A study of underpricing of private equity and venture capital in Europe over the last two decades

A study of underpricing of private equity and venture capital in Europe over the

last two decades

Filippa Bakker, 10401075 Master’s thesis Quantitative Finance Supervisor: Dr. Rafael Perez Ribas

2

Statement of Originality

This document is written by Student Filippa Bakker who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

This paper investigates the impact of private equity and venture capital backed European IPOs on underpricing in the period 1999-2018. Different instruments and time periods are used to test the differences between these groups. The sample consists of IPOs, 5,475 IPOs in total of which 723 PE- or VC-backed. Underpricing is measured as the first day return from closing to the offer price. The results in this paper show that private equity and venture capital backed IPOs do have a lower level of underpricing compared to non-backed IPOs. Both smaller private equity IPOs as well as venture capital IPOs show a significant higher level of underpricing where the result for venture capital is even higher, consistent with younger venture capitalists invest in smaller firms and taking higher risks.

Table of Content

STATEMENT OF ORIGINALITY ... 2

ABSTRACT ... 3

1. INTRODUCTION ... 5

2. BACKGROUND INFORMATION ... 7

2.1REASONS OF GOING PUBLIC ... 7

2.2WHY ARE IPOS UNDERPRICED? ... 8

2.3UNDERPRICING OVER THE YEARS ... 8

2.4MIFID ... 10

3. LITERATURE REVIEW ... 12

3.1UNDERPRICING OF PE-BACKED IPOS ... 12

3.2UNDERPRICING OF VENTURE CAPITAL BACKED IPOS ... 13

3.3POSITIVE RELATION SPONSOR BACKED AND UNDERPRICING ... 13

3.4DIFFERENCES BETWEEN PRIVATE EQUITY AND VENTURE CAPITAL ... 14

3.5DIFFERENCES WITHIN INDUSTRIES, COUNTRIES AND YEARS ... 15

3.6PRIVATE EQUITY BACKED AND OWNERSHIP REDUCTION ... 16

4. DATA ... 18 4.1SUMMARY STATISTICS ... 18 4.2INDEPENDENT VARIABLES ... 19 4. EMPIRICAL METHOD ... 22 5. RESULTS ... 25 6. ROBUSTNESS CHECKS ... 42 6.1CRISIS PERIOD... 42 6.2TECH INDUSTRY ... 43 7. DISCUSSION ... 44 8. CONCLUSION ... 44 9. REFERENCES ... 46 10. APPENDIX ... 50

1. Introduction

The most important issue as well as the biggest challenge in an initial public offering (IPO) is the pricing of the stock. It has been found that IPOs often are underpriced and companies thus hand in some of their true value on the day of going public. Why would companies accept these costs? Venture backed companies the other way, have sometimes an incentive to underprice more, because they want to interest the investors and make the company attractive when they want to grow a lot and need more capital. The debate in the IPO literature which is there for many years and is not solved nowadays, is whether this

underpricing gets higher and is driven by agency problems between issuers and underwriters. It is difficult for the issuer to value its company right. Underwriters use bookbuilding to set the offer price after indications of interest from institutional investors. Because of several deals and multiple bookbuilding, it is not surprising that underwriters have a big advantage here and allocate the shares tactful, so that a high level of underpricing occurs (Ritter, 2011). This problem of asymmetric information escalated and in order to stimulate fair and

transparent markets that are efficient and integrated, the MiFID is introduced in November 2007.

In general, I investigate whether PE- and VC-backed IPOs are less underpriced than non-backed IPOs. I also analyze the impact of the implementation of MiFID for the

underpricing of PE- and VC- backed IPOs, by a DID model. MiFID is introduced in order to make the market more transparent for the investors and thereby decrease information asymmetry. Furthermore, the rules for the allocation of shares and prospectus are stricter, which should lead to a more stable market with less extreme levels of underpricing. The guideline protects investors and take care of the integrity of financial markets. It is implemented in 28 countries in Europe and to see whether this guideline has helped to harmonize the European stock exchange, it is interesting to investigate the effect of MiFID I and the interaction with sponsor-backed IPOs. This is done by a Difference-in-Difference (DID) model. This DID model is also regressed on the interaction between sponsor-backed IPOs and the crisis and bubble years. Then I look to the effect of small- mid- and large sized IPOs on underpricing of PE- and VC-backed companies. PE-investors mostly invest in larger IPOs and focus on a big deal. Therefore, they could reduce the level of underpricing. Further, I analyze

6 the importance of the stake that PE-investors have in the company going public. A higher stake indicates a high quality IPO.

Companies need money in order to grow and they can finance their business in different ways. A company can borrow money from a bank, issue securities in the public market or get money from private investors. Because an IPO could raise a lot of capital, despite of the costs, companies still easily chooses for this option. However, in recent years, more and more firms are bought by private equity. After the crisis years, private equity investors started to invest in and go public with companies. Private equity buys firms with money that largely comes from institutional investors. Because of the low returns on bonds and the high valuations of shares, they put more and more money into alternative investment strategies like these. PE investors then have something to say and could vote for a company going public. They are highly sophisticated and professional, and try to sell the companies for their best price in order to get the highest proceeds. Therefore, more IPO’s are PE funded in the recent years and it is may likely that these IPO’s are less underpriced than non-backed IPO’s. The reason for this is that private equity investors better know the real price of the company. Furthermore, there where non-backed companies would accept the offer price given by the underwriters, PE investors are more sophisticated and aggressive in valuation of their stake (interview investment banker, ABN Amro).

As found in this thesis, the stake of sponsor-backed IPOs is grown over the and the question is whether these PE funded IPO’s are less underpriced and there is less return for investors of the shares after the IPO. A sign for investors is the stake a PE firm has in de company. If the stake is smaller or almost zero after the company has been going public, it could induce that the PE investors could invest in other investments which give them higher returns, instead of keeping a stake in the company. Investors should keep an eye on this and could opted for better investments with help of the results of this thesis.

In section 2, background information about IPO and the MiFID I regulation is given. Section 3 consists of a literature review of private equity and venture capital in special. Then section 4 shows the data used with some descriptive statistics. Section 5 shows the results and the last sections consist of robustness checks, discussion and conclusion. In the end, references and appendix are included.

2. Background information

Underpricing of newly issued shares has been given much attention in the literature about initial public offerings. It seems that new issues are underpriced and many researchers have tried to look at the story behind this underpricing. In recent years, more IPOs are done by PE investors and it seems that they are less underpriced. This section reviews different theories and literature behind underpricing and explains how PE- and VC-backed companies could be underpriced less.

2.1 Reasons of going public

When companies are in an early stage of their life cycle, they can raise equity capital with some help of investors. More capital is needed to finance future investments. The solution for getting this additional capital can be by going public. Ljungqvist (2004) find that providing public capital the main reason of an initial public offering is. Furthermore, going public enhances the liquidity. Beside of gaining capital, Ljungqvist (2004) states that going public gives indirect benefits to the company because it sets the company in the spotlights. The company becomes interesting and can attract a higher level of managers. Furthermore, the increased equity capital reduces leverage of the firm, which makes it safer to borrow money. The transparency of the company will increase after going public and this makes it easier to find potential investors. The reduced leverage together with the higher level of transparency can increase competition among lenders. Zingales (1995) mentions that this will lower the cost of capital and thus increase the firm value, which gives another benefit of going public. However, the transparency also brings us to a less positive point. By the greater transparency, all the competitors can see the information and take advantage of it. Therefore, it is essential that there is a careful policy with regard to the disclosure of information (Ritter, 1998). On contrary concerning transparency, the manager knows which project will increase the value of the company and after going public, he can no longer decide whether to invest. When the ownership is kept private, the manager would keep the power to make these decisions (Boot et al., 2006). Brau & Fawcett (2006) share this idea that keeping control in the decision-making is the most important reason to stay private.

