IPO UNDERPRICING IN THE TECHNOLOGY
INDUSTRY, IS HISTORY REPEATING ITSELF?
by
Michiel van Oijen
s2056712
MSc Thesis International Financial Management January 2016
University of Groningen Faculty of Economics and Business
Supervisor: V. Purice Co-Assessor: H. Gonenc
1 This research tests whether the financial activities that drove the dot.com bubble to burst back in 2000 are similar in today’s situation. It speculates on a current technology bubble and formulates its argumentation based on the results that were described as the symptoms that contributed to underpricing and leaded to the formation and burst of the previous bubble. Underpricing averaged at a 24.6% in the technology industry during the years 2006–2015. The variables underwriter rank, leverage, VC-backed, age at the IPO and selling intensity are tested within this time period. Furthermore, this research controls for the effect of the financial crisis, size and profitability. The results indicate that there are several similarities with the sample period and the bubble of 2000. Although it is hard to prove or speculate the existence of a bubble with a small portion of previous indicators, this research suggests that the creation or existence of a technology bubble is plausible.
2 I. INTRODUCTION
Most of today’s giant technology companies or so-called “unicorns”1
are founded in Silicon Valley. At this moment, there are 90 U.S. out of the 144 global companies that earned the title of being a unicorn, all (U.S.) in sum with a total valuation close to $504 billion, which is about five times as many as three years ago (The Economist, 2015c). Some of these unicorns have grown so big that their current valuation is so high, that even top-rated companies like Facebook or Google could not afford to acquire these new born unicorns. For instance, two private companies Uber and Airbnb are well-known unicorns. The former is a six-year-old company introducing a new and cheaper taxi experience for non-professional taxi drivers, is valued at $51 billion2. Airbnb, a seven-your-old company, which turns people’s homes into hotels, is valued at $25.5 billion2
. Kim et al. (2008) found that especially the technology companies are difficult to value, main reasons being; large amounts of intangible assets, high capital investments (with a questionable value) and negative cash flows. Among these technology unicorns, there are many firms that went public or are considering an initial public offering (IPO) and have considerable initial growth forecasts. This despite the fact that some of these companies make little revenue, are not profitable or even worse, have substantial amounts of "negative" stockholders equity (Ritter and Welch, 2002; Bhattacharya et al., 2010). Furthermore, history (e.g. the 2000 bubble3 and more recent IPOs of Facebook and LinkedIn) tells that such companies are very popular among investors and hedge funds due to the attention of the media and press (Kim et al., 2008). Although, today’s IPO market has shrunk compared to the 1999-2000 ‘hot-IPO market’, it seems that more offerings for technology companies that will go public are close at hand. Nowadays, most of the companies tend to go public after a series of capital funding, generating close to $200 million capital on average per company prior to the IPO. However, due to these large amounts of capital and high valuations, some experts say the proliferation of these unicorns is a sign of the creation of a tech bubble (Christopher, 2015; The Economist, 2015a). The recent developments in the technology sector have a high resemblance with the happenings back in 2000, when the market was faced with the latest technology bubble. The consequences of a bubble can be devastating. In the post-bubble phase, many technology companies saw their value drop to zero. The prevalent failure of Internet firms followed the decline of the stock market in 2000–2002 (Goldfarb et al., 2007). It was without doubt the largest stock market collapse in the history of industrial capitalism (Cassidy, 2002; Mahar, 2003). Shortly after the bubble, Barber and Odean (2001) identified several market conditions that contribute to the formation of speculative bubbles, which are classified as: the availability of large amounts of capital, significant uncertainty regarding firm valuation, and an inexperienced but active investor clientele. With respect to the first two conditions, the importance and relevance in today’s environment has already been introduced. Furthermore, another symptom of a speculative bubble is that history tells
1 Any private company worth more than US$1 billion based on fundraising.
2
Valuations as of November 2015. Source: The Wall Street Journal: http://graphics.wsj.com/billion-dollar-club/ 3
3 that bubbles are often accompanied by periods of high technological innovation (Lansing, 2012; Kindleberger, 1989; Chancellor, 1999). Technology and innovation are both key instruments for today’s organizations and is developing faster than ever expected. As yet has been described, the increase of technology unicorns, together with the extreme high valuations as well as the large availability of capital, explains the anxiety of experts, as the question arises whether history is repeating itself again? In order to investigate the probability of a new technology bubble, this research will focus on the financial activities that happened in the technology bubble in the late 90’s, such as IPO underpricing, and test whether these activities are similar to today’s situation.
The rise and fall of the Nasdaq & New York Stock Exchange (NYSE) composites, together with the technology market capitalization in the last decade of the 20th century, evidently show the burst of the bubble. For this reason, the development of these composites will be shortly addressed. The evolution of the technology bubble (Internet bubble or dot.com period) started around 1995 and saw its implosion in the collapse of the NASDAQ and the NYSE in 2000. The NASDAQ is a stock exchange with a high share of technology firms, attracting many of the firms dealing with the internet or electronics. While the NYSE had much less penetration in the technology sector, in today’s world, it is achieving a better position in captivating technology companies, as more of them are listed on the NYSE. The Nasdaq Composite Index rose from 775.20 at the beginning of 1995 to 2,505.89 in January 1999. Not much later, it more than doubled from this point to its peak of 5,048.62 on 10th of March 2000 when the bubble started bursting. Eventually, two years later on September 2002 the NASDAQ closed at 1,185 (Figure 1).
Figure 1:
Time series chart of the U.S. stock exchange over the period 1995-2015. The light-grey line indicates the NYSE; the dark-grey line indicates the NASDAQ composite. The graph shows, besides the tech bubble, the fall of the NASDAQ and NYSE exchange in 2008, caused by the financial crisis. Source: DataStream Thomson Reuters.
0 1000 2000 3000 4000 5000 6000 0 2000 4000 6000 8000 10000 12000 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Ind ex NASDAQ Ind ex NYSE Years
4 The market capitalization is another indicator that shows the burst of the technology bubble. The technology industry saw a total value loss of close to $4.7 trillion in only two years’ time. The market capitalization peaked on the 10th of April 2000 at $5.23 trillion and dropped to $1.55 trillion on the 26th of September 2001 (Figure 2). This 18-month decline of stock prices resulted into $1.0 trillion loss in Silicon Valley’s 150 largest companies (Cassidy, 2002; Mahar, 2003). In the two years following the dot.com crash of the NASDAQ, there was approximate loss of $8.5 trillion in shareholders wealth (Stiglitz, 2003, p. 6). As can be derived from Figure 1&2, both the NYSE and NASDAQ, as well as the technology market capitalization show a clear upward trend line after the financial crisis of 2008. For some experts, this development is one of the explanations for their concern about the creation of a possible next bubble (La Monica, 2014; Rushe, 2015).
Figure 2:
Time series chart of the U.S. Technology Market Capitalization over the period of 1995-2015. Note: this graph does not take private companies into account. Source: DataStream Thomson Reuters.
