IPO underpricing and the partial adjustment phenomenon : the effect of venture capitalist backing on price revision in the tech sector

University of Amsterdam

Amsterdam Business School

MSc Finance

Track: Corporate Finance

Master Thesis

IPO Underpricing and the Partial Adjustment Phenomenon

The Effect of Venture Capitalist Backing on Price Revision in the

Tech Sector

Author: Tammo Paul Johannes Abbenhuis Student Number: 11933720 Thesis Supervisor: Dr. J. J. G. Lemmen Finish Date: July 2018

Statement of Originality

This document is written by student Tammo Paul Johannes Abbenhuis, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is 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 analyses the effect of venture capitalist (VC) backing on price revision and underpricing in hot and cold IPO markets. A comprehensive dataset on 524 IPOs occurring between 1992 and 2018 in the Tech sector has been established to answer the questions at hand. This study finds a causal positive relationship between upwards price revision and underpricing, confirming the partial adjustment phenomenon (Ibbotson, Sindelar & Ritter, 1988). Other findings include the downward effect of VC presence on both upwards price revision as underpricing in IPOs. This confirms the long term investment interest of VCs. Finally, higher levels of upward price revision and underpricing in hot IPO markets are found – an effect decreased by the presence of VCs. Results are found using both diff-in-diff analyses with propensity score matching as multivariate regression analyses with interaction terms. Keywords: Underpricing, IPO, Price Revision, Partial Adjustment Phenomenon, VCs, IPO Cycles. JEL Classification: C30; G14; G24; G32

TABLE OF CONTENTS

List of Tables

………. p.5

List of Figures

………..

p. 5

Chapter 1: Introduction

………. p. 6

Chapter 2: Literature Review

………. p. 9

2.1 Going Public: Benefits and Costs

……….... p. 9 2.1.1 Why is Underpricing Being Accepted?.……….. p. 10

2.2 Potential Solution to Underpricing: Bookbuilding

……….. p. 12 2.2.1 The Partial Adjustment Phenomenon………... p. 12

2.3 Venture Capitalists

……… p. 14

2.3.1 Venture Capitalist Investment Interests ……….. p. 15

2.4 The Technology Sector

………... p. 17

2.5 IPO Cycles

……….. p. 18

Chapter 3: Data Description

………. p. 19

Chapter 4: Methodology

………... p. 27

4.1 The Causal Relation between Price Revision and Underpricing

……….. p. 27

4.2 The Effect of VC Backing on Price Revision

………... p. 28

A. Multivariate OLS Regression Model……….. p. 28 B. Diff-in-diff Analysis: Propensity Score Matching………... p. 28

4.3 The Effect of VC Backing on Underpricing

……….. p. 30

A. Multivariate OLS Regression Model………... p. 30 B. Diff-in-diff Analysis: Propensity Score Matching………... p. 31

4.4 Differences in Price Revision and Underpricing: Hot and Cold IPO markets

…. p. 31

4.5 Effect of VC Reputation on Price Revision and Underpricing

……….. p. 32

Chapter 5: Results

……… p. 32

Chapter 6: Robustness Checks

……….. p. 43

Chapter 7: Conclusion

………. p. 46

Chapter 8: Limitations

……… p. 47

Reference List

………... p. 48

Appendices

……….. p. 52

LIST OF TABLES

Table 1: Descriptive Statistics……….. p. 21 Table 2: Comparison between the VC and non-VC Backed Sample ……….. p. 22 Table 3: Comparison between Hot and Cold IPO Markets……….. p. 24 Table 4: The Effect of Price Revision on Underpricing………. p. 33 Table 5: The Effect of VC backing on Price Revision……….. p. 35 Table 6: Regression Results: VC Backing on Underpricing Through Price Revision….. p. 38 Table 7: Propensity Score Matching: Differences in Underpricing and Price Revision between the VC backed and matched non-VC Backed Sample ………. p. 39 Table 8: The Effect of Market Heat on Price Revision and Underpricing………... p. 41 Table 9: Propensity Score Matching: Differences in Underpricing and Price Revision of Hot IPO Markets and Matched Cold IPO Markets………... p. 42 Table 10: Propensity Score Matching: Differences in Underpricing and Price Revision of High vs Low Reputation VC backing……… p. 43 Table 11: Robustness Check: Propensity Score Matched Diff-in-Diff Analyses using Two Other Matching Variables……….. p. 44 Table 12: Robustness Check 2: Regression Results. Differences in Hot and Cold IPO Markets ……….. p. 45

LIST OF FIGURES

Figure 1: IPO Volume per Year ……… p. 25 Figure 2: Average Underpricing and Price Revision per Year ……… p. 25 Figure 3: Price Revision and Underpricing (only upwards revision) ………p. 26 Figure 4: Normal Distributions of Propensity Scores………. p. 30

APPENDICES

Appendix A: Full Description of Variables……… p. 52 Appendix B: Distribution of Price Revision Observations……… p. 54 Appendix C: Testing for Multicollinearity in Equation (I) ……….. p. 55 Appendix D: Testing for Multicollinearity in Equation (II) ………. p. 55 Appendix E: Testing for Multicollinearity in Equation (IV & V) ……… p. 55 Appendix F: Testing for Multicollinearity in Equation (VI & VII) ……… p. 56 Appendix G: Testing for Multicollinearity in Table 12……….. p. 56

CHAPTER 1

INTRODUCTION

In November 2013, the major social media platform Twitter went public. Throughout the road show (the process from entering the IPO process to the first trading day) Twitter changed its offering price several times. The volatility of Twitters’ valuation is representative for the technology sector. Valuations are difficult because of the uncertain future of tech companies. Twitter was no different. In September 2013, Twitter valued itself at 20.62$ per common share (Gaprindashvili, 2013). However, only one month later, Twitter was re-valued by their main underwriters JP Morgan, Goldman Sachs and Morgan Stanley at an offer price between 17$ and 20$ (issuing anywhere between 70 and 85 million shares) (Barr, 2013). Twitters’ valuation was based on its growth potential, customer base and business plan. For example, the company experienced a 247% increase in advertising revenue from 2011 to 2012 (Gupta, 2013). Finally, on the 7th _{of November 2013, the company hit the market at an offering price of 26$ issuing 70} million shares. Representing upwards price revision of 40.5% from the midpoint of the filing range. Despite the high upwards revision, the stock closed in at 44.9$, equalling underpricing of 73%. (Pepitone, 2013). The level of underpricing represented a huge amount of money left on the table. Nevertheless, investors were happy with the results of the initial public offering (IPO). Twitter’s IPO process represents the difficulties in valuing a volatile tech company. The IPO results leave us with one of the most asked questions in finance research: why is underpricing (an opportunity cost) being accepted at such a high level? A vast amount of research has been done on underpricing. Most known theories as to explaining underpricing include the Winner’s Curse, introduced by Kevin Rock in 1986 (further explained in chapter 2) and the Signalling hypothesis, which states that IPOs are underpriced on average in order to sell future offerings at a higher price (Garfinkel, 1993). Ljungqvist (2007) groups several explanatory theories under one of the four following headers: asymmetric information, institutional reasons, control considerations and behavioral approaches. Asymmetric models explain underpricing as a consequence of differences in information between the key parties to an IPO (the Winner’s Curse for example). Institutional theories look at the three main features of the market place: taxes, litigation and price stabilization by banks. Control theories argue that underpricing is necessary to stabilize the shareholder base of the underlying issuer once public, minimizing intervention by outside investors. And behavioral theories examine the irrationality of investors and the effect of their biddings on prices. The majority of the literature supports the information asymmetry view, which is also the main area of interest for this thesis. Information asymmetry in the IPO market exists through a difference in knowledge between investors, issuers and underwriters that leads to inaccurate pricing of IPOs. The main information asymmetry theory examined in this thesis is