When going public, shares are more liquid and thus can be sold faster in the market. Where transparency could be positive and negative about going public, according to literature

8 gaining additional capital, lower cost of capital and higher liquidity are the most important motivations of going public.

2.2 Why are IPOs underpriced?

According to different studies, asymmetric information is one of the main reasons an IPO could be underpriced. This asymmetric information theory simply induces that one of the parties involved by an IPO has additional information, which is unknown by the other parties. The three main parties involved are the issuer, the underwriter and the investors.

Rock (1986) is one of these authors who states that underpricing is caused by

asymmetric information in his research into the reason of underpricing of IPOs. His winner’s curse model is a well-known asymmetric information model. He says that a part of the investors is uninformed, while other investors are perfectly informed about the true value of IPO. The underpriced IPOs are more attractive for investors, because buying these shares will generate more profit. This profit comes from the initial return of the IPO. Therefore, the better-informed investors will bid only on good priced IPOs, while the uninformed investors just bid randomly on good and bad ones. The less informed investors will then only get a fraction of the most profit-making shares and the major part of the least desirable shares. They are confronted with the winner’s curse (Rock, 1986). This major part consists of overpriced shares. The average expected return for uninformed investors will therefore be negative and these investors have no reason to stay in the market. In order to keep the uninformed investors, their average expected return should be positive. This could be created by deliberated underpricing. Then, the uninformed investors will purchase the issue. So, on average, IPOs must be underpriced to compensate the uninformed investors for purchasing shares of the lower valued companies (Ljungqvist, 2007). This winner’s curse of Rock (1986) explains why IPOs are underpriced and Chambers and Dimsons (2009) findings about the increased underpricing over time.

2.3 Underpricing over the years

Studies over several years back show that the level of underpricing has changed during the years. Since long ago, underpricing is measured as the first day returns and are showed as a

percentage change from the offering price to the closing price on the first trading day after the company went public. When the returns are positive, the IPOs are called underpriced.

Chambers and Dimson (2009) divide samples in three subgroups. The first before WWII, then a period post-WWII and post-Big Bang from 1986 till 2007. They find that the underpricing of IPOs was only 3,8% on average in the period from 1917 till 1945. The period after, from 1946 till 1986, underpricing grew to 9,2%. This is a major growth and it kept growing in level of underpricing, especially directly after the UK stock market deregulated in 1986. In the last period, they even measure an underpricing of 19,0%.

Underpriced thus has changed significantly over time and many factors causing these changes. As to begin from the book building that started in the 1990s. First, both auctioning and book building were used to price IPOs. Then from end 1990s, book building increased and took over auctioning (Degeorge et al., 2007). In the end of these years the internet bubble occurred and the level of underpricing was outrageous high. As found by Lowry et al., (2002), the fluctuating volatility of shares had not been so high in the years before. There have been diverse hot issues periods before the bubble of 1999 and 2000, but most of the factors being unusual in these bubble, did not occur in the earlier hot issues episodes. They study the difference in auctioning and book building and find a significant difference in IPOs placed by auctioning compared to book building. This result is a lower level of volatility in the initial returns for IPOs placed in auction format. Book building thus can partly explain the high volatility in returns in the bubble years, at least a fraction of it. Furthermore, the arise of phones, computers and internet increases the liquidity over these years. This leads to

investing in shares has become easier. The internet bubble starts from 1997. In 1999, private equity was able to keep the underpricing clearly lower compared to non-backed and the year 2000. When the top was reached in 2000 this was no longer possible and also PE- and VC-backed companies suffered from high levels of underpricing. It could be that the sponsor investors of the issuing firm wanted to go public in these years and held a relatively high stake in the company. They knew the IPO market reached phenomenal levels and could sell their stake after having them bid up, before the bubble burst (Chamber and Dimsons, 2009).

Underwriters take care of a stock demand that is high enough. They define the price and they see how many times the book is overwritten. Ljungqvist and Wilhelm (2003) argue that there was a desire for underpricing in these years by the underwriters. Because they have more information due to the book building mechanism and sees whether there is excess

10 demand (Lowry and Schwert, 2004), underwriters have an evidence on the investors. They used their book building knowledge and the increased liquidity during the bubble. They were reluctant in raising the price of the intended IPO when noting excess demand. Allocating underpriced IPOs directly to friends or family got priority over maximizing proceeds for the issuer (Lowry et al., 2002). Even more, Lowry et al. (2002) find that IPOs where the price was revised by a higher amount resulted in an even higher volatility, and thus initial return. This was not desirable to investors and also PE investors had to suffer from this. PE investors want to reduce the information asymmetry in order to get the best price for the company.

However, they could not do anything to this proposed price by underwriters. Therefore, PE-backed IPOs also suffered from extreme levels of underpricing during the bubble and left a lot of proceeds at the day their company went public. The level of underpricing increased with the extent the book is overwritten. This is because of the allocation that investors get

distributed, which is not their total demand. Investors, mainly institutional, will place an order directly after the IPO has taken place and where the demand of the shares is then higher than the offer, the price of the share will go up directly after a company went public. Underwriters used this advantage during bubble years and after, and allocated shares in their wish. This adverse selection problem leads to extremely underpricing which is not desirable by the issuer (Ritter, 2011). According to Michaely and Shaw (1994), retail investors face an allocation bias here too. Underwriters want to allocate shares in favor to institutional investors. Because these investors have superior information, they bid higher prices and get the best IPO deals.

In order to reduce this winner’s curse problem (Rock, 1986) and prevent for higher levels of underpricing, something had to change in the trading market. Adverse selection and the dishonesty was an enormous concern and therefore the trading rules are strengthened. MiFID I is introduced in 2005 and implemented in November 2007.

2.4 MiFID

The trading rules were thus not strict and underwriters could use advantages in their own favor. They could freely choose the way of allocations after the book building process, could inform investors and set up a price without any control. Initial returns of IPOs can therefore rise enormously. MiFID presents a regulatory that takes a large step toward a capital market that is efficient and competitive. The main objective is to create an environments of market

quality and enhanced transparency. High quality reporting should enhance integrity and strengthens investor protection. Investment firms are required to implement best execution, which not only means the firm has adopted the policy following criteria, but also their IT systems in order to correctly execute orders to what is written in the requirements. This also include trying to limit costs for clients and time taken to complete a transaction. The level of disclosure of information to retail investors increases and there is evidence of enhanced gradual protection of the European Capital Markets. Further, the transaction costs are

lowered through increased transparency trading procedures. With these general but effective improvements, the MiFID represents a revolution in European securities markets that is likely to lead to long-lasting structural changes (Casey and Lannoo, 2008). Because of lower trading costs and more information transparency both pre- and post-trade, this should create a widened financial market. The regulatory requires to operate all under the same regulatory within Europe, which would reduce the complexity and confusion regarding different

mechanisms regarding manipulation and financial instruments. It is interesting to see whether these strengthened rules do have an impact on underpricing.

