the regressions, I delete IPOs for which there is no information available regarding the amount of shares locked up. Finally, some observations are lost because of missing identifiers or erroneous data. The final sample consists of 1046 observations.

Determining the exact type of insider that is subject to a certain lockup would be very valuable in distinguishing between volume and information asymmetry effects regarding the market

(24)

23

reaction around expiry. Thompson One identifies the variable type of lockup using the codes “company lockup”, “management lockup” and “selling shareholder lockup”. Unfortunately, Thompson One does not well document this variable since it uses “company lockup” for almost all observations. I check whether the documented type of lockups by Thompson One are erroneous by looking up the prospectus, but indeed find that most prospectuses do not well specify the exact type of insider. In most cases, the prospectus states that the “directors, executive officers and other shareholders” are subject to the lockup agreement2. This even holds for multiple lockups, where one could expect that the different lockups apply to different kind of insiders. For this group I am only able to specify the exact type of insider for 29 out of the 86 multiple lockups.

Thompson One also provides almost all other independent variables that are needed as input for the regressions. Missing data is completed using Compustat and The Center for Research in Security Prices (CRSP). For 56 IPOs the total assets before the offering were not available through Thompson One and therefore had to be collected through Compustat as the total assets for the end of the quarter in which the IPO took place. The same method is used for the 52 missing book values of equity that are needed to calculate the book-to-market ratio which will be used to match sample firms to a reference portfolio. In addition, CRSP provides the missing first day returns. Corresponding CRSP equal weighted index levels are obtained to compute the corresponding market returns that are needed to determine the market adjusted first day returns. Finally, I had to hand collect 385 incorporation dates to define the age of the company at the IPO.

2_{Likewise, the prospectus stated in appendix 1 does not specify the exact insider type.} Table 2

Summary of sample selection process

Total number of IPOs with lockup agreement in the US for the period 1996-2005 4089

Less: IPOs that are not common stock or concern acquisition companies,

close-end funds, REITs, trusts and unit issues. 1422 Less: Issuing firms for which a company identifier (i.e. ticker or permno) is

missing or have erroreneous observations 32 Less: IPOs for which there is no information available regarding the amount of

shares that is subject to the lockup agreement 1589

(25)

24

The BHAR that is used as the dependent variable into the first regression, is computed using data from CRSP. I gather daily returns to compute the one and three year BHRs for each issuing firm in the sample. The size and market-to-book breakpoints to construct 25 reference portfolios are obtained from the website of French.3 Also the corresponding reference portfolio returns are obtained from the IPO dataset maintained by French.

Eventus, available through Wharton Research Data Services (WRDS), is used to examine the market reaction around expiration of the lockup. Eventus calculates the CARs for the whole sample and for each IPO in the sample which will be used for further cross sectional purposes. Eventus uses daily returns from the CRSP database to compute the CARs.

4.2 Sample Characteristics

Table three provides a description of the distribution of lockup lengths over the years. The decreasing yearly reported average lockup lengths and standard deviations support the finding of existing literature that the period of locking up shares declined over the years to a standard level of 180 days. Keasler (2001) mentions that the increased competition in the investment banking industry is the reason for this convergence towards a standard level. About 80 percent of the sample concern lockups of 180 days, which again is in line with existing literature. The sample of 1046 lockups consists of 92 lockups that are less than 180 days and 126 lockups that are longer than the standard lockup length. Table two also shows that IPO activity during the internet bubble significantly decreased.

Table four presents descriptive statistics for the full sample (panel A) and subgroups (panel B, C and D) based on the length of the lockup. Panel A shows that the average market-adjusted initial return is 24.70 percent, which is somewhat higher than Loughran and Ritter (2004) report. These relatively high initial returns are most likely boosted by the IPOs in the years 1999 until 2001 which have excessive high degrees of underpricing. The average age of issuing firms is 13.37 years and more than 50 percent is backed by venture capitalists. Panel B, C and D show some differences of IPO characteristics for the subgroups short lockups (<180), long lockups (>180) and lockups with the standard lockup length of 180 days. The average proceeds and assets are less for the group with longer lockups and are in line with for instance Brav and Gompers (2003) who mention that size and lockup length are negatively related. Another variable that measures information asymmetry is age. The average age at the offering for longer lockups is lower than for the subgroup with shorter lockups. Once more, this is in line with Brav and Gompers (2003) who argue that firms with a higher

(26)

25

degree of information asymmetry, on average have longer lockups. By comparing the group with long and short lockups, one could notice that the percentage of IPOs that is backed by venture capitalists is twice the size for the group with short lockups. Bradley et al. (2001) find the same pattern in their study regarding VC and lockup expiration. Finally, the profitability measure EBIT/Assets is negative for all subgroups. The low ratio for the group of issuing firms with a short lockup period is remarkably low compared to group with long lockups in place and not in line with what I would expect.

Table four also presents the median for each variable, which is a better measure of distribution than the mean. The relative low median compared to the mean indicates that some variables have a skewed distribution to the right due to outliers. I control for these outliers by winsorizing these particular variables at the 99th percentile of their distributions. Only the ratio EBIT/Assets will be winsorized at both the 1th and 99th percentile. The winsorized variables are used into the regressions presented in section five, which test the expected relationships based on the commitment theory of lockups.

Table 3

Distribution lockup length over time

Distribution lockup lengths over the years. The sample consists of 1046 IPOs in the US for the period 1996-2005. The presented lockup lengths are measured in days. Panel A shows the differences in lockup length across the years and panel B shows these differences between two sub periods; 1996-2000 and 2001-2005.

Year N < 180 180 > 180 Mean Std Dev

Panel A: Distribution over years

1996 219 6 169 44 237.43 145.73 1997 113 8 83 22 223.67 131.56 1998 50 6 35 9 224.60 139.37 1999 187 21 153 13 183.02 57.68 2000 119 6 105 8 191.56 73.25 2001 29 2 24 3 192.59 62.86 2002 11 0 10 1 196.82 55.78 2003 43 3 35 5 196.74 76.71 2004 140 15 117 8 179.79 52.80 2005 135 25 97 13 174.64 58.41

Panel B: Distribution over subperiods

Full Sample (1996-2005) 1046 92 (8.8%) 828 (79.2%) 126 (12%) 201.22 100.78 Subperiod 1996-2000 688 47 (6.8%) 545 (79.2%) 96 (14%) 211.51 115.37 Subperiod 2001-2005 358 45 (12.6%) 283 (79%) 30 (8.4%) 181.44 59.32

The length of the lockup versus long term performance : an examination of the role of the lockup.