A key element related to the collapse of the NASDAQ, NYSE, technology market capitalization and the existence of the dot.com bubble is known as the phenomenon of technology stocks IPO underpricing (first-day returns) (Barron, 2007). After the bubble, many attempts have been made to identify the reasons for the underpricing such as: underwriter ranking, market sentiments, age at the IPO, leverage, price earnings ratio, the offer size, promoters’ holding, selling intensity, VC-backed firms etc., which resulted into extensive literature, explaining the factors that have an impact on underpricing (Rani and Kashik, 2015). This research will investigate and focus on some of these factors that had a significant relationship with underpricing during the bubble and test how they
0 1.000 2.000 3.000 4.000 5.000 6.000 B illi o ns U. S.$ Years
5 developed in the years after it, starting from 2006 until 20154. These factors are identified as: underwriting rank, leverage, VC-backed, age at the IPO and selling intensity (Loughran and Ritter, 2002; Kim et al., 2008, Ljungqvist and Wilhelm, 2003; Liu and Ritter, 2011). The control variables are identified as: financial crisis, size and profitability. Furthermore, the number of technology IPOs and profitability differed highly in the ‘hot-IPO’ market compared to the periods before and after the bubble. This resulted into a theory that explains these characteristics as a contribution to the creation of the bubble. It is for this reason, that this research will discuss and compare the development of these characteristics. This paper does not develop a model that measures a (current) bubble. Instead, it conjectures on similarities of several determents related with underpricing. Though bubbles may be proven to exist in markets, they are still very difficult to predict with certainty in the present.
This research addresses the following research question: “Given the explicit details of the financial
activities that happened in the Internet bubble of 2000, is IPO underpricing for technology companies still an issue and can we possibly describe the era of a new bubble?”
In order to answer the main research question, several sub-questions have been formulated to underpin the argumentation of this research. The formulation of these sub-questions is as follows:
- Which factors had a significant impact on the underpricing of technology firms and were a
contributing factor to the collapse of the NASDAQ and NYSE composite in 2000?
- What are the main differences and similarities between technology IPOs back in the tech
bubble compared to the technology IPOs of today?
- Are we at the eve of a new technology bubble?
This research is distinctive in its field, because very little literature is written about the speculation of a new bubble, the development of the current technology underpricing of IPOs in the U.S., together with its comparison with the bubble of 2000. Perhaps the closest research to the current analysis are papers written by Wilhelm and Ljunqgvist (2003); Kim et al. (2008); Loughran and Ritter (2003); Liu and Ritter (2011). The outcome of this research can contribute towards new insights into the current technology industry for IPOs. There are several reasons why the main focus of this research is based on U.S. listed companies. First of all, the roots of the previous Internet bubble originate in the United States. Secondly, the literature primarily focuses on U.S. companies’ traits that caused the Internet bubble of 2000. Finally, most of the technology companies are founded in Silicon Valley, which results into the fact that of all technology unicorns, 62.5% are U.S. listed. The country in second place, China, accounts for a dozen of the unicorns, which is around 9.7% of the total technology industry. However, predictions are that eventually there will be more billion-dollar startups in China than in the
4
6 United States, though Silicon Valley will have more firms of higher value (The Economist, 2015b). Based on above mentioned reasons, the importance and relevance of using U.S. listed companies for an adequate comparison can be justified.
Although the choice has been made to focus on the U.S. stock market, the outcome of this research can be applied to other stock markets as well, ensuing cross-border perspectives. First of all, it is widely acknowledged that the financial globalization has resulted into the fact that the various markets and countries no longer trade independently. For example, the pattern of the NASDAQ during the bubble of 2000 shows a high similarity with the movements of other (e.g. European) stock markets over the same period (Hon et al., 2007). Secondly, as a result of the intense popularity and high levels of demand in U.S. technology stocks, most private investors and venture capitalists have a global presence (Schertler and Tykvová, 2009). In general, the technology industry is more likely to be financed by foreign venture capitalists than companies in other industries (Schertler and Tykvová, 2009). In the early ‘90s, around 10% of the U.S. venture capital deals had international participation, but climbed to 25% by the end of the Internet bubble (Aizenman and Kendall, 2008). The magnitude for the intra-continental and inter-continental VC-links can be found in Figure 6 of Appendix V. Thirdly, there are various SEC rules that have an European equivalent that issues regulations of a similar nature (the European coordinating authority is the CESR). Based on the above, it can be stated that the American technology bubble has an international aspect and that the outcome of this research has a potential reach that goes beyond the U.S. stock market.
This research is structured as follows. Section II describes the literature related to the IPO underpricing topic and the creation of the technology bubble. Section III describes the methodology, the data, including the descriptive statics and the conception of the regression formula. Section IV examines the model and which of the variables can best predict IPO underpricing. Section V consists of the conclusions related to the main research question. Section VI is used for the reference list and Section VII for appendices.
II. LITERATURE
7 II.I IPO Underpricing & Technology Industry
The pricing of IPOs has been broadly investigated, as the continued finding of positive first-day returns leads to a number of important questions about the reasoning and behavior of issuers, investors, and underwriters during and after an IPO process. Since it is difficult to determine the real value of a technology firm, it is as difficult to set an offer price for the IPO, which is normally a long-term (negotiation) process between issuer and underwriter. In the public offering, the price of the stock at the end of first-day will present the ‘fair’ value of the firm, as by then the market has priced it correctly. The first day closing price is an indicator of what value investors place on the firm: if closing price is higher than the offer price, then the share is considered to have been underpriced.
Within the extensive underpricing literature, a lot of attention is given to the patterns of technology underpricing in the late 90’s. In the bubble, the first-day returns of technology IPOs were remarkably high compared to other industries. Evidence of previous research finds that the first-day return on technology initial public offerings averaged about 17 percent, with a median of 10 percent in 1996. Three years later, in 1999, first-day returns averaged 73 percent, with a median of 40 percent before tapering off to 58 percent, with a median of 30 percent in 2000. The Internet IPOs, which is a specific segment of the technology industry, was booming during the bubble and averaged a striking 89 percent, with a median of 57 percent during 1999 and 2000 (Ljungqvist and Wilhelm, 2003). The mean and median returns of non-internet IPOs over the 1998 to 2000 period were 33.6% and 10.4% respectively. Of the 632 technology IPOs in 1999 and 2000, 29% saw the company double in value in the first day. The most extraordinary example is THEGLOBE.com company, who saw its value rise with 606% after the first day of trading on the NASDAQ (Perkins & Perkins, 2001; The Economist, 2015a). It offered 3.1 million shares at $9 per share, and the first-day closing was an astonishing $63.50 per share. Its market value was $622 million, even though the company lost $11.5 million in the first nine months of 1998 (Perkins and Perkins, 2001). As a result of the introduction of Internet in 1995 to the public, companies who were related in Internet operations saw their value exponentially increase. One of the explanations for this is that hedge funds and other investors all wanted a piece of the cake. Abbreviations as TINA (‘there is no alternative’) or FOMO (‘fear of missing out’) were used to describe what happened in the booming technology segment. The major shift in popularity resulted that companies purposely added the terms ‘Internet’ or ‘.com’ to associate their company name with this hype (Perkins and Perkins, 2001). In sum, the technology sector saw explosive optimistic investor behavior, which leaded to the proliferation and existence of underpricing during the bubble.