Cumming et al. (2015) investigate the effect on liquidity in the EU after

implementation of the MiFID regulation. The rules of the MiFID should enhance trading from countries outside Europe and with this increases liquidity within the EU. They find that the promulgation of the MiFID in 2005 resulted in a positive and significant shift of trade to the EU. The volume of trading in EU thus increased in 2005. However, the effect of the

implementation of the MiFID in November 2007, do not show a significant effect on trading. The effect on the announcement is greater and the market responded here more than to the actual implementation. The economic significance of the increased exchange market share is about 51% (Cumming et al., 2015). The promulgation effect implies that the strengthening of stock exchange rules empowers the EU to increase the volume of trading. An enhanced liquidity should lead to less underpricing. It is not yet tested whether the actual return on IPOs has decreased after implementation. This thesis tests this gap in literature and seeks for an effect in the years after implementation of the MiFID I. The MiFID II regulation is

implemented on 2 January 2018 and the rules build on the regulatory of MiFID I. After introduction in 2016 and implementation, the results are may even increasing with the expectation that it lowers asymmetry and therefore underpricing of IPOs.

12

3. Literature Review

3.1 Underpricing of PE-backed IPOs

An explanation why PE-investors can underprice lower compared to non-backed firms, is because they have less adverse selection problems. They can signal quality to the market about the company going public (Feretti and Meles, 2011). Because they have built up a strong reputation over the years, their signal strength is strong too. Benveniste and Spindt (1989) study the effect of adverse selection and define the choice of a contract when going public in terms of a, the value of investors’ information. A high-a firm is more adversely affected by a firm-commitment and then have pressure from underwriters to presell. Firms do not want that and therefore, they choose best-effort contracts in order to limit the incentives of their underwriter to wholly presell issues. However, because of the bias selection in contract choice, firms with the best-effort contracts are more underpriced on average than firms with firm-commitment contracts, so these firms are not better off. The authors conclude here that companies with higher information asymmetry have more

underpricing on average. This is also the reason that PE backed firms can be less underpriced. PE-backed firms try to lower information asymmetry for the issuer. Of course, they do this for themselves in the end to get the highest proceeds. PE investors are risk averse and because the presales increase with risk aversion, PE backed firms thus choose to sell under firm-commitment contracts and get lower initial returns on average. Sherman and Titman (2002) agree with Benveniste and Spindt (1989) that information asymmetry affect the precision of the price and the initial return. The higher the asymmetry, the more difficult it is to value a company and the higher the level of underpricing. Since PE investors often have private information about the issuing company, they try to reduce this asymmetry to lower the underpricing. They can reduce this asymmetry more than non-backed IPOs. Therefore, the following hypothesis is formed:

Hypothesis 1:

PE-backed IPOs are less underpriced than non-backed IPOs

After the MiFID I is implemented however, the transparency of information for investors is met more. Investors have more information nowadays and firms do not have to be afraid for

underwriters’ incentive to presell, because rules about the prospectus are stricter and underwriters cannot choose allocation anymore. Therefore, there is less availability for PE-sponsors to reduce the information asymmetry. This could result in a smaller difference in underpricing between sponsor- and non-backed IPOs after the implementation of MiFID I, which is not yet tested in literature. I will look for a possible difference in sponsor- to non-backed IPOs after implementation of the MiFID I.

3.2 Underpricing of Venture Capital backed IPOs

The definition of venture capital argues that it is part of the broader private equity activities, which also include leverage buyouts (Cumming, 2013). Venture capitalists focus more on smaller sized and young firms (Feretti and Meles, 2011). Lee and Wahal (2004) agree with this and argue that young venture capital firms are willing to take smaller firms public.

Furthermore, they take higher risks in firms that go public compared to older venture capitalists. Older venture capitals do not need that because their imago is established. The older firms therefore do not have to take risk investments and are more able to define a right price and thus underprice less. Coakley et al. (2009) find that VC-firms were larger

underpriced in the bubble years but in the years after, these firms are less underpriced than non-backed firms. Venture capitalists can namely overcome agency problems and do not want to lose their reputation as good quality investors. This results in lower underpricing. The hypothesis therefore is:

Hypothesis 2:

VC-backed IPOs are less underpriced than non-backed IPOs

3.3 Positive relation sponsor backed and underpricing

There is some research done which tells us that PE backed IPO’s have a positive effect on underpricing. Lee and Wahal (2004) test the hypothesis that the first-day return of VC backed IPOs is higher than non-venture backed IPOs and they find a statistically significant and economically important difference. Outstanding are the results during the internet bubble in 1999-2000, where the difference between VC backed and non-venture backed 25.0

14 studies the effect of young venture capital firms which earlier go public than older firms. His grandstanding theory states that each additional IPO for young VC companies brings more capital than older firms. Older VC firms do not need to signal because they are established. Also, investors believe in their ability because they have observed them for several years. The incentive to raise new funds directly after the IPO is therefore smaller for them. Young venture capital firms accept higher underpricing in order to signal quality of their VC funds. It should help in raising more capital for their funds in the future (Lee and Wahal, 2004). If PE firms wants to grow and invest more in the future, they could use this strategy too. PE funded firms then accept higher underpricing with the idea of signaling quality to the market and attract investors.

3.4 Differences between private equity and venture capital

Although different explanations have been found for PE- and VC-backed IPOs with respect to underpricing, Coakley et al. (2009), Lee and Wahal (2004), and Ferreti and Meles (2011) agree that both PE- and VC-backed IPOs have a significant lower level of underpricing compared to non-sponsored IPOs. There are some differences however. Venture capitalists tend to take younger firms public and in general smaller firms than private equity investors. PE focus on large firms and invest all their money in one firm, trying to get the best offer price and thus proceeds, through operational efficiency. However, this study sees comparable roles in VC-and PE-backed IPOs. Both try to optimize price, signal quality of the firm VC-and their reputation by determining a right value for the company going public and underprice as little as possible. Since PE-investors focus on larger companies in general, and information asymmetry is less as the firm size is larger (Benveniste and Spindt, 1989), I come to the following hypothesis:

Hypothesis 3:

A large IPO has a higher the stake of PE-investors and a smaller the level of underpricing

It is not clear from literature how large the stake of younger or older venture capitalists is. Therefore I have no idea what their proportion of investments is in smaller of larger firms. What is known from above, is that younger venture capitalists take higher risks and invest in smaller firms. Further, it seems that older venture capitalists can price better and invest in

larger companies in general. To test whether the older venture capitalists indeed can price better, I test the next hypothesis:

Hypothesis 4:

