Stock Performance of Social Media
IPOs
Master Thesis
Joachim Wisman Groningen, August 2015 University of GroningenFaculty of Economics and Business MSC International Financial Management
Uppsala University
Faculty of Social Sciences MSC Business and Economics
Student number (RUG): 1673343 Student number (UU): 880326-P539
2
Stock Performance of Social Media
IPOs
I know that you are rational, you know that I am rational, and I know that you know that I am rational (Thaler, 1999).
Abstract: This paper focuses on Initial Public Offers (IPOs) of internet firms after the Dotcom bubble. I examine the two-year performance. My sample has a international reach and identifies a clear underperformance of these internet firms. I find that existing literature concerning IPOs is not able to explain the underperformance. Also I find higher deal values of IPOs result in higher underperformance. Dorn (2009) classifies this as investors being attracted to glamour stocks. Amongst my sample there are a lot of firms that could be perceived as glamour stocks (Facebook, Groupon). Lastly I make use of a local, regional and global benchmark to adjust the returns.
JEL classification: G10 L25 L86
Key words: information and internet Services, long-run abnormal returns, stock issues,
3
Contents
I. Introduction ... 4
II. Relevance ... 5
III. Literature ... 6
IV. Hypothesis Development ... 11
V. Methodology ... 13
V.A Data collection... 16
V.B Regressions... 17
VI. Descriptive statistics ... 19
VI.A Regressions ... 20
VI.B Robustness ... 22
VII. Conclusion & implications ... 24
VIII. Limitations & extensions ... 26
IX. Bibliography ... 27
Appendix ... 31
Table 4 Correlation Matrix ... 31
4
I. Introduction
At the change of the last century into the new millennium, most of us were afraid of a digital crash. A crash of our operating systems, which had become quickly embraced by the public and our increasingly dependency of these desktops had grown in a similar direction. In the midst of this pending doom where our overreliance of these so-called ‘personal’ computers would backfire on us, another computer related danger was lurking the dark.
It was not our beloved personal computer that stopped functioning, but the appreciation of dotcom stocks. These stocks had received a lot of attention halfway during the nineties. Attention in the form of media exposure, (over)appreciation by financials and ultimately in the public their valuation of these stocks. This fad in our recent past has much been researched1. The question on everyone’s mind in the wake of such a crisis, is whether it can repeat itself and if so when it will take place and what shape it will take on.
However this research is not about ‘fad exploration’, it is focused on the so-called second generation of internet stocks. The dotcom burst did not put a hold on the development of the Internet. Since then it has continuously been expanding its reach, not just geographical, but it has also become more integrated in our daily lives. In essence it has made us able to get and stay in contact with everyone who is digitally connected. We have also seen the downside of this progress, whether it is the NSA being able to digitally hack into the daily lives or conventional business/ earning models being under pressure by firms like Groupon that make use of the bargaining power of the collective. Google has made everything searchable, but has also made businesses increasingly reliant on being found through Google. This in turn has developed a complete new branch of firms and professions. Most of these firms provide a service, which is intangible by nature. It is not just solely the technology behind it, which makes it hard to understand the performance, but also the impact of these services on existing industries. A recent example is Uber, which provides an alternative to consumers that need a ride. It pools ‘self-employed do it yourself’ cabdrivers and ultimately brings down the cab fares. In response to this, conventional cabdrivers resist to this change, which has even led to riots in Paris, where Uber cabdrivers saw their cars go up in flames.
The previously mentioned examples are just a pick out of a vast variety of services that are changing. The vast amount of research about the Dotcom crisis made it painfully clear, that it is hard for financials to value these services and businesses. Surprisingly not much academic research has been done about this novel, but increasingly more important industry. This paper intends to shed a first light about the stock price performance of this industry.
1 The underlying causes of the dotcom bubble have been researched in depth amongst those
5
II. Relevance
The performance of listed firms has well been document. Also the underperformance of firms after their Initial Public Offerings (IPOs) has been extensively researched and lastly the run-up and aftermath of the previously mentioned internet stock crash. One could say enough material to draw conclusions for this author on ‘second generation’ of internet stocks. The opposite is true, the mechanics of their business models is drasticly different. Internet portals or more commonly known as social communities are still in their infancy phase.
Higson & Briginshaw (2000) classify the second generation of internet stocks as new economy stocks, based on the idea that these stocks have new underlying business models. They argue that valuations have been off target, because too much emphasis was placed on growth prospects, rather than economic fundamentals. Hand (2000); Trueman et al. (2000); Hand (2001); Trueman et al. (2001) and Trueman et al. (2006) underline the explanatory value of economic fundamentals to explain and forecast stock behaviour and revenues of internet stocks. However they also explore the idea of web traffic variables. Their assumptions are that web traffic variables are driving revenues and ultimately internet stock prices. However their research proof the opposite of this assumption. They actually find that economic fundamentals and growth prospects are able to explain the stock price behavior.