8 firms went public for opportunistic reasons, extracting a surplus of cash from these investors. This assumption can be derived in the growing number of public offerings related with the dot.com period. The number of technology IPOs in the US market rose especially during the years of 1999-2000, so much, that it was named as a ‘hot-IPO period’. Noticeably, the amount of IPOs and investments in these events grew heavily in this period (Ofek and Richardson, 2003). The start of a rise in technology offerings can be seen from 1995 onwards, the year in which the Internet saw its first appearance. This rise can be seen from the total technology IPOs in that period. In 1991 there were 70 technology IPOs which was just a small part of the total of 286 IPOs in that year. However, in 1995 there was a dramatic increase of public offerings, resulting into 204 IPOs in the technology sector. Furthermore, a total of 369 technology companies went public in 1999, which was 77.52% of all the IPOs in that year (Ritter, 2014). Research held by Ljungqvist and Wilhelm (2003), report that the number of Internet IPOs increased from 19 in 1996 to 22 in 1997 to 39 in 1998 and then to 257 in 1999 and 135 in 2000. The burst of the technology bubble is generally considered to be the trigger for the much broader technology market recession that followed after the bubble (Bhattacharya et al., 2010). This a potential explanation for the major drop of technology IPOs in the period following the dot.com burst, starting form 2006 (Appendix I). Another explanation for the fall in public offerings is that more and more companies decide to stay private for a longer period (Appendix III). Furthermore, technology companies do not need to IPO due to the large availability of capital. Private investment rounds have filled the cash requirements of most technology companies (Titcomb, 2015).
II.II Independent Variables
This subsection will discuss the main factors affecting IPO underpricing. It will introduce the theory related to these variables and its development in the bubble of 2000. After each variable, a hypothesis is conjectured explaining the relationship with IPO underpricing. At the end of this subsection an overview will be given, containing a summary of the expected relationship of each variable.
Underwriting Rank
In general, IPOs are arranged by a lead underwriter (investment bank), which gauges institutional investor demand by extensive marketing research and also arranges a syndicate of other banks to share the underwriting risk and bundle experience5. A summary of the most important responsibilities of the lead underwriter is formulated as: advise the issuer on pricing the issue, both at the time of issuing a preliminary prospectus that includes a file price range, and at the pricing meeting where the final offer price is set (Loughran and Ritter, 2002). The market and stock pricing research is a very expensive process for the underwriter. For example, during the bubble, some of the largest underwriters are spending close to $1 billion a year on equity research (Rynecki, 2002). Part of the way that these costs
5
9 are covered is by charging issuers of securities explicit (gross spread, which is commonly 7% a share) and implicit (underpricing) fees (Perkins and Perkins, 2001; Loughran and Ritter, 2002). Furthermore, history tells that (prestigious) underwriters are very keen to the underpricing of IPOs (Liu and Ritter, 2011). If underwriters receive a compensation from both the issuer (gross spread) and the investors (quid pro quo for leaving money on the table6), then the underwriter has an incentive to advocate a lower offer price than when the compensation was just the gross spread (Loughran and Ritter, 2002). This leads to a situation with agency conflicts between the issuing company and the investment bank, which is known as the agency hypothesis (Loughran and Ritter, 2002). In this case, the underwriter benefits from money left on the table through the rent-seeking activity of buy-side investors, whereby investors compete for allocations by offering the underwriter side-payments (Loughran and Ritter, 2002; Liu and Ritter, 2011). Basically, what this means is that the underwriter has side-line agreements with the investor (this can be both on past as well as on future businesses) that some part of the underpricing profit will flow back to the underwriter. The next example show how beneficial such a quid pro quo agreement can be for an investment bank. Consider that Merrill Lynch has alleged to receive commission business equal to one-half the profits of investors received from the IPO of the THEGLOBE.com company. The IPO involved 3.1 million shares at $9.00 a share, with a gross spread of roughly speaking 7 percent, which is equal to $0.63 per share. If the investor bought shares at the offer price and sold its share after the first-day of trading (606%), capital gains would have amounted $54.50. Based on the commission business equal to one-half of the profits, the total underwriter compensation for Merrill Lynch would have amounted to $0.63 plus $54.50, or $55.13. This example reflects the notion that the underwriter is eager in setting a lower offer price, especially during ‘hot-IPO’ markets characterized with high underpricing. However, these quid pro quo arrangements are strictly forbidden by the SEC7, but appeared to happen during the late ‘90s. Loughran and Ritter (2002) argue that these arrangements and IPO allocations strongly attributed to the increase in underpricing, as underwriters became more profitable due to the increased willingness of issuers to explicitly leave money on the table. A possible explanation behind this logic is that issuers received good news about the prospects of their personal wealth. As such, managers of issuing firms reduced their bargaining power as they were complacent in setting a lower offer price due to the increase in valuations. At hence, these companies received public attention and insiders simultaneously increased the possibility in increasing their personal wealth. Another reason why underwriters have an interest in some level of underpricing is because the market associates underpricing with a successful IPO. Underwriters off course like the idea that their name will be related to the successes of public offerings.
6 Money on the table is the number of shares issued multiplied by the difference between the first closing market
price and the offer price. It is the dollar value of underpricing. 7
10 Prestigious underwriters are generally reluctant in taking young, unproven and small firms to the public, because of the higher risk associated with these firms. Loughran and Ritter (2002) document that during the bubble, prestigious underwriters relaxed their underwriting standards and an increasing number of very young, unprofitable companies were taken public. They document a significant relationship between underwriter and underpricing during the dot.com bubble. Since the beginning of 1998 throughout the end of 1999, Internet IPOs underwritten by (the prestigious) Goldman Sachs have surged an average first-day return of 62 percent. Morgan Stanley, another highly prestige investment bank, helped IPO’s during that same period to a first-day return of 46 percent. Most entrepreneurs and start-ups tend to put either one of these banks on their dream underwriter list. Merrill Lynch, a bank that increased its presence in the technology sector scored an average of only 17 percent after the first day of trading (Perkins and Perkins, 2001). The IPOs managed by high-prestige underwriters during the 1990s and the internet bubble are associated with more underpricing than IPOs managed by lower prestige underwriters (Loughran and Ritter, 2002). Furthermore, highly ranked underwriters tend to be associated with deals by older and larger firms (firms with larger offerings), as well as firms that are backed by venture capitalists (Kim et al., 2008). Based on above literature (explicitly the agency hypothesis) and previous results, this research expects that high prestige underwriters are associated with more underpricing than IPOs managed by lower prestige underwriters. The distinction between low and high prestigious underwriters will be discussed in the Section Methodology. An overview of all the lead underwriters who managed the IPOs throughout the period of 2006–2015 is given in Appendix IV.