The larger the VB-backed IPO, the lower level of underpricing

3.5 Differences within industries, countries and years

With level of underpricing that may differ with different contracts, there appear to be

differences in initial return within industries and control factors as well. Stivers (2003) studies return dispersion and explains idiosyncratic volatility. The results show that the dispersion of returns is correlated with firm-level and divergent returns are seen as reflecting uncertain stock markets. Specifically, defining the right value of a firm when going public is harder when the firm expose high uncertainty or risk. Lowry and Schwert (2010) also study volatility, the standard deviation in firms that go public and the standard deviation in market-wide conditions. They find that more asymmetry gives a considerably higher volatility of initial returns and find a positive relation between the mean and volatility of underpricing. The harder it is to value a company, the more volatile the initial return. An example are biotech companies that need money to invest and can, after years of making losses, suddenly come with a great invention. Small or young companies have more instability in initial return too. The opposite is when a lot of information is known and the information asymmetry is small. IPOs done in real estate can therefore have lower levels of underpricing. The rental prices and house prices are fixed and thus the value of the company can better be determined with lower volatility, especially compared to an IPO done in an uncertain industry. (Wong et all., 2011) find that real estate IPOs are indeed less underpriced. Because the asymmetry is already very low within this industry, it is harder for PE investors to lower the underpricing than it is for other industries. Both non-backed and PE- and VC-backed companies do have lot of information about the value of the company. Therefore, there is less information that sponsor backed companies can release more than the non-backed. This should result in a smaller difference in underpricing. Freybote et all. (2008) investigate property IPOs in Europe. There are large differences in rules and they specifically look to the change after the

introduction of REITs and the eastern market that lies behind. They find a lack of quality in the Eastern European countries that have a need in going public. The supply of the IPOs is a

16 problem and this could be solved by introducing sharper rules in Europe. The authors also find a significance lower level of underpricing for property industry compared to other industries (Freybote et all., 2008). In short, more uncertainty should lead to more volatility and less uncertainty to lower volatility in the level of underpricing (Sherman and Titman, 2002). Besides of that, price movements and thus the daily return after an IPO appears as well. This depends on the level of uncertainty about the economic state. Times with higher uncertainty about the economy, gives a higher volatility and thus higher movements in price (Stivers, 2003). Therefore, as reasoning as above, in bear and bull markets the prices can increasingly go up or down. For the European IPOs, the different markets bear and bull will be tested in this thesis with respect to underpricing of sponsor-backed IPOs and their interaction effect.

3.6 Private equity backed and ownership reduction

Private equity firms in Europe do not prefer an exit by an IPO. PE firms sell more often to public or another private equity firm than exiting via an IPO. Equity markets in the US are more developed than in Europe. The Europe market is less liquid, which make exit

opportunities harder (Schwienbacher, 2005). This could explain why PE firms in the US have more exits through IPOs compared to Europe and following Black and Gilson (1998) it is the reason why venture financing lags behind. Companies where investors exit through IPOs are generally associated with higher returns (Gomper, 1995). In the dataset used, only 2.9% of the PE backed IPOs did a full exit after the IPO. PE firms in Europe thus not fully exit in most of the cases, but there is an ownership reduction. This reduction can be small or large,

depending on the choice of the PE firm and the market demand for the stocks. The difference between the pre-IPO stake of the PE firm and the post-IPO stake is the ownership reduction. When the reduction is large, this could be a sign for the market that the PE investors have some better investment opportunities. Teoh et al. (1998b) study earnings management prior to IPOs. When high earnings are reported at the time of the offering, this leads to overvalue the newly shares. They state that when PE investors are in for a quick exit, they try to manipulate earnings where possible to maximize their exit value. In the case the investor keeps a sufficient stake in the company after the IPO, he presumably sees future potential in the company, thus expect rising share prices. Therefore, there will be fewer incentives to manipulate the numbers and increase proceeds. The strategy of the PE investors thus matters

and the reputation as well. The sponsors want to give a sign of good quality to investors too. To signal quality, Wong et all. (2011) use the degree of underpricing and the proportion of shares retained by the sponsor. They find a positive relation between the quality of the IPO and the sponsor ownership. Further, there appears to be a positive relationship between the fraction of shares by the PE investors and the level of underpricing. Therefore, it could be an indication of high quality when the PE investors have a higher stake. Quality is defined as the level of underpricing, where a low level of underpricing indicates good quality. Investors can choose to keep shares of that company based on this information. Especially when the PE investor is well known and thus very sophisticated, this could be a reliable sign the firm will still growing. Wong et all. (2001) tested this in the Asian market on only real estate IPOs. I want to test the effect of quality on ownership reduction over all industries. Having done that, I see whether the ownership reduction can give a signal of the quality of the European IPOs generalized to all industries. All above leads to the following hypothesis:

Hypothesis 5:

4. Data

The sample consists of observations from IPOs that took place in Europe, in the period from January 1991 till May 2018. The information is gathered from Dealogic, an online platform which powers the global capital markets and has all the financial data needed. For this thesis, this data includes company name, total value and market value both pre and post the deal. Further, the offer and closing price is obtained as well as the deal specific industry, in order to analyze the level of underpricing in the different industries. To see the distinction in

underpricing between private equity and non-backed IPOs, the investment background is gathered from Dealogic. In total, 7757 companies went public in Europe in the period January 1990 till June 2018. For various reasons, data is dropped. Industries that has no PE backed firms involve are dropped. Further, in the first years of the sample till 1998, only 6 IPOs are PE backed. Therefore, this data is not relevant for the thesis. Next to this, in these first years of data there are several variables without values and therefore the data used for this thesis start from 1999. Finally, from 5,475 IPOs left, the level of underpricing is measured. The regressions are made based on this final number of IPOs.

4.1 Summary Statistics

Table 1 reports descriptive statistics for the various variables used in the regressions

Table 1 Descriptive statistics

Variable N Mean Median Std. Dev. Min Max Initial Return (%) 5,455 16.57 5.43 39.01 -34.5. 240.91 - if PE backed (%) 556 6.51 2.98 16.16 -34.5 240.91 - if VC backed (%) 284 13.19 4.32 36.70. -34.5 240.91 - if PE or VC backed (%) 723 8.50 3.33 24.87 -34.5 240.91 - if Non-backed (%) 4,752 17.76 5.82 40.54 -34.5 240.91 Offer price 5,455 9.82 2.89 49.47 0.0028 2680.31 Shares offered (*million) 5,269 55.3 6.85 701 10 38000 Proceeds (*million) 5,269 167 20 2030 87.4 136000 Deal Value (*million) 5,288 137.4 20.45 478 0.00009 16549,85 Net revenue (*million) 5,455 3.14 0.58 9.40 0 272.907 Market value (*million) 5,455 385.5 39.54 2027.92 0 54182.24 % of company sold 5,455 34.59 29.03 25.05 0 103.83 # lead managers 5,455 1.48 1 1.17 0 15 Financial sponsor Pre stake 323 61.37 65.18 28.70 3.53 100 Financial sponsor Post stake 257 31.26 30 19.63 0 100

From the Financial sponsor pre- and especially post-stake, some observations are missing in the data set. Table 2 (appendix) shows the level of underpricing for all the different industries within the sample, separated for non-backed, PE- and VC-backed and a column with the average per industry. T-tests find a few significant differences within industries. More precisely, a lower level of underpricing for some PE- or VC-backed IPOs compared to non-backed IPOs within the corresponding industry. There seems to be a difference in information asymmetry in these industries. Private equity investors have an advantage of private

information with which they could reduce the level of underpricing. This is the case for the industries Computer & Electronics, Dining & lodging, finance, healthcare, Metal & Steel, Publishing, Professional services and retail. The PE-backed companies are significant less underpriced in these industries.