the Partial Adjustment Phenomenon, introduced by Ibbotson, Sindelar and Ritter in 1988. The theory states that price revision from the initial offer price towards the final offer price, directly affects levels of underpricing. This could provide an explanation as to why Twitter’s IPO was so underpriced eventually. Although further explained in the literature section, the theory provides further insight as to why levels of underpricing have increased, despite the introduction of bookbuilding in the US in the 90s. Bookbuilding was an IPO method introduced in the 90s in order to decrease information asymmetry in the IPO market by conducting a road show to inform on investor interest in the market. Thereby it should decrease underpricing since the offer price can be set closer to the market valuation, yet empirical evidence showed the contrary. This paper seeks to establish a causal relationship between upwards price revision and underpricing. A relationship examined before by other researchers, yet not fully confirmed and not yet examined in the technology sector. Other than examining the partial adjustment phenomenon, this paper adds some insight to the correlation between price revision and underpricing by examining the effect of VC presence. The presence of VCs on underpricing has shown contradictory findings in existing literature. Several researchers find that VC backed IPOs lead to higher levels of underpricing, whilst others show the exact opposite. Researchers arguing that VCs lead to higher levels of underpricing build on the theories that VCs have a short term interest in their investments, whilst opposite parties claim a long-term interest by VCs. Both parties build on reasonable theories to explain their findings, yet a triumphant side is not yet determined. Although the effect of VC presence on underpricing has been examined, the effect on price revision hasn’t. This study also distinguishes between hot and cold IPO markets in the Tech sector and examines differences between the two. A final contribution to the literature is the usage of a different methodology than existing papers. This thesis uses a comprehensive dataset on IPOs occurring in the Tech sector between 1992 and 2018. Data has been collected from Thomson Reuters, Compustat, VentureXpert, CRSP, Jay Ritters’ IPO database, the Federal Reserve Bank of Philadelphia and Yahoo Finance. The final dataset consists of 524 IPOs. The correlation between price revision and underpricing has been examined by conducting multivariate OLS regressions. The effect of VC backing and differences between hot and cold markets are researched through diff-in-diff analysis with propensity score matching and OLS regressions. A full explanation of the data and the methodology is given in chapters 4 and 5. In short, this thesis examines the effect of VC backing on price revision, and its indirect effect on IPO underpricing in the technology sector. Literature has shown contradictory results and no research has been done on the effect of VC backing on price revision. A final contribution to the literature lies within examining differences in effects between hot and cold IPO markets.

This study confirms the partial adjustment phenomenon by finding a statistically significant positive causal relationship between upward price revision and underpricing. Next, it finds that VCs have a long-term investment interest rather than short-term returns. This is concluded after finding that VCs decrease upward price revision and underpricing. Lastly, the results show that IPO market heat increases upward price revision and underpricing, an effect decreased by the presence of VCs. This paper continues as following: first it examines the theoretical framework around previous research on the relevant subjects. These subjects include the costs and benefits of going public, IPOs, Bookbuilding, the Partial Adjustment Phenomenon, VCs, the Technology sector and IPO cycles. Based on the written literature, 5 hypotheses are formulated to indicate what is expected to be found through the analyses. Afterwards, a descriptive look into the data collection is given, highlighting some interesting first findings and differences across several datasets. Some key characteristics and preliminary results can be estimated by eyeballing the dataset. This paper continues by explaining the steps taken to answer the formulated hypotheses in detail. Methods used and empirical regressions are clearly stated in this section: methodology. Finally the results are given and discussed, after which a conclusion shows the overall findings of the paper. Final limitations and robustness checks are indicated at the end.

CHAPTER 2

LITERATURE REVIEW

In order to answer the main questions asked in this paper, a look at the literature done on underpricing, VCs, price revision and hot and cold markets needs to be taken. In this chapter the concept of underpricing is introduced, after which several explanations are examined as to why underpricing is being accepted. Afterwards the paper dives into the main theory of this research: the Partial Adjustment Phenomenon. A look into literature done on VC investment decisions and IPO cycles is given in order to explain the hypotheses formulation.

2.1 Going Public: Benefits and Costs The main reason why firms go public is to obtain external financing without increasing debt levels. Other reasons include providing an exit for initial investors (and the original owners) or the improvement in monitoring by the stock market, which in turn could prove create value for shareholders (Lopez de Silanes, 2018). Brau & Fawcett (2006) conducted a survey where they asked 336 CFOs questions about firm incentives to go public. Along with the previously listed reasons, they find strong evidence that one of the main reasons for conducting an IPO is the option of having shares for possible future acquisitions. The acquisition of a target can be financed using two methods of payment: stock or cash (or a combination of both). When a firm is private, stock payment is not an option. By going public, private firms obtain the possibility of financing future acquisition deals using stock. Other frequently given answers as to why firms choose to go public included attaining a value to your company for potential targeting purposes and enhancing the reputation of your company. Even though there are beneficial reasons to go public, there are costs accompanied with the IPO process as well. The costs of going public come in two forms: direct costs (underwriter fees, lawyer costs, listing costs) and indirect costs: underpricing. Ritter (1987) shows that both costs are economically significant and account for about 21.22% in firm commitment offers and 31.87% in best efforts offers of gross proceeds. Direct costs seem to account for a smaller amount of total costs. More specifically, 7% of all costs in the going public process are direct costs, meaning the majority of costs go to underpricing (Chen & Ritter, 2000). Direct costs seem to be equal across different underwriters, despite the variation in their ranking and reputation (Cliff & Denis (2004); Ritter, (1987)). But what is underpricing exactly and why is it being accepted if it represents such high costs of an IPO? Underpricing occurs when the closing price of the stock on the first day of trading exceeds the initial offering price. For example: a firm goes to an underwriter for assistance in the going public process. The underwriter advises the client on a share price and amount of shares which should be offered on the market. Now underpricing occurs when, on the first day of

trading, stock closes at a market price higher than the initial offer price. For the firm going public this is an opportunity cost. Say, for example, that the stock hits the market at 10$, and underwriter fees equal 7%. The firm receives equity equal to 9.3$ a share. At the end of the day stock trades at 14$ (a 40% first day return). This means the firm could have received 13.02$ a share. The stock price was underpriced by 40% (Ritter, 2011). In light of the number of IPOs over history, it is interesting to look at why firms are still determined to go public, despite these huge cost levels. 2.1.1 Why is Underpricing Being Accepted? Several theories have been developed trying to explain why underpricing has been accepted by issuers. As stated in the introduction, the main area of interest for this thesis is the asymmetric information theories that explain underpricing. The information asymmetry view states that one of the key players in an IPO transaction (the issuing firm, the underwriting investment bank and the investors buying the stock) has significantly more information than the others. Information asymmetry can be best explained by looking at Akerlof’s (1970) Market for Lemons. Akerlof used the car market to explain information asymmetry and its effect on the average market quality and price of the cars. Suppose you would like to buy a second hand car. And all cars in the market can be qualified either as a “plumb” (a good quality second hand car) or a “lemon” (a bad quality car). You cannot know for sure whether a second hand car is a plumb or a lemon. Since you are not able to distinguish between the two, you are willing to pay nothing more than the average value of a car in the entire market. The seller however, has more information than you do. He knows which cars are plumbs and which are lemons. There is a situation of information asymmetry. If the seller sells you a plumb, he will lose value, since you are not willing to pay more than the average car in the market. Idem ditto: if the seller sells you a lemon, he will gain value. Therefore, the seller decides to sell less plumbs and more lemons. This in turn, decreases the average quality of the car in the market and thus the average value an uninformed buyer is willing to pay. The seller sells less plumbs and more lemons, creating a downward spiral. One solution to decreasing information asymmetry in the market is by increasing transparency. In the car market this can be achieved by test rides, warranties and a quality check by a third party. Rock (1986) applied Akerlofs’ lemon theory to the IPO market. The Winner’s Curse model shows that uninformed investors are at a disadvantage as to informed investors in IPO markets. Informed investors have knowledge of the true value of the issued stock, whereas uninformed investors do not. This difference in information affects their investment decisions. An informed investor will only place a buy order on IPO stock if he/she believes the offer price lies below or at the true value of the underlying asset. This as opposed to uninformed investors,