Leverage – Information Asymmetry
11 detailed information about the firm, meaning that the company has to provide this information. Moreover, the funder could also demand to be involved or at least informed about future developments, like budgeting and planning. Therefore, IPOs that have a higher level of asymmetric information (thus a low leverage) and valuation uncertainty should be more underpriced. Previous studies, find that issuing debt (high leverage) before an IPO signals to the market that the firm is of high value (James and Wier, 1990; Habib and Ljungqvist, 2001; Schenone, 2004). These authors hypothesize that issuing debt reduces asymmetric information, resulting into a lower IPO underpricing. Additionally, managers of highly leveraged firms face stricter budget constraints and have less control over the firm’s cash flows. As a consequence, less valuable firms are unwilling to adopt much debt because they are more likely to be forced into bankruptcy.
So far, previous research has been inconclusive. On the one hand, prior research generally suggests that regardless of the firm’s underlying characteristics, higher financial leverage results into lower underpricing. Ritter and Welch (2002) found that all theories of underpricing based on information asymmetry share the prediction that underpricing is positively related with asymmetric information. On the other hand, research held by Su (2004), finds a positive relationship with IPO underpricing and firm leverage at the book year before the IPO, at a significance level at ten percent. Kim et al. (2008) found that leverage was positively related with IPO underpricing for high-technology firms only and slightly negative related with IPO underpricing for low-technology firms, both at a significance level of five percent. A possible explanation can be found in the ‘winner’s curse’ hypothesis that uninformed investors face (Rock, 1986; Beatty and Ritter (1986). According to this theory, informed investors will only bid for securities that are underpriced and uninformed will bid on overpriced securities. In this case, the uninformed investor is aware of the possibility that he faces a chance to receive a higher portion of the overpriced security than the informed investor would. As a result, IPOs in general must be sufficiently underpriced to compensate the ex-ante risk or adverse selection bias of the IPO. This means that the higher the ex-ante information asymmetry of the IPO (high leverage), the higher the underpricing (Su, 2004). However, a high leverage also has its limits (higher than 1, resulting into negative stockholders equity) and is not necessarily a good signal. In such case, companies face a substantial higher risk and costs of financial distress (e.g. bankruptcy). Based on the literature of Rock (1986), and Beatty and Ritter (1986), this research expects to find a positive relationship between leverage and underpricing.
VC-backed – Investment patterns
12 in the internet and other information technology sectors. Throughout the bubble, VC funds have poured billions of dollars into hundreds of startups. The bubble period was an ‘easy money’ period for technology companies, where venture capitalists gave much more money to firms, many of which did not scored outstanding growth results and had a poor profit track record (sometimes even resulting into liquidation). This overoptimistic behavior of these so called “capital cowboys” is clearly given in the results of Ritter (2014). He found that during the periods 1997-1998, VC-backed technology IPOs averaged at 22 percent. Two years later, the range of VC-backed technology companies varied between 64 and 72 percent. The total proceeds for venture capitalist was a $23,254 billion in 1999, almost six times higher than four years before. In line with these results, most hot-IPO markets are backed by venture capitalists (Kim et al., 2008). It seems that the venture capitalists might be a key contributor to the emergence of the bubble, as Pitch Johnson argues: “Without doubt, the massive capital venture inflows are a major cause for the creation of the bubble” (Perkins and Perkins, 2001, p. 54). The problem is that the huge inflow of capital ultimately results in too much money chasing too few company deals with high quality, which result that the supply-demand side is unbalanced. Eventually, this leads into higher stock prices and pumps up the value of the company resulting into overvaluation of the company.
Venture capitalists generally focus on post-issue coverage (meaning that their return will be realized after the IPO); this is why they typically invest in young and risky companies with a high potential of growth opportunities. Most commonly, they do not sell shares in the IPO, but make distributions to their limited partners beginning when the ‘lock-up’8
period expires (Liu and Ritter, 2011). As a consequence of this, one would expect that combined with the theory that underpricing compensates for risk, venture capital backed IPOs is related with higher underpricing. This theory is known as the changing composition hypothesis, which describes that venture capitalist typically invest in riskier IPOs and therefore will be more underpriced than less-risky IPOs (Loughran and Ritter, 2002). On the other hand, much like prestigious underwriters of an IPO, venture capitalist would certify the quality of a company when taken public. The certification theory predicts less underpricing for VC-backed firms because venture capitalists can certify the fairness of IPO pricing due to reputation concerns (Chemmanur and Loutskina, 2006; Megginson and Weiss, 1991). This assumption reflects the notion that venture capitalists are concerned about their reputation in the IPO market and have the ability to monitor and screen firms in the pre-IPO stage. As a result of this, they want to price the equity of IPOs backed by venture capitalists closer to the intrinsic value. As a result of these two conflicting theories, empirical evidence is mixed. Previous research finds that, prior to the burst of the technology bubble, the first day return for IPOs backed by VC firms was double that of non-VC backed IPOs (Bradley et al., 2014; Loughran and Ritter, 2002). It seems to be the case that venture capitalists affect the valuation of the IPOs backed by them by attracting high quality market players to the IPO, which, in turn, increases the heterogeneity in investor beliefs about the firm’s future prospects. In line with the
13 certification theory mentioned earlier, Megginson and Weiss (1991) postulate the opposite and find that VC-backed IPOs are less underpriced than non-VC-backed IPOs. However, these findings are based on sample period of 1983 until 1987. This research draws its expectation based on the findings of Loughran and Ritter (2002), since their main focus is on technology IPOs during the bubble. Therefore, this research expects to find a positive relationship between VC-backed IPOs and underpricing.
IPO Age – Date of Incorporation
If the arrival of the Internet in 1995 can be regarded as an increase in the rate of technological progress, Michelacci and Suarez (2004) forecast that younger technology companies will go public and the number of startups will increase. This is exactly what happened between 1996 and 2000 as the age at issue date declined over the period (Kim et al., 2008; Loughran and Ritter, 2002). The average issuer for technology firms was 14 to 17 years old in 1996 through 1998 versus 9 to 10 years in 1999 and 2000. The median age fell by about a third, from 8.5 to 9 years in 1996 through 1998 to 4 to 6 years in 1999 and 2000, as the number of young tech companies that went public rose considerably (Ljungqvist and Wilhelm, 2003; Kim et al., 2008; Loughran and Ritter, 2002). Among the technology companies, the very young firms outperformed the older firms in underpricing, particularly during the 1999 to 2000 IPO period (Loughran and Ritter, 2002). However, during the bubble, research held by Clark (2002) found a statistically significant positive relationship between firm age at the IPO and long-run aftermarket performance, meaning that the chance of mortality is higher for these younger firms. Ljungqvist and Wilhelm (2003) found that age prior to the IPO was negatively related with first-day returns at a significance level of five percent. Furthermore, Loughran and Ritter (2002) documented a higher level of first-day returns for younger firms than older firms. They propose a potential explanation for the cross-sectional pattern between a firms’ age prior to the IPO and underpricing, which is that younger firms are more risky firms, and investors need to be compensated for that risk. Combined with the hypothesis that underpricing compensates for risk, previous research held by Clark (2002) found a statistically significant positive relationship between firm age at the IPO and long-run aftermarket performance. This explains that younger firms are indeed riskier and have a higher mortality rate. Based on these findings, this research expects to find a negative relationship between a firm´s age at the time of going public and underpricing.