Table 3 (appendix) shows statistics for the levels of underpricing per year, separately by sponsor-backed and non-backed. The table shows the level of underpricing for every year within the sample, separated for non-backed, PE- and VC-backed and a column with the average level of underpricing per year. T-tests find some significant difference in underpricing within years, between the sponsored and non-backed IPOs. The sponsored IPOs show a significant lower underpricing for the years 1999, 2004, 2005, 2013, 2014, 2015, 2016 and on average. Remarkable here is the level of underpricing in 2010. Sponsor backed IPOs are 28.91% underpriced, significant more than the non-backed IPOs in this year. Except that it was the crisis period and a cold market expects less underpricing, there is no theory from literature which explains this suddenly high level of underpricing. Sponsors can try to signal quality and growth for their companies going public. It could be a strategy to attract investors with this during the crisis period. Graph 1 (appendix) corresponding to Table 3 shows the level of underpricing for each category non-backed and PE- or VC-backed IPOs, belonging to Table 3. Chambers and Dimson (2009) find an increasing level of underpricing over the years. To create a visual image before making regressions, the line in graph 2 seems to decline slowly over the years. This will be tested in the regressions in section 5.

4.2 Independent variables

The variable of most interest is the one for sponsor backed IPOs. The dummy PE and VB represents the private- and venture-backed IPOs respectively. Because of the size of the sample of PE backed IPOs, and the small difference in characteristics, I take PE- and

VB-20 backed IPOs together. The regression in table III and the result of regression (2) specifically, also explains the reason for this.

Then I use some control variables that explain underpricing in the literature. For the size, offer price and shares offered are used. Proceeds are measured as shares offered times the offer price. The natural logarithm of proceeds is commonly used as control variable. It is a measure for information asymmetry and according to Beatty and Ritter (1986), smaller offerings are more speculative than larger ones. They have a higher ex ante uncertainty than larger companies. The bigger a company is, the more information available. Thus, lower uncertainty and less information asymmetry (Benveniste and Spindt, 1989). Therefore, I expect that size has a negative effect on underpricing.

The independent variable company sold is the percentage of the company that is sold to the public. This variable shows the percentage of the companies’ shares offered and is calculated as shares offered divided by shares outstanding times hundred. Aggarwal et al. (2001) state that the higher the stake in ownership is, thus a low percentage sold, the higher the level of underpricing. However, when the firm sells a greater portion of the shares at time of the IPO and want to maximize their proceeds, this proceeds effect can dominate. Habib and Ljungqvist (2001) agree with Aggerwal et al. (2001) that the higher the stake offered to the public the lower the level of underpricing is. Therefore, we expect this variable to have a negative effect on initial return.

Then the offer price is an independent variable. Lower offer prices are associated with smaller firms and as explained for the size, this results in higher levels of underpricing.

Namely, lower offer prices are likely to have lower proceeds (Bradley et al., 2006). Therefore, a low offer price is mostly paired with a small size company. Fernando et al. (2004) find significant higher levels of underpricing for IPOs with lower offer prices. They find that the relation of underpricing and offer price is u-shaped. A high offer price also increases the level of underpricing. Therefore, the offer price as variable itself is not very interesting, but it should be divided by subgroups. One subgroup consists of offer price below 5 euro. Then a subgroup between 5 and 15 euro and a high offer price group from above 30 euro.

Another control variable is for industries. In some regressions, industry fixed effects are used for every industry. In others, any of them are used as control variables. Technology firms expose a higher level of risk and the combination of a high return and more volatility leads to a positive effect. Therefore, a dummy for the computer & electronics industry is

included which I expect to be positive, because it seems harder to estimate. The opposite is true for the real estate industry, which is easier to value because of more symmetry. From this smaller degree of information asymmetry and thus better idea of right valuation of a company in this industry, we expect the real estate industry to have a negative effect on underpricing.

Fernando et al. (2004) use the natural logarithm of market value as size. The market value contains more information than the deal value. Furthermore, PE- and VC-investors are interested in the value of the firm before going public and their growth capacities, not to proceeds from the IPO only. Because size and underpricing seem to have a relation, dummies for the market capitalization are used to see whether there is a difference under PE- and VC-backed firms. With these variables, hypothesis 3 and 4 can be tested. As seen in section 3, PE-investors concentrate on a single company instead of dividing their money over a few smaller companies. Therefore, the expectation is that PE-investors invest more in the companies with a higher market value. A dummy small market value is included for a market capitalization smaller than €100.000.000, -. A dummy mid-market value is included for a market

capitalization between €100.000.000, - and €1.000.000.000, - A dummy large market value is included for a market capitalization greater than €1.000.000.000,-.

4. Empirical Method

In order to answer the research question and hypotheses, data related to IPOs is collected. To examine the hypotheses, which all have to do with underpricing, an ordinary least squares regression model and panel data with fixed effects is used. Furthermore, some DID models are tested. The regressions consist of the dependent variable initial return, which is a yield that measures whether the IPO is underpriced, overpriced or correctly priced. As seen in existing literature, the initial return is calculated by the two variables offer price and closing price of the stock on the first trading day. It calculates the level of underpricing of the IPO. To get this yield, the following calculation is done:

Level of underpricing

=

* 100%

Where closing price stands for the aftermarket closing price of the first trading day and offer price for the offer price.

I use panel data with fixed effects to test the relationship described above, between the level of underpricing and the independent variables:

!_" = %_" + '₍)_("+ '_*)_*"+ … . + '_-)_-"+ '_./₍+ ℰ_"

Where !" denotes the level of underpricing ‘Y’ at company ‘i’. This means that the dependent variable Y stands for the level of underpricing for each individual company ‘i’ that went public in the period that is used. The independent variables are represented by )(1 + )*1 + … + ).". % is a constant that gives the point where the dependent variable crosses the regression line and variable ℰ denotes the error term.

In order to control for fixed effects for industry and year, the variables ∑*: 34/56789 ";( _" and ∑?@(!<=8_>

>;( are added to the model. There are 24 different industry intercepts for each industry and 20 different time intercepts, one for each year, represented by binary variables. These fixed effects absorb the influences of all omitted variables that differ from one time and industry to the other. The combined fixed effects eliminate omitted variable bias arising from unobserved variables that are constant across industries and over time (Stock and Watson, 2012, p. 402). Another way to control for year fixed effects is including a crisis

dummy which equals one if the period is during the financial crisis from 2007 till 2012 and zero otherwise. Furthermore, the dummy for the bubble is included for the internet bubble in 1999 and 2000. One model uses the crisis and bubble dummy and the other model the year dummy variable to fix the effect of time. To test whether the MiFID has a significant effect on level underpricing, a dummy MiFID represents the sign that the regulatory has on

underpricing. Since MiFID is introduced, rules have become stricter and underwriters could choose less the allocation of shares to investors. It implemented new measures such as pre- and post-trade transparency requirements. Because the information asymmetry should be lower after implementation of the MiFID, I expect that the closing price is closer to the offer price and thus a lower level of underpricing after incorporation. This results in an expected negative sign after the MiFID.

I first run a univariate regression on private equity and venture capital because this is the major question and I do not want other variables to take over the effect by small

correlations. Then I run the regression by starting with one of the dependent variable on the control variables and then adding the independent variables. This to see the incremental explanatory power of the independent variable. Fernando et al. (2004) also tests first on underpricing and his explanatory variable itself. Later, they run a multivariate regression of underpricing on the variables, after including the control variables.