Thus the question arises whether these internet firms are actually unique compared to the old economy stocks. From an academic standpoint the contribution of this research lies in understanding the price patterns these firms display. Secondly the price pattern itself might indicate whether it represents informational efficiency. Lastly, initial returns of IPOs have been found anomalous, therefore it is of interest if a reversal can be observed in the long-run.
6
III. Literature
The research concerning IPOs can be roughly divided in two fields. Firstly, the research area of the period before and around the IPO, I will refer to this as the pre-IPO period. The second area is the long-run performance of these IPOs. Through reviewing this literature more closely, the pre-IPO period can be divided in three focal points of academic attention. This division is mainly based on the three parties involved with IPO listings. Which are the issuers (firms pursuing a listing), the underwriters (investment banks that facilitate listings) and investors (purchasers). Mcdonald & Fisher (1972) research initial underpricing of new issue stocks in the period between ’69-’70 and find positive underpricing. Which implies positive returns on the first day when the stock is issued. Ibbotson (1975) continues on this research and finds similar underpricing results, but expands his scope for the period between ’60-’69. Since then more researchers have focused on this phenomenon and have tried to provide possible explanations. Rock (1986) finds that underwriters are sensitive to the question of the allocation process of shares. Implying that this process is not transparent. In the case of under subscription, the ability and efforts of the underwriter can be questioned. According to Rock (1986) this can be either due to the lack of promoting the shares or the ability to set the price. Thirdly when the shares are over subscribed, underwriters might be tempted to give a discount, to sell the shares easily. This last might be contradictional, but the subscription process or bookbuilding process does not happen overnight, but over a short period. Also subscribers have the chance to change their subscription before the IPO. Therefore an identifiable risk for underwriters arises, not being able to sell all shares. Because of this they might be triggered to price shares lower, to mitigate this risk.
However this does not explain why the issuing party does not get upset about this discount. Researchers frame this as ‘’money being left on the table’’, by the issuing party. Because any discount is directly paid out of their own pocket, namely an increase of share price at the first day, could also have been adjusted by increasing the initial offer price. Hence, this can be perceived as an instant loss when the firm goes public. And secondly the issuer pays the underwriter for their services. Loughran & Ritter (2004) provide a possible explanation why issuers do not get upset. They find for the period ’90-’98, that on average USD$9.1 million is left on the table, which they perceive as little. Secondly when more money is left on the table, it most often occurs when the initial offer price was already set higher. Thus the issuer is already receiving more, than anticipated. Or as Loughran & Ritter (2004) frame it, the issuer is left happy, even though they have just been victimized.
7
need to be met. Firstly the uncertainty of the market price of the new issue for the underwriter. Secondly the reputation of the underwriter is put on the line for issuing the stock. Lastly the underwriter generates earnings based on this reputation. Thus in a situation where the underwriter cheats in the pricing of a new issue, it will backfire on his reputation and subsequently his earnings. Hence, when the underwriter underprices too much, it will lose new clients wanting to go public. And when the underwriter underprices too little it will lose investors buying their stock.
Benveniste & Spindt (1989) investigate how underwriters can determine more accurate price setting and the allocation of shares to investors. They find that when underwriters make use of the information investors are giving, they can actually reduce underpricing. They do this based on the premise of two contrasting ideas. One is the idea that the issuing firms is better informed about the actual value of its firm and the other is proposed by Rock (1986), that actually the investor is better informed. The latter is better informed because it is better aware of the market. Based on these conflicting ideas of information asymmetry between parties Beneveniste & Spindt (1986) develop a model that explains the role of underwriters in the IPO process and the integration of information for pricing and allocation of shares. Similar to Beatty & Ritter (1986), they underline the role of underwriters and argue that they make the pricing of IPOs more efficient.
Ibbotson (1975); Franklin & Faulhaber (1989) and Welch (1989) find that underpricing could actually be a signal to investors. According to Franklin & Faulhaber (1989) It should signal superior prospects about the issuing firm. This can be achieved through a low IPO and quantity. Welch (1989) argues that the reason to sell for more attractive prices to investors is to leave a good taste in their mouth for future seasoned offerings. In his research he finds evidence that about one third of the IPOs actually reissue, but the seasoned issues are triple in size compared to the initial offering. Franklin en Faulhaber (1989) argue that underpricing is a credible signal that the firm is good to investors, because only good firms can recoup initial losses endured by an IPO and bad firms will not be able to regain the lost ground by going IPO. Hence, bad firms going IPO cannot afford underpricing as a signal to investors. Thus they conclude that underpricing is actually a signal for the quality of the firm.