who will place their purchase orders on their own perceived value of the stock. More specifically, informed investors will increase or decrease the amount of stock they wish to purchase according to the differential between the true value and the offer price, whereas such a differential does not exist for uninformed investors. As a consequence, when the stock turns out to be overpriced (the offer price exceeds the true value) uninformed investors will receive all purchased shares and informed investors receive none (as they haven’t committed purchase orders). Uninformed investors end up with overpriced shares. In the case of underpricing (the true value exceeds the offer price), informed investors increase their purchase order accordingly with the amount of underpricing and the majority of the shares will be allocated to the informed investors. In extreme cases this would lead to 100% of the shares going to uninformed investors in cases of overpriced IPOs. This would lead to the unwillingness of uninformed investors to participate in IPO auctions, therefore leaving only informed investors in the market. Rock states that the presence of uninformed investors in markets is necessary since the demand of informed investors is insufficient to buy all the shares in even underpriced IPOs. This requires that all IPOs need to be underpriced on average. This won’t solve the informed-uninformed investor problem, but it will make sure that uninformed investors won’t be making losses on average. A possible solution to information asymmetry in IPO markets is similar as to the one in Akerlof’s second hand car market: decreasing information asymmetry by increasing transparency. A way of achieving this is by hiring a high quality underwriter. Prestigious underwriters are only willing to accommodate IPOs of high quality firms, thereby their attendance certifies the high quality of the issuer. Evidence on this matter is mixed though. But it also sheds light on another explanation of the acceptance of underpricing in IPOs since underpricing is beneficial for the lead underwriter. Liu and Ritter (2011) talk about the All-Star Analyst Coverage Theory. A possible explanation of the acceptance of underpricing is the indirect payment for all-star analyst coverage by investment bankers. Cliff & Denis (2004) argue that lead underwriters have the possibility of allocating IPOs to preferred clients in cases of underpricing. They might allocate these IPOs to clients in exchange for future business opportunities. Thus, a lead underwriter benefits if an IPO is underpriced. The issuing company might be willing to accept this underpricing, in exchange for all-star analyst coverage. Cliff and Denis show that 80% of all completed IPOs between 1993 and 2000 receive a post-IPO recommendation by the analysts of the lead underwriter, of which 95% is either strong buy or a buy recommendation. They also show that issuers switch underwriters between the IPO and a potential SEO if there has been an unexpected amount of post-IPO coverage. This indicates that issuers indeed pay for all-star analyst coverage through underpricing, the biggest and indirect cost of an IPO.

2.2 Potential Solution to Underpricing: Bookbuilding Bookbuilding has been introduced on the US IPO market in the 90s. Bookbuilding was supposed to provide a solution to information asymmetry in the IPO market. It is a method used in order to signal to the market the true value of the issuers stock. The bookbuilding procedure starts with the lead underwriter indicating a non-binding price range of the offer price based on a valuation of the issuer (Cornelli & Goldreich, 2001). This price is also known as the expected offer price. After setting this price, the IPO enters the “waiting period”, the time period between setting the expected offer price, and the final offer date (Hanley, 1993). In this period the underwriter conducts a “road show”, where the investment banker screens the market for potential investors and gathers information as to how many shares they would be willing to buy at what price. This information is essentially used to produce the “book”. The book contains information of preliminary bids by potential investors. These investors can submit three types of bids: strike bids, limit bids and step-by-step bids. An investor, who wants to buy shares regardless of the final issue price, places a strike bid. A limit bid gives the maximum price an investor is willing to pay for shares and a step-by-step bid represents an investors demand curve, specifying at which price the investor is willing to buy how many shares. An example of a finished book can be found in table 2 of “Bookbuilding and Strategic Allocation” by Cornelli and Goldreich (2001), p. 2342. After finishing the book, the investment banker uses the bidding information to set the final issue price. The issue price is set so that the demand of shares exceeds the number of shares issued, assuring oversubscription. 2.2.1. The Partial Adjustment Phenomenon Even though bookbuilding was introduced in order to reduce underpricing and set the issue price nearer to the true value of the stock, empirical data shows no decline in underpricing since the introduction of the IPO method in the 90s (Loughran & Ritter, 2004). Researchers started to question this phenomenon and came up with several explanations as to why this was the case. One of the main theories behind explaining the levels of underpricing despite the bookbuilding method is the “Partial Adjustment Phenomenon”, introduced by Ibbotson, Sindelar and Ritter in 1988 and further researched by Hanley in 1993. Theoretically, bookbuilding should decrease levels of underpricing because it decreases information asymmetry in the market and better reflects investor interest in the issuers stock. For example, if the market shows positive interest in the IPO, underwriters can upwardly revise the offer price, putting the offer price closer to its true value. However, this price revision directly affects the level of underpricing. Hanley (1993) finds that positive price revisions in the offer price actually increase the level of underpricing, instead of decreasing it.

The problem lies within the signaling of investor interest to the market through price revisions. If the price is revised upwards, this signals investor interest to the market. It signals that the stock is highly valued, and therefore more investors place bids on the issuers’ stock. This is only true if investors truthfully reveal their investment interests. As an investor with a positive investment interest, you rather not reveal this positive interest since it can decrease your returns. This is because investors benefit from underpricing, whilst the issuer does not. If you show your interest in the issuers stock, the underwriter will revise the offer price upwards, decreasing your potential returns. Therefore, a problem lies within truthfully extracting information from the market by the underwriter. Benveniste & Spindt (1989) develop a model where they show how underwriters truthfully extract investor information from the market. The model shows that a tradeoff between share allocation and information exchange exists between the underwriter and the potential investor. Investors who truthfully show their positive investors interest, are allocated a larger amount of the issued shares by the underwriter. Investors will do so if, the profits from the larger amount of allocated shares exceed the losses in terms of underpricing caused by the upward price revision. The investor benefits by receiving a larger share of smaller overall returns. The problem lays within the fact that empirical papers show that upwards price revision leads to higher levels of underpricing, instead of lower ones. Therefore, investors who show positive investment interest receive even higher returns: an increase in allocation of shares and an increase in underpricing. Hanley (1993) shows that first day returns of firms going public are respectively 20.7% for upward revised offer prices, 10.0% for offer prices within the anticipated range and 0.6% (economically insignificant) for offer prices below the anticipated range. The phenomenon is termed the partial adjustment phenomenon because the price is raised, but not fully towards its true market value. The remaining part of price fulfillment will be reached through underpricing. Hypothesis (1): Considering Hanley’s findings on the partial adjustment phenomenon, and the vast research done on underpricing, this study expects to find a causal relationship between price revision and underpricing. One where upwards price revision leads to higher levels of underpricing, due to the signaling to the market of high investor interest. Hypothesis (1) is therefore formulated as: H1: Upwards price revision leads to higher levels of underpricing, as according to the partial adjustment phenomenon