Selling Intensity
14 During the dot.com period 1996 through 1998, more than one third of IPOs included shareholders shares offered. In 1999, 19.2 percent of all technology IPOs included shareholder sales and the fraction declined further, to 8.5 percent, in 2000. As a consequence, the average fraction of pre-IPO shares outstanding sold at the IPO declined, from 4.9 percent in 1996 to 0.7 percent in 2000, as did the share of shareholder sales in the average offer, from 9.8 percent in 1996 to 2 percent in 2000 (Ljungqvist and Wilhelm, 2003). Furthermore, studies held by Ljungqvist and Wilhelm (2003) found that first-day returns are larger when insider ownership stakes are smaller and more fragmented when insiders sell fewer shares at the offer price. This is known as the ‘insider stake’ hypothesis, which posits that insiders have more incentive to reduce underpricing when they sell a large fraction of the firm. A low level of shareholders shares offered suggest that the owners are confident about the future, which will send a positive signal to the market and increases the demand for the issue, resulting into a higher underpricing (Otchere et al., 2013). As a result, research finds that managers and executives strategically underprice IPOs in order to maximize personal wealth from selling shares when the lock-up period expires (Aggarwal et al., 2002; Ang and Brau, 2003). This empirical view is slock-upported by the increase of retained shares of insiders during the bubble. Regarding to the aforementioned, several studies found evidence that insider selling is negatively related with the underpricing of IPOs (Grinblatt and Hwang, 1989; Brennan and Franks, 1997; Ljungqvist and Wilhelm, 2003). However, others postulate a positive relationship between insider selling and underpricing. For instance, Ang and Brau (2003) argue that insiders (managers, venture capitalists, etc.) cofound the negative signal, but at the same time send a positive signal with the commitment of selling no shares during the lock-up period. This research does not distinguish the identity between the owners (managers, venture capitalists, others) who sell their shares. Based on the existing literature and the ‘insider stake’ hypothesis explicitly, this research expects to find a negative relationship between selling intensity and underpricing.
II.III Control Variables
In order to control for firm characteristics that might influence underpricing, several other variables are included and are formulated as: Financial Crisis, Size and Profitability. Along with these control variables, there is a substantial difference associated with them during the bubble compared to the years before and shortly after the bubble. This is consistent for both within and outside their industry segment.
Financial Crisis
15 markets are characterized with lower underpricing and a substantial reduction in the amount of offerings (Helwege and Liang, 2004). The financial crisis leaded to a fall in the general stock market activity, as it suffered from rapid declining security prices and a widespread of investor pessimism. This resulted that the years 2008 and 2009 corresponded with extreme low IPO activity. In the results of this research, there is a clear drop to only seven IPOs in the years 2008 and 2009 compared to forty-three IPOs in the year 2007 (Appendix I). As a consequence, this period is valued as a ‘cold’ IPO market and therefore will be added as a control variable in the research. This research proposition is to find a negative relationship between underpricing and the dummy variable of the financial crisis.
Size
Size is known as an attribute that affects the IPO performance (Mikkelson et al., 1997; Baker and Gomper, 2003). Small and relatively younger companies have a larger substantial of risk associated to the IPO, and like explained earlier; their shares would be expected to be more underpriced because of the risks related with the IPO. Typically, a larger firm could exercise more power and should be better in influencing the issue price in negotiations with the lead underwriters, which would possible lead to a reduction of ‘money left on the table’ and underpricing (Otchere et al., 2013). Based on this assumption, expectations are that size will have a negative effect on underpricing.
Profitability – Net Income
16 profitability during the bubble, this research postulates to find a negative relationship between IPO underpricing and profitability.
As mentioned shortly at the start of this section, an overview of the postulated relationships of each independent variable with the dependent variable can be found in Table 1. The outcome of the regression is documented in Section IV Results.
Table 1: Summary of Postulated Relationships on Underpricing
Includes expected relationships from 2006 throughout 2015 with the dependent variable underpricing. The Bubble Period column indicates the relationship between variables based on the results and literature in the bubble. The Hypothesis column indicates the postulated relationship between variables. Note: + / – means that literature and/or previous results report mixed evidence and is inconclusive. N.D. stands for: not defined.
III. METHODOLOGY & DATA
This section starts with introducing the measurement of the dependent and independent variables. The next part will contain the method for data gathering and how the data is obtained. The third part will formulate and discuss the descriptive statistics. Finally, the main quantitative methodology tested in this research will be presented.
III.I Measurement of Variables
IPO Underpricing
The IPO first-day initial returns (underpricing) can be defined as the increase from the offer price to the closing price on the first day of trading (Beatty and Ritter, 1986). In other words, it is the
9Previous research documented the bubble years as a ‘hot’ IPO market, suggesting a positive relationship
between underpricing and the dummy variable for this period. The financial crisis is found to be a ‘cold’ IPO market and as a result, a reversed relationship is expected.
Variables Bubble Period Hypothesis
17 difference between the offer price and the market closing at its first-day trading. Meaning that when the closing price of stock ‘i’ is higher than the offer price of stock ‘i’, the stock experiences underpricing. The return of stock ‘i’ at the end of the first trading day ‘d’ is calculated as (Bhattacharya et al., 2010; Ritter and Welch, 2002):
(1)
where, on the first day of trading ‘d’, the closing price of stock ‘i’ is Pid ; the offer price of the same
stock ‘i’ is OP ; and first-day initial return of the same stock is Rid.