Then I use a difference in difference (DID) model to test the interaction with some time dependent variables and PE- and VC-backed IPOs. The DID model estimates the

interaction with the control variable PE- and VC-backed and the bubble years, the crisis, and the MiFID separately. A DID is easy to use here to test for both cross-sectional and time-dependent effects. From this, the following model is obtained for the interaction with MiFID:

34A7A=B C<7584_",> = '_E + '₍FGHI_"+ '_*JAK3L_"+ '_MFGHIJAK3L_"

The DID model for PE- and VC-backed IPOs with the bubble:

34A7A=B C<7584_",> = '_E+ '₍FGHI_" + '_*N5OOB<_"+ '_MFGHIN5OOB<_"

Then the effects of PE- and VC-backed IPOs for the crisis are tested apart for the period before, during, and after the crisis. Because the sample consists data over a long period of 20

24 years, there could be significant difference in underpricing in the years before the crisis, under PE- and VC-backed IPOs compared to the years after the crisis. The way of valuing companies has changed over time. Besides of that, one would expect that better techniques improves the accuracy of valuation. Therefore, it is interesting to see the different results from the interaction of sponsor-backed companies with years before the crisis, during and years after. This leads to the following three DID models, before, during and after the crisis respectively:

34A7A=B C<7584_",>= '_E + '₍FGHI_"+ '_*F8<P8A6A6_"+ '_MFGHIF8<P8A6A6_"

34A7A=B C<7584_",> = '_E+ '₍FGHI_" + '_*I8A6A6_"+ '_MFGHII8A6A6_"

34A7A=B C<7584_",> = '_E+ '₍FGHI_" + '_*FQ67P8A6A6_"+ '_MFGHIFQ67P8A6A6_"

Then, regressions on small- mid- and large companies show the differences in underpricing between size. The models will be runned for VC-backed and separately for PE-backed companies, to see whether there is also a difference within the sponsor backed firms and to test hypothesis 3 and 4. Lastly, an OLS regression of only PE-backed companies tests what influence the stake percentage difference has on the initial return in order to test hypothesis 5.

5. Results

In this section are the main results of the OLS regression and panel data regression with fixed effects presented. Before testing the models, the correlation between all the variables are tested for multicollinearity. No high correlations have been found that could influence a regression outcome negatively.

Underpricing for private equity and venture capital IPOs

The first hypothesis is that PE-backed IPOs have a lower first day initial return than non-backed IPOs. In table (4), the univariate regression (1) to test this test hypothesis is done. The result shows a constant of 17.19% initial return minus 6.41% in the case the firm is PE-backed. This gives on average 10.78% initial return for a PE-backed IPO. This is not exactly the same as found in the summary statistics in section 4, because fixed effects for years and industries are included in the regression. The result is statistically significant on a 1% level. PE investors have more information than non-backed companies and can lower the information asymmetry. Less asymmetry and a better knowledge of the exact value of the company leads to less underpricing. The result is totally what we expect, PE-backed firms are less underpriced.

To test hypothesis 2, regression (2) shows the univariate regression of initial return on VC-backed IPOs. As seen in the descriptive statistics, VC-backed IPOs do have a lower level of underpricing. However, the difference is not significant and the difference is smaller than the difference from PE-backed to non-backed IPOs. VC investors do want high proceeds and thus try to lower underpricing. However, as explained in section 3, they sometimes want a higher level of underpricing, to show the market they are growing and attract investors.

Furthermore, they take higher risks in general than PE investors do and there is therefore more uncertainty about the pricing of the company. This contradictory effect gives a difference with non-backed IPOs that is too small on average. In combination with a low number for the sample of VC-backed IPOs and accounting for fixed effects, this result is not significant.

To make the sample bigger and because there are some reasons given in section 2 why we can treat VC-backed the same as PE-backed IPOs, I take the samples together and run a regression (3) which results in a significant lower level of underpricing of -5.08%, on a 1% level. This means an IPO is 5.08% less underpriced when the company is PE- or VC-backed, compared to the non-backed IPOs. For all other regressions I take PE- and VC-backed

26 together in one dummy. So, absorbing for fixed effects and thus allowing for correlation with the independent variable over time and industries, underpricing is lower for PE- and VC-backed IPOs. The hypothesis that PE-VC-backed IPOs are less underpriced than non-VC-backed IPOs is proven here.

TABLE 4 Underpricing on PE- and VC- backed IPOs

Effect of private equity funding and venture capital funding IPOs on the level of underpricing.

This table reports coefficients from OLS regressions of initial return on an indicator for whether the firm is private equity backed, venture capital backed or one of both. The dependent variable initial return is the return from the offer price to the closing price on the first trading day at the IPO date. All models include 5,455 IPOs in Europe between 1998 and 2018. The independent variable PE dummy for model (1) takes on a value of 1 (zero otherwise) if the IPO is private equity backed. The independent variable VC dummy for model (2) takes on a value of 1 (zero otherwise) if the IPO is venture capital backed. The independent variable PEVC dummy for model (3) takes on a value of 1 (zero otherwise) if the IPO is private equity or venture capital backed. P-values are reported in parentheses.

Dependent variable: Initial return

(1) (2) (3) PE -6.41*** (0.001) VC -0.89 (0.728) PEVC -5.08*** (0.002) Constant 17.19*** 16.58*** 17.21*** (0.000) (0.000) (0.000)

Firm Fixed Effects Yes Yes Yes

Year Fixed Effects Yes Yes Yes

N 5,475 5,475 5,475

F-statistic 11.73 0.12 9.15

Prob > F 0.001 0.7284 0.000

Adj. R-sq 0.0430 0.0408 0.0425

*** denotes significance at the 1% level; ** notes significance at the 5% level; * denotes significance at the 10% level.

Effect on underpricing including control variables

Then the regressions including control variables show the effects of these variables on underpricing in Europe, accounting for year and industry fixed effects. The results are shown in Table (5). The first regression is a regression on the control variables, as explained in the empirical method. Regress underpricing on the natural logarithm of the deal value, offer price, industry effects and the percentage of company sold to test on underpricing on control variables. The regression shows indeed a negative significant effect for size. When the deal value is bigger, the IPO is less underpriced. A price between the 5 and 15 euro also results in a significant lower return. This is in line with the expectation of the price following Fernando et al. (2004), who find that the share price has a u-shaped curve in relation to underpricing in the US. The mid-price is on the point of the curve where it lowers underpricing and that is indeed true for this sample of the European market. The technology industry does have a positive impact but is not significant. From the regression, it can not be said that this effect is positive. Real estate shows a negative significant effect from -4.40% on the 1% level. As expected, the real estate industry has a low level of information asymmetry because a lot of information is generally known. Therefore, the underpricing within this industry is also lower. The percentage of company sold gives a negative significant effect of -0.10%. The higher the stake of the company sold to the public with the IPO, the lower the underpricing. This

corresponds with the findings of Aggarwal et al. (2001). The firm probably wants to maximize their proceeds when selling more shares to the public. When they keep a sufficient stake, they benefit from increasing price after the IPO too. The independent variables in regression (1) with control variables explain 9.58% of the variation in underpricing, measured by the R-squared. Then for regression (2), the size and percentage of company sold are about the same as in the first regression and significant on a 1% level. The difference is the offer price. To test whether the prices in Europe could also be seen as u-shaped to the level of

underpricing, the offer price till 5 euro is included here, and a high offer price from above 30 euro. Both variables show a higher level of underpricing and significant. In combination with the mid-price segment, the price in Europe thus follows a u-shaped curve to underpricing. The adjusted R-squared is here a little lower, 8.69% of the variation in underpricing is

explained by the independent variables. It is probably that the most prices are in the range of 5 to 15 euro per share, so this result contains more IPOs and is reliable.