2.3 Venture Capitalists Recent literature has shown contradictory effects of VC presence in IPOs. VCs could contribute to revealing the true value of an issuers’ stock and therefore maximizing such firm’s returns and minimize the level of underpricing. VCs have the intention of targeting high-growth potential firms and assisting in the growth process. Since VCs are mainly targeting small and medium enterprises (Kitchen, R. 1989), they are simply better at valuing start-up firms than investment banks. As VCs assist in the valuation of the company, they could decrease underpricing. Fried & Hisrich (1994) study VC investment criteria and the investment decision process. They argue that VCs follow a six-step investment process in which they assess whether the investee is worth investing in or not. On average, it takes 97.1 days for an investee to pass all six stages and receive funding. The study shows the extensive thought process conducted by VCs and indicates that VCs are good at targeting undervalued firms and finding their true value. Megginson & Weiss (1991) examine this phenomenon by deploying a matched pairs methodology on the return and performance of VC-backed IPOs versus non-VC backed IPOs. By matching VC-backed to non-VC backed IPOs on three-digit SIC code and offering amount, they find that VC backed IPOs experience lower levels of underpricing, attract better underwriters and receive greater interest of institutional investors. They find consistent evidence with the certification hypotheses presented by Booth and Smith (1986), which states that VCs have the intention of certifying an issuer’s true value and maximizing IPO returns. This is because VCs have a long-term interest in an underlying investment and want to maximize returns for potential future investment opportunities. Megginson & Weiss also find that VC backed IPOs attract more experienced underwriters. Underwriters in VC-backed IPOs have an average market share of 4.4%, in comparison to 3.0% for non-VC backed IPOs. The lead underwriters of VC backed IPOs brought 9 more IPOs to the market on average than lead underwriters of non-VC backed IPOs. They also find that underpricing averaged 7.1% for VC-backed IPOs against 11.9% for the matched non-VC backed IPO sample. This difference is significant at the 1 percent level. Contradicting results are found by Lee & Wahal (2004), who find higher levels of underpricing for VC-backed IPOs in comparison to their matched non-VC backed IPOs. They argue that previous studies tend to be wrong since they do not control for the endogenous choice of investment by VCs. If the distribution of investments of VCs were random, previous studies might be correct. Lee & Wahal argue that this is not the case and that this selection bias needs to be accounted for. Lee and Wahal control for this selection bias by conducting a first stage regression to predict the receipt of venture funding (a method previously conducted by Dehejia & Wahba, 1999, who looked at the evaluation of training programs in analyzing their effect on post-intervention earnings), after which they feed the estimates into methods of matching VC-backed IPOs to non-VC backed IPOs. This method is known as propensity score

matching. Instead of matching on firm characteristics (as done by Megginson & Weiss, who match on SIC code and total offering amount) they match on the non-endogenous propensity score (the predicament of VC backing). The propensity score represents the predicted probability of receiving VC backing, estimated using firm characteristics. This method is further explained in the methodology section. Lee & Wahal generate propensity scores based on three-digit SIC codes and log of net proceeds. After controlling for the selection bias they find that levels of underpricing tend to be significantly higher for VC backed IPOs compared to non-VC backed IPOs, findings contradicting previous ones. They find that underpricing is between 5.0% and 10.3% higher for VC-backed IPOs. Their results are consistent with the Grandstanding Hypothesis introduced by Gompers in 1996. The grandstanding hypothesis states that VCs have a limited time frame in which they must return proceeds to investors, because venture capital funds are usually organized as limited partnerships with a fixed lifetime (Gompers 1995). Since the larger proportion of returns of VCs come from IPOs, VC firms depend on their reputation regarding their ability of taking firms public. Since taking a firm public signals quality and increases their reputation, which critical for future fundraising – VCs are willing to bear the cost of underpricing. According to this theory, younger VC firms should accept higher levels of underpricing since their reputation is still to be built. Lee and Wahal confirm this after testing the differences in underpricing between two sub-samples of VC-backed IPOs that differ in reputation and experience of the underlying VC. They find that younger and less reputable VCs engage in IPOs with higher levels of underpricing. 2.3.1 VC Investment Interests In order to choose a party regarding the research done on the effect of VCs on underpricing, a closer look into VCs’ investment interests needs to be taken. VCs are active investors, whose intention is to add value through active involvement in the underlying business plan. They tend to invest in younger companies and take a leading role in working with these companies (Gompers, 1995). The backing VC makes use of its network, assets and experience to help the investee grow and increase in value. Furthermore, VCs often have one or more representatives serving on the companies’ board of directors. Also, as opposed to buyout funds, VCs take minority positions in companies and do not use leverage to finance their position (Kaplan & Sensoy, 2015). VCs hold a strong position in equity after the IPO (34.3% of equity before the IPO and 24.6% afterwards on average). Also, 58% of all VCs don’t even sell any shares in the offering. Keeping their equity position signals the strong belief of VCs in their underlying investments. The presence of VC representatives in board positions of investees also indicates active participation in the growth of their investees. 85% of issuers have a VC representative sitting on their board at the time of the IPO and continue to hold their board positions for more than a year

afterwards. This indicates active monitoring and long-term interest of VCs in their investments (Barry et al., 1990). Brav & Gompers (1997) investigate the market reaction to VC backed IPOs against non-VC backed IPOs. They argue that, if VCs indeed have a long-term interest in the underlying investment, and conduct active participation, the market should value this and VC backed IPOs should outperform non-VC backed IPOs. They look at IPO performance and find indeed that VC backed IPOs do outperform non-VC backed ones in the long run (five years) on average. Indicating that VCs are not using IPOs as an exit strategy to sell of their underlying investment. Cumming (2008) shows that VCs have high control rights when they use convertible debt or convertible preferred equity in financing target firms. He finds that convertible preferred equity-financed firms have a 12% lower likelihood of an IPO exit; which is in line with the inclined theory that VCs have a long-term interest in their underlying investment by having higher control rights. The same line of reasoning is used by Kaplan & Strömberg (2003), who also find that VC firms invest in convertible preferred equity. This results in higher cash flow rights and ownership stakes, also indicating more interest in the long-term performance of investments. Gompers et al. (2016) survey 885 VCs to learn more about VCs investment decisions. How they monitor their portfolio after investments are made and what VC exit strategies look like. For example, their study showed that 87% of surveyed VCs provide strategic guidance to their investments, post-investment. Kaplan & Strömberg (2004) show that VCs have the intention of adding value to the business before making an investment decision. Gompers et al. (2016) find that 60% of surveyed VCs meet at least once per week with their portfolio companies and that 72% help their companies network with potential future investors. Indicating that VCs use their expertise to help their portfolio companies be successful. Finally, 58% of VCs provide aid in hiring board members whilst 46% provide aid in hiring employees. Hypothesis (2) Considering that the majority of the research presented indicates that VCs participate in active monitoring of the underlying investment and do not use IPOs as one of their main exit strategies, the assumption is made that VCs have a long-term interest in their investments. Therefore, this study expects to find the following effect of VC presence on underpricing: H2: Due to the long-term investment interest of VCs, their presence in IPOs will reduce underpricing in order to maximize capital gains.