Underwriting Rank
When a tech firm goes public, the underwriting section of the prospectus lists all of the investment banking firms that are part of the underwriting syndicate, along with the percentage or total of shares that each participant underwrites. More prestigious underwriters are listed higher in the underwriting section, with the underwriters in higher brackets (8.001 or higher) underwriting more shares (Loughran and Ritter, 2002). For underwriter prestige rankings, this research focuses on the CM-ranking model defined by Carter and Manaster (1990). In general, underwriters with a CM-rank of 8.001 to 9.001 (on a scale of 1.001 to 9.001) are considered to be prestigious national underwriters. Those with a rank in a range of 5.001 to 7.9 are considered to be quality regional or niche underwriters. Underwriters with a rank of 1.001 to 4.9 are generally linked with penny stocks; while many of those with rankings below 3.0 have been charged with market manipulation by the SEC (Loughran and Ritter, 2002). The CM-ranking model is the most commonly used ranking measure for underwriter prestige. More importantly, this measure is recently updated by Ritter (2014), but does not include rankings for all the underwriters over the period of 2012 till 2015. Therefore, the value of the previous period will be used if values are missing for some underwriters. In most of the circumstances, two or more underwriters are tied to the IPO of the company. One of the reasons for adding more lead managers is to secure more analyst coverage and increase the playing power (Liu and Ritter, 2011). In such a case, the average of the underwriters’ rankings will be used to estimate the total underwriting rank. To distinguish between prestigious underwriters and the rest, a dummy variable is constructed taking a value of one for IPOs tied with an average ranking of 8.001 or higher and zero otherwise. This is consistent with literature, as a previous research contained a proxy for prestigious underwriters (Liu and Ritter, 2011).10
10Robustness-test variable: when average rankings are used in the regression, significance levels of variables
18 Given the literature and findings of previous research, this research will add the following variables to the regression model: Leverage, Age at IPO, Selling Intensity, VC-backed IPO, Financial Crisis, Size and Profitability. Because these variables are less distinctive and complex than underwriting rank and underpricing, they will not be discussed that extensively. A clear overview of all other variables used in this research, together with the description of the measurement is given in Table 2.
Table 2: Overview of the variables.
The description of the variables is consistent with previous literature.
Variable Description
Underpricing ([first day close – offer price] / offer price).
Underwriting Rank CM-ranking measure of underwriter prestige values from 1,001 to 9,001, with the value of 1 if the average ranking is higher than 8,001 and 0 when lower.
Leverage Ratio of total debt to total assets in the book year before the IPO.
Age at IPO Computed from the date of incorporation to the date of the IPO. Measured as Ln(1+Age at IPO).
Selling intensity Equal to the number of shareholders shares sold at the offer divided by the number of outstanding shares prior to the IPO.
VC-backed An indicator that equals 1 if an issuing firm is backed by venture capital, 0 otherwise.
Financial Crisis Dummy variable with the value of 1 for the period 2008 and 2009 and 0 otherwise.
Size Ln(Assets in the book year before the IPO in $ millions).11
Profitability Dummy variable with the value of 1 if a company makes a profit in the book year before the IPO and 0 if otherwise.
Firm Characteristics
Book value Assets Total assets in the book year before the IPO. Book value Debt Total liabilities in the book year before the IPO. Net Income after Tax Total net income in the book year before the IPO.
III.II Data
The data utilized in this research will contain IPOs of technology, high-technology and internet firms with four-digit SIC codes 3571, 3572, 3575, 3577, 3578 (computer hardware), 3661, 3663, 3669 (communications equipment), 3674 (electronics), 3812 (navigation equipment), 3823, 3825, 3826, 3827, 3829 (measuring and controlling devices), 4899 (communication services), and 7370, 7371,
11
19 7372, 7373, 7374, 7375, 7378, and 7379 (software), as used in previous studies by Loughran and Ritter (2002). A more recent article suggest that the definition of technology stocks has changed over time, which results that the SIC codes 3559 and 7389 (business services) are added to the technology industry (Ritter, 2014). Technology IPOs with a stock offer price below $5.00 dollar or traded on the OTCBB12 market, will be eliminated from this research as the same assumption has been made in previous literature (Loughran and Ritter, 2002; Ritter, 2014). Technology companies that went public will be obtained from Bureau van Dijk (Orbis), which includes all U.S. listed companies in the period of 2006 – 201513. This research started with 349 companies, but eliminated 123 firms due to an incorrect date of the IPO, missing or inconsistent data of variables, trading on the OTCBB market or an offer price below $5.00. This leaded to a total sample of 226 companies. First-day trading prices, the closing price after day one of the IPO and the trading volume on the day of issue are obtained from Thomson Reuters (DataStream). This database includes stocks listed on the New York Stock Exchange (NYSE) and NASDAQ. However, Thomson Reuters does not contain all the information of the selected companies, such as the offer prices and ownership structure. As a consequence, missing offer prices and selling intensity is withdrawn from EDGAR online and the NASDAQ website. Data related to IPO deal characteristics such as; venture capital structure and leverage are retrieved from the US SEC government source: filing data of IPO prospectus file type S-1, S-1/A or SB214. The information on the date of incorporation is found on Bureau van Dijk (Orbis). Finally, some of the missing data, such as the variables underwriter rank, VC-backed and closing price of the offering was hand collected.
12 Over-the-counter (OTC) equity securities that are not listed on the NASDAQ, NYSE or any another national
stock exchange.
13 The regression sample runs until September 2015, since that was the outset of this research. This explains the
fall in of IPOs associated within the year 2015 (Appendix 1). 14
20 III.III Descriptive Statistics
Table 3: Descriptive Statistics
The descriptive statistics includes all 226 observations from 2006 throughout 2015. All variables show a significance level of one percent. Note: Age at IPO, Underwriting Rank and Net income after Tax are reported in their original values, while they are later transformed for the estimation. The original values reveal more information associated with the IPO than the dummy variables.
Minimum Median Mean Maximum Std. Dev.
IPO Underpricing (0.47) 0.18 0.25 2.22 0.35
Underwriting Rank 1.001 8.34 8.67 9.001 1.13
Leverage (0.41) 0.76 0.86 10.03 0.84
VC-backed 0.00 1.00 0.76 1.00 0.42
Age at IPO (years) 0.09 7.9 9.25 73.93 7.76
Selling Intensity 0.00 0.01 0.05 0.75 5.80
Net Income after Tax (M$) (179.3) (1.41) (2.44) 696.8 5.79
Book value Assets (M$) 0.18 82.39 404.69 7,999 1,065
Book value Debt (M$) (63.04) 61.52 337.56 9,173 1,042
Revenue (M$) 0.00 52.45 169.51 3,297 430.36
Table 3 suggests that from 2006 throughout 2015, IPO underpricing was quite excessive, as it averaged at 24.6%. During the years 2008 until 2010, the number of IPOs and first-day returns dropped significantly. The best explanation is probably the occurrence of the financial crisis, which increased the risk associated with the IPO. In 2008, underpricing averaged at only 7% and the total sample of technology companies that went public in 2008 and 2009 was 14 (Appendix I). For these reasons, this research will add a dummy variable for the financial crisis. From 2012 onwards, there is clear increasing trend in average IPO underpricing, as it scores a 23%, 35% 28% and 36% until 2015 (Appendix I). This increase in underpricing suggests that optimistic investor behavior and overvaluation is still evident. Furthermore, Table 3 also reports the profitability of technology companies in the book year prior to the IPO and scored an average loss of $2,500,000. The incapability of profit-making operating fell dramatically from 2013 onwards, as the average fell, suggesting higher losses. From 2013 to 2015, the average profitability was denoted as a loss of: $17,600,000; $19,100,000 and $21,500,000 (Appendix III).