28 When then including the variable of interest PE- and VC-backed firms in regression (3) the control variables are comparable to the previous regressions. However, the variable of interest changes from sign and is the only variable not significant. As seen in Table (4) the sponsor backed IPOs are lower than the non-backed. When looking at the table with

distribution small-, mid- and large market values, the PE- and VC-backed IPOs invest in larger firms overall. Deal value as used in this regression is less than the market value used for that table (9), however they are correlated. Their correlation is 0.86, so highly correlated and as explained in the data section, market value is a good measure to test for sponsor-backed IPOs. PE-sponsors invest more than twice as much IPOs with mid-market value and the VC-backed twice as much than the non-VC-backed. The non-VC-backed sample invest about 70% of the total deals in small firms compared to only half of it on average by the sponsor-backed companies. Therefore, it is possible that firm size and private equity and venture capital interacts in a way not fully captured by the regression. When then measuring size with the natural logarithm of shares offered and offer price in regression (4), the variable for sponsor backing is significant and negative as expected. However, this do lower the explanation of variance in underpricing, as seen in the lower adjusted R-squares of 6.14%. The deal value of the IPO is thus a better measure for size than shares offered, which we could expect from literature and logical reasoning. Shares offered does not take into account the offer price so this measure of size is not the best. Since deal value and shares offered are highly correlated, it will be used as substitution. The interaction between deal value and sponsor backing disappears then. As the expectation, sponsor backing is then negative and significant in

regression (4). Regression (5) runs on only PE- and VC-backed IPOs and has therefore only 709 observations. Size has a lower impact than in the other regressions, but the constant is also lower. As seen in Table (9), the size of this sample should be higher on average too, as the sponsor-backed IPOs has a two times larger stake in the larger firms. The impact of size is negative and significant and thus the interaction of sponsor backed IPOs is total captured by the sponsor IPOs here. Therefore, the sponsor backing dummy in regression (3) does not take over the effect of size from non-backed firms. As expected and explained above, the price range between 5 and 15 euro is negative and has a significant effect on the 10% level. The percentage of company sold is not significant anymore. This can be due to the fact that PE- and VC-investors care less about this percentage. They care more about their own stake and the proceeds from that. The explanation of the model is highest here, 27.86% of the variation

in underpricing is determined by these independent variables. The constant is lower and the sponsor backed firms have a negative effect on underpricing with the control variables in the regression (4). Further, Fernando et al. (2004) use a sample from US between 1981 and 1998 to find the relation between offer price and underpricing. I find a u-shaped curve as well for Europe between 1999 and 2018.

Table (6) shows the level of underpricing for the sponsor backed IPOs within the sample, separated PE- and VC-backed and tested to the non-backed IPOs. T-tests confirm the significant differences between the sponsor- and the non-backed firms, as found in the regressions. More precisely, a lower level of underpricing of 6.51% for some PE-backed IPOs compared to 17.67% for the non-backed IPOs in the sample, significant on a 1% level. Private equity investors do have an advantage of private information with which they could reduce the level of underpricing. This result leads to rejection of the first hypothesis:

R_E: PE-backed IPOs have the same level of underpricing as non-backed IPOs R₍: PE-backed IPOs have a lower level of underpricing as non-backed IPOs

Private equity backed IPOs do have a lower level of underpricing than non-backed IPOs. Then Table (7) contain test statistics from the T-test of VC-backed IPOs to the non-backed IPOs. The difference between those groups is smaller, however still statistically significant on the 10% level. The smaller difference could be due to small-cap VC-backed IPOs that have a relative high level of underpricing within the VC-backed IPOs. However, these results lead to rejection of the second hypothesis.

RE: VC-backed IPOs have the same level of underpricing as non-backed IPOs R₍: VC-backed IPOs have a lower level of underpricing as non-backed IPOs

TABLE 5 OLS regressions on underpricing including control variables

This table reports coefficients from OLS regressions of initial return on a dummy for PE- or VC- backed IPOs and control variables. Models (1), (2) and (3) include 5,065 IPOs, model (4) 5,288 IPOs and model (5) include 709 PE- or VC-backed IPOs. All in Europe between 1998 and 2018. The dependent variable initial return is the return from the offer price to the closing price on the first trading day at the IPO date. The independent variable PEVB models takes on a value of 1 (zero otherwise) if the IPO is PE- or VC-backed. The ‘Offer price low’ dummy takes on a value of 1 (zero otherwise) if the IPO offer price is smaller than or equal to 5 euro. The ‘Offer price mid’ dummy takes on a value of 1 (zero otherwise) if the IPO offer price is between 5 euro and 15 euro. The ‘Offer price high’ dummy takes on a value of 1 (zero otherwise) if the IPO offer price is higher than 30 euro. The ‘Tech’ dummy takes on a value of 1 (zero otherwise) if the IPO is in the industry group ‘Computer & Electronics’. The ‘Real Estate’ dummy takes on a value of 1 (zero otherwise) if the IPO is in the industry group ‘Real estate’. The variable ‘% of company sold’ indicates the stake from the shares offered by the shares outstanding, multiplied by 100. P-values are reported in parentheses.

Dependent variable: Initial return

(1) (2) (3) (4) (5)

Size -3.96*** -4.05*** -3.98*** -2.46***

(0.000) (0.000) (0.000) (0.000)

Shares offered 1.11***

(0.001)

Offer price low 2.87** 9.70***

(0.024) (0.000)

Offer price mid -5.21*** -4.81*** -3.29*

(0.000) (0.000) (0.087)

Offer price high 9.40*** 5.82**

(0.000) (0.017) PEVB 1.20 -4.36*** (0.475) (0.009) Tech 1.88 (0.158) Real estate -4.40** (0.037) % of company sold 0.10*** 0.09*** 0.08*** 0.01

(0.000) (0.001) (0.001) (0.797)

Constant 80.00*** 78.32*** 80.52*** 28.01*** 53.36***

(0.000) (0.000) (0.000) (0.000) (0.000)

Year Fixed Effects Yes Yes Yes Yes Yes

Industry Fixed Effects No Yes Yes Yes Yes

N 5,065 5,065 5,065 5,288 709

F-statistic 58.30 56.08 55.33 17.84 6.91

Prob > F 0.000 0.000 0.000 (0.000) (0.000)

Adj. R-sq 0.0958 0.0869 0.0864 0.0614 0.2786

*** denotes significance at the 1% level; ** notes significance at the 5% level; * denotes significance at the 10% level.

Table 6 Underpricing between PE- backed and non-backed IPOs

Year Non-backed PE-backed

Mean 17.67% 6.51%*** Std. Dev. 40.6 16.7 Observations 4,919 556 Df 5,473 t-statistic 6.43 P(T<=) one tail 0.000

*** denotes significance at the 1% level; ** notes significance at the 5% level; * denotes significance at the 10% level.