VC backed firms to endure lower levels of underpricing, since VCs want to make sure issuers get the maximum returns possible from the IPO in order to finance future investment possibilities thus minimizing this opportunity cost. Hypothesis (3) Taking the partial adjustment phenomenon into account – which states that upwards price revision leads to higher levels of underpricing – and the expected effect of VCs on underpricing, this study expects to find the following effect on price revision: H3: VC presence will decrease upwards price revision, in order to reduce underpricing and maximize capital gains. 2.4 The Technology Sector The growing presence of the tech industry over the past decades has increased productivity and efficiency worldwide. Technology has reached levels where jobs are being automatized, production is being changed entirely and new concepts in the business world are introduced on a monthly basis. Look at self-driving vehicles, face recognition technologies or the introduction of brain implants in the medical world (Relegado, 2017). Duysters & Hagedoorn (2000) report that technological core capabilities are necessary in order to generate significant performance differentials. Griliches (1998) indicates the importance of innovation and technology by finding a strong relationship between R&D investment and firm productivity. The technology sector distinguishes itself from other sectors due to its rapid changes, importance of innovation, high competition and uncertainty in future growth possibilities. The Herfindahl Index can be used to measure the level of competitiveness within an industry (Nawrocki & Carter, 2010; Rhoades, 1993). The Herfindahl Index (HHI) is calculated by summing the squared market shares of all firms within an industry and will vary between zero and one. Where a value of one indicates the lowest level of competition and represents a monopoly. 𝐻𝐻𝐼 = 𝑠_!! ! !!! Nawrocki & Carter show that the Herfindahl Index is significantly lower in the software and application industry, the telecommunications industry and the computer industry (compared to the energy and oil industry). This indicates that the used sample should have significantly higher levels of competition. The tech sector has been researched in terms of productivity and M&A performance. However, there is a lack of research concerning underpricing in this sector. Given

the growing importance of technology, the decision is made to focus this study on the tech sector only. Note that this study does not intend to compare levels of underpricing, price revision or venture capital effects with other sectors. It merely limits itself to the tech sector for data handling reasons and the importance of the sector. 2.5 IPO Cycles IPO cycles refer to swings in IPO market volume. Lowry & Schwert (2002) find a pattern in IPO volume and returns between 1960 and 2001. Periods of high returns are followed by increases in IPO activities, which in turn are followed by periods of lower returns, therefore showing cyclical movement in the market. They find that IPO activity increases after periods in which underpricing is at its highest. This pattern is deemed unusual, since issuers tend to obtain the highest returns possible in an IPO, and underpricing is interpreted as an opportunity cost. Lowry and Schwert argue that the level of underpricing in the market at the time of filing an IPO, is not related to the level of eventual underpricing of the issuer. They argue that the reason IPO activity increases following periods of high underpricing, is because underpricing is related to positive information leakage during the registration period of those offerings. This is in line with the partial adjustment phenomenon presented earlier. Since the return in the market at the time of the IPO contains no information about the issuer’s eventual returns, there is no evidence that companies can achieve lower levels of underpricing by filing going public during times of low returns. Yung et al. (2008) find the same positive correlation between IPO volume and underpricing. They argue that the higher levels of underpricing are a consequence of more bad quality firms entering the IPO market due to a positive shock to the economy: “Consider a positive shock to the economy. Improving investment opportunities raise the price at which a fixed cohort of firms would be able to sell securities. These higher prices increase the temptation of bad firms to pool. In equilibrium, more bad firms do pool.” - Yung et al. (2008, p. 193) The clustering of IPO volume has been explained by Brailsford et al. (2000) and Ibbotson & Jaffe (1975). They argue that issuers can time their IPOs accordingly to the market in order to maximize the amount of capital raised. Helvege & Liang (2004) use investor optimism to explain IPO clustering. They find that the clustering of IPO volume is unilateral across all industries. Hot markets are not characterized by clustering in a single industry but occur when many industries have hot markets. They argue that hot markets occur when there are higher levels of investor optimism, a view supported by Ljungqvist, Nanda and Singh (2006).

Hypothesis (4) Considering the increase in investor optimism in hot IPO periods, this study expects to find a certain correlation between market heat and price revision. H4: Prices will be revised more upwards during hot IPO markets, than during cold ones. This is due to the increased investor optimism, captured by underwriters in the bookbuilding period. Hypothesis (5) Considering the means of VCs to maximize capital gains of their investees in IPOs, this study expects that their presence will reduce the overall effect of market heat on IPO returns. H5: VC backing in IPOs will decrease the effect of market heat on underpricing, this is because VCs have the intention of minimizing the opportunity cost of underpricing.

CHAPTER 3

DATA DESCRIPTION

The sample consists of data collected on IPOs occurring between the first of January 1992 and the first of January 2018. IPO data has been collected from the Thomson Reuters SDC database. The dataset consists of only completed IPOs occurring in the Tech sector. Firms going public within the Tech sector are those identified with the following primary SIC codes: 372, 376, 357, 283 and 36 as according to Cloodt et al. (2006). Utility firms, financial firms, real estate companies and partly government owned firms are therefore excluded. Using SDC, information has been retrieved on variables as first trading day, issuer name, offer price, shares offered, filing price, offer price, a dummy variable indicating VC backing and company financials. The dataset is complemented with stock data using the Center for Research in Security Prices (CRSP) database. Such as opening and closing prices on the first day of trading, size of overallotment option and an indicator for the stock exchange the issuer is trading on after the IPO. IPOs with an opening price beneath 4$ have been excluded. Underwriter rankings and information on company founding dates have been retrieved from Loughran & Jay Ritters’ (2004) dataset as well as Field & Ritters (2002) dataset. Rankings from 2014 have been assumed to representative for current rankings. The rankings are an adjusted ranking system based on the Carter & Manaster (1990) rank. Rankings range from 0 to 9, where a higher ranking indicates a higher quality underwriter. Missing values on company financials as assets have been supplemented using the Compustat database. Data on market returns on IPO dates

have been retrieved from Yahoo Finance. Finally, information on size and age of VCs has been hand collected from the VentureXpert database. Also, dummy variables have been included indicating whether the IPO occurred during a boom or recession. In order to recognize business cycles in the US, the Leading Index for the United States (USSLIND) has been used as a variable for boom and recession recognition. Booms are recognized as periods with positive USSLIND, whilst recessions are recognized as periods with negative or zero USSLIND growth. Hot and cold IPO markets have been recognized using two measures: one by level of underpricing, where underpricing above 25% indicates a “hot” IPO market, as done by Helwege & Liang (2004). The second where we denote an IPO occurring in a “hot” period if in that year the number of IPOs has exceeded the mean yearly value for the entire dataset. VC backed IPOs are also attained a dummy equalling one if the VC is categorized as a “high reputation VCs”. The dummy is attained a value of one if (1) the age of the VC is in the second or above quartile, (2) the number of companies the VC invested in is in the 3rd _{or above quartile and (3) the total sum on investment} done by the VC (in dollars) is in the 3rd _{or above quartile. Market returns on the day of the IPO} have been retrieved from Yahoo Finance. The final dataset consists of 524 IPOs. Data has been winsorized at the 1% level. Descriptive statistics are shown below in table 1, and a full explanation of each variable has been provided in Appendix A. Underpricing has been calculated as: !!!!! !! , where 𝑃! equals the closing day on the first day of trading and 𝑃_! the offering price. Price revision is calculated as: !!! !! !! , where 𝑃! equals the midpoint of the filing range. Average underpricing in the sample equals 17.7% and average price revision equals minus 1.8%. Average price revision of minus -1.8% seems fairly low, considering that we would expect this number to be positive. The mean lies so close to zero because the observations of price revision follow almost a perfect normal distribution, with the mean around zero (Appendix B). Upward revision observations are offset by downward revision observations. This doesn’t mean that price revision does not affect underpricing though. Price revision dummies indicate whether there has been upward, downward or no price revision in the IPO process. Hanley’s (1993) cut-off point is used, where an offer price within the initial price range indicates no revision. The upward or downward revised dummies are attained a value of 1 if the final offer price was above or below the initial price range. 57.6% of all observations undergo no price revision, 18.5% is upward revised and 23.9% downward. Average age of issuers going public equals 11.5 years and 333 out of 524 IPOs are backed by a VC. 118 out of 524 IPOs are backed by a “high reputation” VC. 517 out of 524 IPOs occur in a boom.