III.IV Model used for regression
21 1)
However, other macroeconomic and company specific dimensions will be used as control variables, since they could have a significant relationship and increase the considerable power that explains the realized IPO underpricing. In this section, several implications concerning the relationship with IPO underpricing will be tested. In particular, the financial crisis caused the IPO market to decrease resulting into less underpricing, during the years of 2008 and 2009. As a result, this period was defined as a ‘cold’ IPO market. For that reason, the dummy variable financial crisis will be tested in OLS (2) of Table 4 and, as of in Formula 2.
2)
In addition, size is found to be influential and affects the level of underpricing and will be tested in OLS (3) of Table 4. This research adds this variable as a control variable, since literature did not found significant influences during the bubble, but does argues that size is an important indicator of underpricing. Size does not correlate with age and all other explanatory variables; therefore testing the effect of this variable is valued as sufficient. This assumption will be tested as in Formula 3.
3)
Finally, there could be an explanatory relationship between first-day return and profitability that is not well documented in the literature. Profitability is found to have an effect on the long-run performance of companies, but does not explain the relation on the short-run performance, as in underpricing. Profitability will be denoted as a dummy variable, taking the value of 1 if a company was profitability in the book year prior to the IPO and 0 otherwise. This assumption will be tested in OLS (4) and is given in Formula 4, together with the dummy variable for the financial crisis and the size of the company.
4)
22 The outcomes of OLS (1) – OLS (4) are discussed in the next section. Current literature suggests that besides profitability, all other variables have an explanatory power in determining the underpricing of an IPO. However, there can be an undocumented explanatory relationship between underpricing and profitability. Furthermore, previous research held by Kim et al. (2008) and, Wilhelm and Ljungqvist (2003) used a two-stage least square model (2SLS) that focuses on a potential selection and endogeneity bias. Due to the complexity of the model and scope limitations, the two-stage Heckman model will not be tested in this research. The regression specification established in Formula 1 to 4 ignores the potential selection bias, which could be present in this analysis. Nonetheless, the selection bias will be further discussed in the section Results.
IV. RESULTS
In this section, this paper discusses the results of the models described in the previous section. This research first elaborates on the linear restriction tests related to an ordinary least square (OLS) method. The second part will contain the discussion of the regression output. The last part will contain the discussion of a potential selection problem of the model.
IV.I Multivariate Regression
In the empirical analysis, the main focus is on the variables that influence IPO underpricing, as determined in the literature. Table 4 presents the OLS regressions in which the level of IPO underpricing (first-day return expressed as a percentage) is the dependent variable. The main model of underpricing is tested in OLS column (1) of Table 4. The coefficient estimates are stable and slightly differ across the models, reported in column (1) to (4). However, the explanatory power in OLS (1) to OLS (4) is considerable lower than previous research. Liu and Ritter (2011), Ljungqvist and Wilhelm (2003), and Kim et al. (2008) found an adjusted-R2 of 28%, 45% 48% in their results. A possible explanation could be the larger sample size and the use of more additional variables that contributes in the determination of underpricing. The value of Prob(F-statistic) is the probability that the null hypothesis for the full model is true, meaning that all independent coefficients tested are equal to zero. The Prob(F-statistic) in OLS (1) and OLS(4) is 0.036 and 0.006. This means that this research rejects the null hypothesis at a high confidence level, which leads to the conclusion that at least one of the regression coefficients in the model is not equal to zero. Both the adjusted-R2 and the Prob(F-statistic) increase when control variables are added into the model, meaning that the fitness of the model improves.
23
Testing for Heteroscedasticity
Since multiple independent variables are used in the regression, standard errors could be inappropriate and hence any inferences could be misleading in the model. Whether the standard errors used in the usual formula are too large or too small will depend upon the form of the heteroscedasticity. The White-test will be used in the regression to test for possible signs of heteroscedasticity with the errors. The null hypothesis in this test can be formulated as H0: no heteroscedasticity. The result finds an
F-value of 0.2, meaning that this research does not reject the null hypothesis and finds that heteroscedasticity does not seem to be an issue.
Testing for Autocorrelation
The Durbin-Watson (DW) is typically known as a test for first order autocorrelation. More explicitly, the DW tests the H0: ⍴ = 0 and H1: ⍴ ≠ 0. Based on the results of the analysis, this research does not reject the null hypothesis, since the DW is close to 2 in OLS (4) (DW=2.016), which means that there are no patterns in the residuals and that there is no sign of autocorrelation.
Normality
In a regression analysis, many statistical tests have been proposed to find out whether the error term is normally distributed or not. These statistical tests are typically constructed using OLS residuals. In addition, to test for normality in the distribution of the residuals, this research conducts a Histogram – Normality test (Jarque-Bera). The null hypothesis of the Jarque-Bera test is a joint hypothesis of the skewness being zero and the excess kurtosis being zero. The Jarque-Bera of the residuals in OLS (4) is 782.59 and 636.19 (both with a p-value = 0.0000). The accepted values of skewness and kurtosis are 0 and 3, but this research reports values of 1.9 and 10.3, which is outside the range values for normal distribution. As a result, this research rejects the null hypothesis that the residuals are normally distributed, and indicates that the assumption of normal distribution is violated. The residuals show a sign of right skewness and kurtosis. Given the fact that this research focus on IPOs, one could assume that there is no full adherence to the norm of normality. This could eventually lead that the results of the regression model are not totally efficient, and worse case, could be shortly biased.
Multicollinearity
24 Appendix VI. This research does not detect any sign of (perfect or near) multicollinearity, as the highest correlation between the independent variables is 0.31.
IV.II Discussion
This subsection will discuss the results presented in Table 4. The first part will contain the discussion of the relationship with the independent variables of OLS (1). The second part will focus on the explanatory power of the control variables in OLS (2) to OLS (4).
Table 4: IPO Underpricing Regression Table.
The sample in column (1) includes all explanatory variables listed in the literature review. Column (2) includes a dummy variable for the financial crisis. Column (3) includes a variable for measuring the effect of the size in the regression, defined as Ln(Assets). Column (4) includes a dummy variable for profitability. The dependent variable in all OLS regressions is the percentage first-day return. The std. error is denoted in the interference, which is underneath the coefficient of each variable. The significance level (p-value) is expressed as: * at a level of 10%, ** at a level of 5%, *** at a level of 1%. The total sample consisted of 226 observations.