Table 7 Underpricing between VC- backed and non-backed IPOs

Year Non-backed VC-backed

Mean 16.72% 13.19%* Std. Dev. 39.1 36.7 Observations 5,191 284 Df 5,473 t-statistic 1.48 P(T<=) one tail 0.0688

*** denotes significance at the 1% level; ** notes significance at the 5% level; * denotes significance at the 10% level.

Difference-in-Difference models

The regressions in table (8) include the DID models. Separate models are tested between years before, during and after the crisis. Furthermore, regressions on the MiFID and bubble years show the effects on underpricing. All on the interaction with PE- and VC-funded IPOs. Regression (1) shows a level of underpricing of 13.62%. When occurring in the pre-crisis and is non-backed, this return is 22.15% and significant. This is in line with the expectation because the bubble years are included in the years before the crisis which gets the average up. The PE- and VC-backed IPOs impacts the underpricing with -6.20%. The interaction term gives the result of an IPO which is PE- or VC-backed and took place before the crisis. In this case, the IPO gives a significant lower return of -5.65%. A PE- or VC-backed IPO done before the crisis, therefore gives an initial return of 10.3%, compared to 22.15% for the non-backed IPOs before the crisis. The sponsor backed IPOs were thus 11.85% less underpriced than non-backed IPOs before the crisis. Then regression (2) gives the result for the DID model in the crisis period. Being a sponsor backed IPO gives a significant lower return of -10.72% on the constant of 18.38%. The crisis years have a significant impact of -2.21% and the interaction of a sponsor funded IPO that took place during the crisis gives a 6.86% higher initial return. As you expect the cold market to have a lower level of return, it is a bit surprising that the interaction term of the sponsor term with the crisis period shows a significant higher level of underpricing. The result of 12.13% underpricing of sponsor backed IPOs during the crisis, is relative high for sponsor backed IPOs, because they show an initial return of 8.50% over the years on average, as found in the statistics in section (4). Then regression (3) on the post-crisis period gives a significant initial return for sponsor investments of -9.07%. When the IPO took place after the crisis, from 2013 on, the underpricing decreases significant with 9.42%. Then the interaction of private equity and the period after the crisis is 3.62%. However, the interaction is not significant. The level of underpricing for non-backed firms after the crisis is 10.47%. For sponsor backed firms this is -9.07% return less. The interaction term makes this around 5% underpricing for sponsor backed IPOs. However, this could not be concluded for sure because of the insignificance of the interaction term. The fourth regression shows the effect of underpricing on PE- and VC-backed IPOs, MiFID and the interaction. The sponsored IPOs are again significant on a 1% level, - 10.45% initial return compared to the non-backed IPOs. The dummy for the MiFID I regulation gives a significant lower level of underpricing. After the MiFID I is implemented, the IPOs have -7.42% underpricing on average, significant

on the 1% level. The interaction variable of sponsor backed IPOs and the MiFID is 3.58%, however not significant. Since the MiFID regulation should lower the levels of underpricing according to literature, and sponsor backed IPOs too, it is surprising that the interaction is not. It could be that the difference between non-backed and sponsor funded IPOs is smaller after the MiFID I due to transparency and increased liquidity. The last regression (5) of table (8) gives the effect of the bubble years, sponsor backed IPOs and their interaction. The bubble years give a significant higher level of underpricing of 17.52% compared to all other years. The PE- and VC-backed IPOs are -6.39% and significant too here. Interaction of these IPOs with the bubble is -2.57% but this is not a significant lower return. This could be due to the difference between sponsor- and non-backed within these years. Table (3)shows this difference within the bubble years. As seen there, the difference for 1999 is large and the sponsor backed IPOs were significant lower. Year 2000 shows an opposite result. The sponsor- and non-backed are practically the same here. Because the bubble years include only two years, the average result of the interaction does not reveal a significant difference between the samples. The difference could be due to the internet bubble which attracted lots of investors for all the internet related companies that went public. For this sample, 40% of the IPOs were tech companies during the bubble and these were highly underpriced, for sponsor backed IPOs especially in 2000. According Chamber and Dimsons (2009), sponsor backed companies may kept a stake and sell the remainder after the share price was bidden up. Furthermore, it was harder to underprice for sponsor backed companies too in 2000 because of the immense market. At least, the interaction of sponsor-backed and the bubble dummy is not significant for the combination of these two bubble years.

Table 8 Difference-in-Difference model for PE- and VC-backed IPOs of the MiFID, bubble and crisis period on the level of underpricing

This table reports coefficients from OLS regressions of initial return on a dummy for PE- or VC- backed IPOs and on different time intervals and their interaction. All models include 5,475 IPOs in Europe between 1998 and 2018. The dependent variable initial return is the return from the offer price to the closing price on the first trading day at the IPO date. The

independent variable PEVB for all models takes on a value of 1 (zero otherwise) if the IPO is PE- or VC-backed. The crisis dummy in model (2) takes on a value of 1 (zero otherwise) if the IPO took place in the crisis period between 2007 and 2012. The PEVB*crisis dummy takes on a value of 1 (zero otherwise) if the IPO was PE- or VC-backed and took place in the crisis period between 2007 and 2012. The pre-crisis dummy in model (1) takes on a value of 1 (zero otherwise) if the IPO took place before the crisis period. The PEVB*precrisis dummy takes on a value of 1 (zero otherwise) if the IPO was PE- or VC-backed and took place before the crisis period. The post-crisis dummy in model (3) takes on a value of 1 (zero otherwise) if the IPO took place after the crisis period. The PEVB*postcrisis dummy takes on a value of 1 (zero otherwise) if the IPO was PE- or VC-backed and took place after the crisis period. The variable MiFID takes on a value of 1 (zero otherwise) if the IPO took place after the MiFID I is implemented in November 2007. The PEVB*MiFID takes on a value of 1 (zero otherwise) if the IPO was PE- or VC-backed and took place after the implementation of the MiFID I. The bubble dummy in model (5) takes on a value of 1 (zero otherwise) if the IPO took place during the bubble years in 1999 or 2000. The PEVB*bubble dummy takes on a value of 1 (zero otherwise) if the IPO was PE- or VC-backed and took place in the bubble years. P-values are reported in parentheses.

Dependent variable: Initial return

(1) (2) (3) (4) (5) PEVB -6.20*** -10.72*** -9.07*** -10.45*** -6.39*** (0.000) (0.000) (0.000) (0.000) (0.000) Pre-crisis 8.53*** (0.000) Crisis -2.21* (0.079) Post-crisis -9.42*** (0.000) MiFID -7.42*** (0.000) Bubble 17.52*** (0.000) PEVB*crisis 6.86* (0.069) PEVB*precrisis -5.65* (0.072) PEVB*postcrisis 3.62 (0.263) PEVB*MiFID 3.58

(0.247)

PEVB*bubble -2.57

(0.676)

Constant 13.62*** 18.38*** 19.89*** 20.86*** 14.25***

(0.000) (0.000) (0.000) (0.000) (0.000)

Firm Fixed Effects Yes Yes Yes Yes Yes

Year Fixed Effects No No No No No

N 5,475 5,475 5,475 5,475 5,475

F-statistic 30.49 12.93 29.01 25.99 61.51

Prob > F 0.000 0.000 0.000 0.000 0.000

Adj. R-sq 0.0291 0.0198 0.0283 0.0156 0.0452