The table shows no real irregularities. Underpricing of 17.7% is in line with most previous findings. The mean value of underwriter ranking indicates that most IPOs have a high rated underwriter, which could also be an explanation for higher levels of underpricing according to the analyst lust theory (Cliff & Denis, 2004). The fact that most observations show negative price revision though, could indicate that there is not a positive relationship between price revision and underpricing. Nonetheless, in order to make such a conclusion, several analyses must be done. Since this study looks at the effect of VC backing on underpricing and price revision, the mean values of some of the most important variables between the two samples are compared. Table 2 shows these differences and includes the p-values of the t-statistics.

The p-values indicate that there is a significant difference in values of all variables except for the log of total assets. The table shows for example, that the level of underpricing (first day returns) is significantly lower in the VC-backed sample (12%) compared to the non-VC backed sample (27%). There are lower levels of upwards price revision (-5%) for the VC-backed sample compared to the non-VC backed sample (4%). Anyhow, since the table has not yet applied propensity score matching, hypothesis one cannot yet be confirmed. It seems to indicate that there is lower underpricing for VC-backed IPOs though. The table shows that VCs indeed prefer to invest in younger companies (9.49 years old on average) and that the waiting period for VC-backed IPOs is significantly shorter than non-VC backed IPOs. The higher ranking of underwriters also indicates that VC-backed issuers more efficiently and more successfully enter the IPO process, finish it faster and generate greater returns. Even though hypotheses cannot yet be confirmed, the table shows promising results for our final findings.

Chemmanur et al. (2011) show that VCs target firms with high growth potential and already outperform non-VC backed firms, pre-investment. They find that VC firms are more efficient and that total factor productivity (TFP) grows faster for firms who receive financing from VCs compared to non-VC backed firms. The increase in efficiency and productivity is mainly due to the monitoring role VCs take as explained in chapter 2. Although the increase in efficiency is partly due to their monitoring, it is also the case that VCs are good at screening potential investees. Chemmanur et al. find that low reputation VCs increase value of investees mainly by screening the right targets, whilst high reputation VCs mainly contribute in terms of monitoring. VCs tend to improve efficiency by increasing sales, increase in total production costs (indicating higher production), increase in wages for VC backed firms compared to similar non-VC backed firms and thus overall product market performance. Some insight into the effect of VC experience on underpricing and price revision is given in the results section. Though not a main area of interest for this study, it gives some insight into the effect of VCs on these variables. The descriptive statistics show that about half of all VC backed IPOs were backed by a high reputation VC. In order to examine differences across hot and cold IPO markets, table 3 is created. The table shows the differences in the mean values for hot and cold IPO markets. Panel A describes an IPO as being in a Hot market if the underpricing was above 25% (Helwege & Liang, 2004), panel B uses the mean value of IPO volume per year as a threshold for hot and cold IPO markets.

Both panels indicate that in periods of hot market activity, issuers tend to go public at a younger age than at times of cold periods. Panel A indicates that, if the underpricing threshold is used, there is more downwards price revision in cold IPO periods than in hot ones (-5% against +10%). Since the % of VC-backed IPOs remains similar, it could indicate that VC presence

decreases upwards price revision. Which is interesting, is that in panel B, there are higher levels of underpricing for cold IPO periods than hot ones. Yet, this difference is insignificant. Even though insignificant in panel B, the results also show lower levels of upwards price revision in cold IPO markets. Also, the waiting period is considerably longer during cold IPO periods (as indicated by both panels). Percentage of VC backed IPOs is larger in hot periods in both panels, yet doesn’t differ much across panels. This indicates that VC investment decisions might be slightly driven by hot IPO markets. Which is interesting is that changing the IPO threshold changes the number of observations in each sample drastically. In panel A only 115 out of 524 observations are considered “hot” whilst this number increases to 418 in panel B. Finally, figures 1 and 2 show IPO volume, underpricing and price revision over the years. Figure 1 indicates that there are notably less IPOs during the financial crisis (2008 – 2009) and after the dotcom bubble (2001 -2002). Figure 2 shows higher levels of underpricing as IPO activity increases. Which is clearly shown by the peak just before the dotcom bubble. Which is interesting though, is that this pattern isn’t repeated in 2014, where IPO activity increases, yet underpricing and price revision remains the 0 20 40 60 80 To ta l n u mb e r o f IPO s 1990 2000 2010 2020 Year

Figure 1: IPO Volume per Year

-. 5 0 .5 1 1.5 *1 0 0 (% ) 1990 2000 2010 2020 Year

Underpricing (%) Price Revision (%)

same. Perhaps history has shown that issuers tend to learn from their mistakes, and try to avoid being underpriced too much during hot IPO periods. If price revision and underpricing are correlated, similar patterns in the levels of underpricing and price revision in figure 2 should be seen. However, a clear correlation cannot be identified solemnly by this graph. It could be the case that such a relationship cannot be identified because observations of upward price revision are offset by observations with downward price revision. Therefore the decision is made to create the same table where only observations with upward price revision are included. Figure 3 shows the outcome. Figure 3 shows a slight correlation between underpricing and upwards price revision. Nonetheless, it doesn’t show a clear pattern. In order to estimate a causal relationship, OLS Regressions and Propensity Score Matching methods need to be conducted. Results of these methods are shown in chapter 5: Results.

Overall the descriptive statistics show that there are significant differences between the VC backed sample as the non-VC backed sample. Indications of less upwards price revision and underpricing for the VC backed sample are found, indicating that hypotheses 2 & 3 might be confirmed. The descriptive statistics also show that VCs tend to invest in younger and smaller companies and also smoothen the IPO process (indicated by a smaller waiting period and higher underwriter ranking). Significant differences in hot and cold IPO markets are found using two different thresholds. Hot IPO markets tend to have higher levels of upwards price revision and underpricing, yet a non-differential amount of VC backing in both samples indicates that their choice of investment is not reflected in their post-IPO intentions of proceeds. If higher levels of VC backing were to be shown in panel A, that would indicate that VCs contribute to higher levels of underpricing, as opposed to lower ones. A final observation from this dataset is that all previous findings stated in chapter two, are representative in the dataset. Be that as it may, multivariate regression analyses and several econometrical methods need to be conducted in order to examine the effect of VC backing on price revision, underpricing and differentials between hot and cold IPO markets. 0 .5 1 1.5 2 /1 0 0 (% ) 1990 2000 2010 2020 Year Price Revision (%) Underpricing (%)