OLS (1) OLS (2) OLS(3) OLS (4)
Underwriting Rank 9.67* (5.30) 10.03** (5.27) 13.94*** (5.64) 13.56*** (5.63) Leverage 6.65** (2.94) 7.26*** (2.94) 5.27* (3.11) 4.32 (3.17) VC-backed 4.66 (5.68) 4.75 (5.65) 1.52 (5.87) -0.27 (5.99)
Age at the IPO -6.47**
25
Independent Variables
Among the explanatory variables, Underwriting Rank has a positive and significant relationship with underpricing (p-value = 0.069), which is consistent with the findings of Kim et al. (2008) and Loughran and Ritter (2002). A possible explanation is that prestigious underwriters are very selective in the companies they take public, but as soon the deal is made, the underwriter is eager to act in the best interest of the bank, which means that an IPO is more likely to be underpriced. In such a case, the underwriter is typically better off in engaging a lower offer price, because by then the IPO is considered to be a success. This pattern of higher IPOs underpricing associated with prestigious underwriters is consistent with the agency theory defined by Loughran and Ritter (2002). They argue that underwriters have an incentive to strategically underprice the IPO if the investment bank receives compensation from both the issuer and investor. In an industry with an average underpricing of 24.6%, one could assume that these so-called arrangements between underwriter and investor exist, because the generated commission for the underwriter is typically larger than merely the gross spread. It seems that the agency conflicts documented in the bubble are also evident within the time period of this research. Therefore, the suggestion can be made that similarities in the underwriter segment occur between both IPO periods.
Table 4 indicates that underpricing is positively related with Leverage, meaning that a 1% increase in leverage results into a 6.65% increase in underpricing (p-value = 0.024). This finding is consistent with the results of Kim et al. (2008) in which they found a negative relationship between low-technology and underpricing, but a positive significant relationship between high-low-technology and underpricing. The results suggest that firms with a lower leverage set their IPO offer price closer to their intrinsic value, which would lead to lower underpricing. Additionally, firms with a high leverage set their offer price below their intrinsic value to compensate investors for their exposed ex-ante uncertainty and informational risk. As it seems that IPOs that are highly funded by debt (as well for venture capitalists), the funder may have better information about the true value or risk of future cash flows than investors. In this case, underpricing may be used as a signal the company’s ‘true’ value. This is consistent with Rock’s (1986) ‘winner’s curse’ theory that assumes that some investors are better informed about the true value of the shares offered than other investors in general.
26 previous bubble as VC-backed IPOs rose especially during the ‘hot-IPO’ period, which at hand, could lead to a contributing factor of creating a bubble. Venture capitalists are indeed dreaming big, as the total venture investments in 2014 was $48 billion, more than any year since 2000. Although, this is not as high compared to the bubble period, when close to $100 billion in venture capitalist investments occurred. According to Bill Gurley, the venture-capital community is taking on an excessive amount of risk right now, which is unprecedented since the late ‘90s. As this moment, only a few capitalist have had high success rates with unicorns that went public (Fitzgerald, 2015). As a result, the mood among some venture capitalist backers of startups has become more cautious. It seems that fewer specialist technology investors are taking part in new financing rounds. But startups are found to have no problems in raising their capital, because general investors such as hedge funds, assets-management firms, oligarchs and other funds are filling the gap (The Economist, 2015d). The disposition of withholding investments is however a long-term process, as most of these startups are clearly not in the running for an IPO in the near future.
27 This research finds that Selling Intensity is positively related with underpricing but it is far from significant (p-value = 0.826). This is contradicting with the expectations and the ‘insider stakeholder hypothesis’. Moreover, it is inconsistent with the literature (Grinblatt and Hwang, 1989; Ljungqvist and Wilhelm, 2003), suggesting that selling intensity should decrease the level of underpricing. A possible explanation of this could be that the range of selling behavior within shareholders is quite low as it averaged at only five percent. However, this time period shows similarities with bubble, as the selling intensity decreased harshly from 1997 towards 2000. It seems that investors are still known to be very optimistic about long-term prospects of technology firms. This could possibly lead to a situation that investors ignore the selling behaviour of shareholders at the IPO. It appears that the lock-up period is a regular practice within this type of industry and like explained, its sends a positive signal to the public, which could increase the expectations and demand in shares of investors.
Control Variables
In the second model OLS (2) of Table 4, the ‘cold’ IPO market during the years 2008 until 2009 is added to the regression model as a control variable. This research finds that this cold period of IPOs, known as the Financial Crisis, has the largest negative relationship (19.32%) with IPO underpricing (p-value = 0.053), which is consistent with both the literature and the expectations (King and Banderet, 2011). During a recession and in periods of financial distress, issuers and underwriters are very reluctant in taking a company to the public due to pessimistic investment behavior and the jeopardy of failure. Furthermore, the average age of companies in this time period is relatively large (thirteen years compared to nine on average). Given that, the sample size of public offerings in the financial crisis is relatively small, this results that it is difficult to perform a robustness test with different time periods.
In the third OLS (3) model of Table 4, this research finds a negative significant relationship for underpricing and the control variable Size (p-value = 0.061). These results indicate that bigger and well established firms (e.g. Facebook) have more bargaining power in setting the final offer price. It seems that they are more capable in setting the offer price close to its real intrinsic value, which reduces the money left on the table and increases the total capital funding. Another explanation for this finding is that there is a positive correlation between the numbers of shares issued (proceeds) and the size of the firm. So relatively speaking, it requires much more investor demand to achieve a certain level of underpricing at a bigger firm than at a smaller firm.
28 the companies were profitable. In the year 2014, 88% of the technology IPOs had negative earnings, which is at its highest since 2000. The negative relationship with underpricing shows that underpricing is likely to be higher if the company is not profitable. This confirms the literature of investor behavior on high ex-ante uncertainty related with the IPO, leading into higher underpricing. Furthermore, the clear fall in profitability within the technology sector seems to be a normal practice in today´s practice, as the gross of these companies are non-profitable organizations. This stems mainly from the investors, pumping money in the technology industry, which make that these organizations have the habit of spending all the money in an attempt to expand their market share. Such practices are only likely to stop in a new situation, in which investors exercise more control, or that the funding of these firms dries up, or perhaps worst, a (new) bubble will burst.
IV.III Selection Bias
One potential problem that may be evident in this research is the existence of a self-selection bias that creates inconsistent parameter estimates in the regression. For example, consider the following situation in the firm’s choice (of preference) of underwriter prior to its IPO. Whether the firm has an established relationship with a bank that can underwrite IPOs or with a bank that cannot underwrite IPOs is, perhaps, not a random choice. The issue manager, who is acting in the interest of the technology firm, believes that future IPO underpricing will be lower if the firm has an established relationship with a bank that can (successfully) manage IPOs. In this case, the managers’ preference will be to engage a prestigious bank that has underwriting capabilities and hence eventually can lead the firm into a successful IPO. This will result that the larger and leading technology firms will pick, because of their bargaining power and their well-established market position, the prestigious underwriters. The aforementioned makes the choice of the leading underwriter no longer a “random” choice, which is a so called self-selection bias. Moreover, the sample of this research only contains technology firms that actually completed an IPO. Firms that aborted the IPO for whatever reason (e.g. recession, take-over or poor advice underwriter) are excluded from the data sample. So it could be stated that not the entire population was observed.