CHAPTER 4

METHODOLOGY

Following the main methodology of Lee & Wahal (2004) and Hanley (1993) this paper estimates the effect of VC backing on underpricing through price revision. This is done in four steps. First, it establishes a causal relationship between price revision and underpricing. (1) An OLS regression model is conducted with underpricing as dependent variable. OLS is used instead of Estimation Variables (IV) or Generalized Method of Moments (GMM) due to the lack of panel data and truly exogenous instruments (Wintoki et al., 2012). The method used expands on Hanley’s method by including a measure of underwriter ranking. Findings are checked and corrected for multicollinearity issues. (2) Next, this study looks at the effect of VC backing on price revision. An OLS regression model with price revision as a dependent variable is conducted. Once again the model is corrected for multicollinearity issues. This effect is further examined by conducting a diff-in-diff method where three different propensity score matching models are applied. This is done in order to control for a selection bias, a method similar to the one conducted by Lee and Wahal. (3) Price revision and VC backing are directly connected to underpricing by conducting an OLS regression and diff-in-diff methods. The OLS regression includes interaction terms between VC backing and price revision. Both the continuous variable of price revision as binary dummy variables representing upwards or downward revision are included in the model. The effect of VC backing on underpricing is also examined by conducting a diff-in-diff using different models of propensity score matching. (4) Finally, the results are checked for differences between hot and cold IPO markets. Two measures to define a “Hot” IPO method are used, as explained in the data description section. Once more, an OLS regression is conducted, where dummy variables of IPO heat and interaction terms with VC backing are included. Finally, a diff-in-diff method using various models of propensity score matching is conducted. 4.1 The Causal Relation between Price Revision and Underpricing To examine the effect of price revision on underpricing, an OLS regression (I) is conducted. The method followed is the one similar to Hanley’s (1993), where price revision is regressed on underpricing. Underpricing has been calculated as:!!!!! !! , where 𝑃! equals the closing day on the first day of trading and 𝑃_! the offering price. Price revision is calculated as 𝑃!_{− 𝑃}!_/𝑃!_{, where 𝑃}! equals the opening stock price on the first trading day and 𝑃! _{equals the midpoint of the filing} range, calculated as the average of the high and low filing range prices, as according to Hanley (1993). Extending on Hanley’s method the model includes a variable that measures the ranking

of the underlying underwriter. To control for heteroskedasticity, robust standard errors are applied. 𝐼 𝑈𝑛𝑑𝑒𝑟𝑝𝑟𝑖𝑐𝑖𝑛𝑔 = 𝛽!+ 𝛽!𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛 + 𝛽!𝐼𝑠𝑠𝑢𝑒𝑑𝑆ℎ𝑎𝑟𝑒𝑠 + 𝛽!𝑂𝑣𝑒𝑟𝑎𝑙𝑙𝑜𝑡𝑚𝑒𝑛𝑡𝑂𝑝𝑡𝑖𝑜𝑛 + 𝛽!𝐿𝑜𝑔𝑜𝑓𝑁𝑒𝑡𝑃𝑟𝑜𝑐𝑒𝑒𝑑𝑠 + 𝛽!𝑅𝑒𝑡𝑢𝑟𝑛𝑜𝑛𝑀𝑎𝑟𝑘𝑒𝑡 + 𝛽!𝑊𝑎𝑖𝑡𝑖𝑛𝑔𝑃𝑒𝑟𝑖𝑜𝑑 + + 𝛽!𝑅𝑒𝑡𝑢𝑟𝑛𝑜𝑛𝑆&𝑃500𝐼𝑛𝑑𝑒𝑥 + 𝛽!𝑈𝑛𝑑𝑒𝑟𝑤𝑟𝑖𝑡𝑒𝑟𝑅𝑎𝑛𝑘𝑖𝑛𝑔 + 𝛽!𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝐴𝑠𝑠𝑒𝑡𝑠 + 𝛽!"𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝐸𝑞𝑢𝑖𝑡𝑦+ 𝛽!!𝐼𝑠𝑠𝑢𝑒𝑟𝐴𝑔𝑒 + 𝜀! There might be some endogeneity going on between the log of net proceeds, size of overallotment and total shares issued though. Testing for multicollinearity will be done using VIFs, where a threshold of below 4 is determined as having no multicollinearity issues. If the results prove to be endogenous, the model will be adjusted accordingly. 4.2 The Effect of VC Backing on Price Revision A. Multivariate OLS Regression Model To examine the effect of VC backing on price revision, an OLS regression similar to equation (I) is conducted. The dependent variable is price revision and a dummy variable for VC backing is included. Robust standard errors are used and the model is being checked for multicollinearity issues using variance inflation factors. The initial model (not yet corrected for multicollinearity) is represented by equation (II). If the model turns out to show multicollinearity issues (for example between number of shares issued and log of net proceeds), the model will be adjusted accordingly. 𝐼𝐼 𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛 = 𝛽!+ 𝛽!𝑉𝐶𝐵𝑎𝑐𝑘𝑒𝑑 + 𝛽!𝐼𝑠𝑠𝑢𝑒𝑑𝑆ℎ𝑎𝑟𝑒𝑠 + 𝛽!𝑅𝑒𝑡𝑢𝑟𝑛𝑜𝑛𝑀𝑎𝑟𝑘𝑒𝑡 + 𝛽_!𝑅𝑒𝑡𝑢𝑟𝑛𝑜𝑛𝑆&𝑃500𝐼𝑛𝑑𝑒𝑥 + 𝛽_!𝑈𝑛𝑑𝑒𝑟𝑤𝑟𝑖𝑡𝑒𝑟𝑅𝑎𝑛𝑘𝑖𝑛𝑔 + 𝛽_!𝐿𝑜𝑔𝑜𝑓𝑁𝑒𝑡𝑃𝑟𝑜𝑐𝑒𝑒𝑑𝑠 + + 𝛽!𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝑅𝑒𝑣𝑒𝑛𝑢𝑒 + 𝛽!𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝐴𝑠𝑠𝑒𝑡𝑠 + 𝛽!𝐼𝑠𝑠𝑢𝑒𝑟𝐴𝑔𝑒 + ε! B. Diff-in-diff Analysis: Propensity Score Matching Next, Lee & Wahal’s (2004) method is followed, who look at the effect of VC backing on IPO underpricing. A diff-in-diff method is conducted where levels of price revision in VC-backed IPOs are compared to those of non-VC backed IPOs. By calculating a simple t-statistic between the VC-backed and non-VC backed sample, a significant difference between the levels of price revision in the two groups can be determined. However, there is an endogeneity concern going on that should be taken into account. As mentioned in chapter 2, VCs do not randomize their investments. Ang & Sorensen (2012) recognize the selection bias in VC investments. They state that VC investments are done in multiple rounds and that better performing companies tend to

receive more investment rounds than their underperforming competitors. Companies with better un-observed characteristics receive funding by more experienced VCs (Sørensen, 2007). Ideally, one would examine a difference in underpricing between an issuer that went public without a VC backing the company, versus the level of underpricing for the same issuer had it had VC backing. Since this is impossible (an IPO can only occur once) a matching method needs to be applied where VC-backed IPOs are matched to similar non-VC backed IPOs. In order to control for this non-randomization of investment decisions, propensity score matching is applied to match VC-backed IPOs to similar non-VC backed IPOs. Propensity score matching is a matching method where we estimate the predicted probability of treatment (in this case VC backing) using background characteristics of each observation (Dehejia & Wahba, 2002). You can choose on which background characteristics you’d like to match. The decision is made to match on the log of net proceeds and the issuer’s age at time of IPO. Probit regressions with VC backing as dependent variable are then conducted to estimate the likelihood of VC backing of IPOs. (III) 𝑃𝑟𝑜𝑏𝑖𝑡(𝑇 = 1|𝑋!, 𝑋!, … , 𝑋!) Where T = 1 if the IPO is backed by a VC and 0 if not. The X’s represent background characteristics of the underlying issuer. The predicted probability is called the propensity score. Finally, non-VC backed observations are matched to VC-backed observations based on their propensity score. Three different models of propensity score matching are used to examine the differences in price revision between the matched samples. (1) Nearest Neighbour Matching (in its 1 on 1 form) matches each treated observation to its control observation with the smallest distance. This study uses a 1 on 5 form of nearest neighbour matching, which means that an observation in the treatment group (VC backed) will match to its 5 nearest observations based on propensity score in the control group (non-VC backed). (2) Radius Matching matches each treated observation to a control observation within a specified radius. A radius of 10% is applied, meaning that each observation in the treatment group will be matched to an observation in the control group which has a propensity score lies within a radius of 10% near the propensity score of the treated observation. (3) Kernell matching, which is basically the same as radius matching, only attains higher weights to observations closer to the propensity score of the treated observations (Stuart, 2010). Lee and Wahal match on net proceeds and primary SIC code. Since the used sample consists of mainly the same primary SIC codes (tech sector), the decision is made to match on the log of net proceeds as of a quarter before the IPO and firm age on issue day. In order to deem

the propensity score matching method efficient, the matched samples need to overlap in terms of distribution. Figure 4 shows the normal distributions of the two matched samples. The propensity scores of both samples show approximately the same normal distributions. Both have a mean value of 0.75 and the shape of the normal propensity score is the same. Since they do not overlap perfectly: propensity score matching is helpful and not irrelevant. If there were a perfect overlap, propensity score matching wouldn’t be useful. 4.3 Effect of VC Backing on Underpricing A. Multivariate OLS Regression Model An OLS regression is conducted to analyse the effect of VC backing on underpricing through price revision. Two types of models are used: Equations (IV) and (V), which are both adjusted versions of equation (I) and (II). Model (IV) includes an interaction term between the binary dummy variable: VC backing and the continuous variable: price revision. Model (V) essentially shows the same, however it includes interaction terms between binary dummy variables for upwards or downwards price revision (based on Hanley’s method) and the binary dummy variable for VC backing. Both models are then tested for multicollinearity and adjusted accordingly. Robust standard errors are applied to control for heteroskedasticity. The dummy variable trap is avoided in model (V) because we omit a dummy variable for no price revision. 0 5 10 0 .2 .4 .6 .8 0 .2 .4 .6 .8

non-VC Backed VC Backed

Density

Kdensity Propensity Score Normal Propensity Score

D e n si ty Propensity Score

Graphs by psmatch2: VC Backed Assignment

𝐼𝑉 𝑈𝑛𝑑𝑒𝑟𝑝𝑟𝑖𝑐𝑖𝑛𝑔 = 𝛽!+ 𝛽!𝑉𝐶𝐵𝑎𝑐𝑘𝑒𝑑 + 𝛽!𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛 + 𝛽!𝑉𝐶𝐵𝑎𝑐𝑘𝑒𝑑 ∗ 𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛 + 𝛽!𝐼𝑠𝑠𝑢𝑒𝑑𝑆ℎ𝑎𝑟𝑒𝑠 + 𝛽!𝑅𝑒𝑡𝑢𝑟𝑛𝑜𝑛𝑀𝑎𝑟𝑘𝑒𝑡 + 𝛽!𝑅𝑒𝑡𝑢𝑟𝑛𝑜𝑛𝑆&𝑃500𝐼𝑛𝑑𝑒𝑥 + 𝛽!𝑈𝑛𝑑𝑒𝑟𝑤𝑟𝑖𝑡𝑒𝑟𝑅𝑎𝑛𝑘𝑖𝑛𝑔 + 𝛽!𝐿𝑜𝑔𝑜𝑓𝑁𝑒𝑡𝑃𝑟𝑜𝑐𝑒𝑒𝑑𝑠 + 𝛽!𝑊𝑎𝑖𝑡𝑖𝑛𝑔𝑃𝑒𝑟𝑖𝑜𝑑 + 𝛽!"𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝐴𝑠𝑠𝑒𝑡𝑠 + 𝛽!!𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝐸𝑞𝑢𝑖𝑡𝑦 + 𝛽!"𝐼𝑠𝑠𝑢𝑒𝑟𝐴𝑔𝑒 + ε! 𝑉 𝑈𝑛𝑑𝑒𝑟𝑝𝑟𝑖𝑐𝑖𝑛𝑔 = 𝛽!+ 𝛽!𝑉𝐶𝐵𝑎𝑐𝑘𝑒𝑑 + 𝛽!𝑈𝑝𝑤𝑎𝑟𝑑𝑠𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛𝐷𝑢𝑚𝑚𝑦 + 𝛽!𝐷𝑜𝑤𝑛𝑤𝑎𝑟𝑑𝑠𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛𝐷𝑢𝑚𝑚𝑦 + 𝛽!𝑉𝐶𝐵𝑎𝑐𝑘𝑒𝑑 ∗ 𝑈𝑝𝑤𝑎𝑟𝑑𝑠𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛𝐷𝑢𝑚𝑚𝑦 + 𝛽!𝑉𝐶𝐵𝑎𝑐𝑘𝑒𝑑 ∗ 𝐷𝑜𝑤𝑛𝑤𝑎𝑟𝑑𝑠𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛𝐷𝑢𝑚𝑚𝑦 + 𝛽!𝑅𝑒𝑡𝑢𝑟𝑛𝑜𝑛𝑀𝑎𝑟𝑘𝑒𝑡 + + 𝛽!𝑈𝑛𝑑𝑒𝑟𝑤𝑟𝑖𝑡𝑒𝑟𝑅𝑎𝑛𝑘𝑖𝑛𝑔 + 𝛽!𝐿𝑜𝑔𝑜𝑓𝑁𝑒𝑡𝑃𝑟𝑜𝑐𝑒𝑒𝑑𝑠 + 𝛽!𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝐴𝑠𝑠𝑒𝑡𝑠 + 𝛽!"𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝐸𝑞𝑢𝑖𝑡𝑦 + 𝛽!!𝐼𝑠𝑠𝑢𝑒𝑟𝐴𝑔𝑒 + ε! B. Diff-in-diff Analysis: Propensity Score Matching To examine the effect of VC backing on underpricing, the same propensity score matching model is applied to underpricing as the one on price revision, as described in 4.2.2. 4.4 Differences in Price Revision and Underpricing between Hot and Cold IPO Markets Finally, differentials between hot and cold IPO markets are examined using two different measures of market heat. One based on underpricing, where underpricing above 25% is defined as a “hot” IPO period, and one where a hot market is defined as one where IPO volume exceeds the mean value of yearly IPO volume in the entire sample. Two OLS regressions are conducted including interaction terms between the two binary variables: VC backing and Hot IPO dummies. Dependent variables are underpricing and price revision respectively. Equation (VI) shows the model used to examine the effect of market heat on price revision and is essentially an altered version of equation (II). The model uses robust standard errors. 𝑉𝐼 𝑃𝑟𝑖𝑐𝑒𝑅𝑒𝑣𝑖𝑠𝑖𝑜𝑛 = 𝛽_!+ 𝛽_!𝑉𝐶𝐵𝑎𝑐𝑘𝑒𝑑 + 𝛽_!𝐻𝑜𝑡𝐼𝑃𝑂 + 𝛽_!𝑉𝐶𝐵𝑎𝑐𝑘𝑒𝑑 ∗ 𝐻𝑜𝑡𝐼𝑃𝑂 + 𝛽_!𝑅𝑒𝑡𝑢𝑟𝑛𝑜𝑛𝑀𝑎𝑟𝑘𝑒𝑡 + 𝛽_!𝑈𝑛𝑑𝑒𝑟𝑤𝑟𝑖𝑡𝑒𝑟𝑅𝑎𝑛𝑘𝑖𝑛𝑔 + 𝛽_!𝐿𝑜𝑔𝑜𝑓𝑁𝑒𝑡𝑃𝑟𝑜𝑐𝑒𝑒𝑑𝑠 + + 𝛽_!𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝑅𝑒𝑣𝑒𝑛𝑢𝑒 + 𝛽_!𝐿𝑜𝑔𝑜𝑓𝑇𝑜𝑡𝑎𝑙𝐴𝑠𝑠𝑒𝑡𝑠 + 𝛽_!𝐼𝑠𝑠𝑢𝑒𝑟𝐴𝑔𝑒 + ε_! Equation (VII) shows the regression model analysing the effect of market heat on underpricing. The model is an altered version of equation (I) and also uses robust standard errors.