8
market will have lower post IPO productivity and profitability, but larger cash holdings. The first could be explained by the arguments of Franklin & Faulhaber (1989), that the quality of firms decrease during peaks of IPO cycles in relation to the post-performance.
Understanding what these cycles trigger is beyond the scope of this article, however it has partly to do with the investor. Ljungqvist et al. (2006); Campbell (2008) and Dorn (2009); explore this concept. Ljungqvist et al. (2006) classify the investors in s-type investor as the sentiment investor and the r-type, the rational investor. They model both behaviors and observe that the ability to generate abnormal initial returns have a lot to do with the ability to arbitrage. When the market consists out of both s-type- and r-type investors and going short is unlimited, prices of IPO will move to the actual value. However there are a lot of technicalities, contractual agreements and legalities that prevent investors to drive the price to its actual value during IPOs. According to Engelen & van Essen (2010) the legal and institutional framework to which issuers, underwriters and investors need to abide by are country specific and impact underpricing in their own unique way. Dorn (2009) has a unique sample where he applies these country specific variables and how they influence underpricing. Although his sample is limited to 1 underwriter and his IPOs span between August ’99 and May ’00, he is able to isolate the behavior of s-type investors. He concludes that s-type investors are willing to pay a considerable premium for IPOs. Even though institutional investors have agreed upon that the price of the IPO is lower than the price s-type investors are willing to pay. Furthermore he researches the 6-month aftermarket returns and finds a negative performance in relation to the underpriced IPOs. Hence, he observes high initial returns to be followed by a stock performance reversal.
Cook et al. (2006) continue with the premise of s-type investors and extends it to attracting these investors through promotion. He finds in the context of IPOs, that there is a positive and significant relation between pre-issue promotion and the initial return of a stock. Secondly he argues that pre-issue promotion leads to upward revisions of the IPO price. Lastly he observes that there is a positive relation between underwriter compensation and pre-issue publicity. Thus it can be in the interest of the underwriter to invest in promotional activities.
The above reflects on the roles of the different parties involved in IPOs. All of the authors observe anomalies at the initial return of IPOs. At the background of this are two theories that can be used to explain financial anomalies. The first is the efficient market hypothesis (EMH) and the other is behavioral finance (BF).
Ritter (2003); De Bondt (2004) and Stracca (2004)2 argue that behavioral finance originated in the wake of anomalous events in the ’80s and ‘90s. They could not be explained by the EMH. Therefore the attention shifted to the individual behavior involved. Ofek & Richardson (2002) state that the efficient market hypothesis cannot explain the anomalies observed and therefore the
2 Stracca (2004) provides a summarized overview of the development of behavioral finance
9
behavioral finance context can possibly be useful. Hence, they question in their study the rationality of markets and find evidence of craziness amongst investors during the creation of the Dotcom bubble. Therefore they cannot reject behavioral motives in explaining this fad.
According to Ritter (2003) there is a distinction to be made in BF. One is based on cognitive psychology and the other is limitation of arbitrage. Stracca (2003); de Bondt (2004) and ap Gwilym (2010) frame cognitive behavior as decision heuristics and focus on the individual decision making of how these are triggered, influenced and can be observed.
Cook (2006) argued the influence of marketing on investors and the relation it has on the initial return. However Dorn (2009) presents that even though the initial return might be observed as anomalous, a reversal can be seen within 6 months. This raises the question whether IPOs are actually underpriced in relation to the long-run performance. Among many Chalk & Peavy (1987); Ritter (1991); Brav et al. (2000); Schultz (2003); Purnanandam & Swaminathan (2004) research this hypothesis.
The second research field up for discussion is the long-run performance, which has also received much academic attention. I have chosen to highlight some of the authors that relate the long-run performance back to newly issued stocks. The issue with measuring the long-run performance is two-folded. One part is about finding explanatory variables and the other about how to measure long-run performance and thus correctly identify possible patterns.
This long-run performance is amongst many documented by Ibbotson (1975); Chalk & Peavy (1987); Ritter (1991); Brav et al. (2000); Schultz (2003); Purnanandam & Swaminathan (2004); Ljungqvist et al. (2006) and Zheng (2007). They all find different explanations of initial underpricing and the long-run performance. Ritter (1991) analyzes a 3-year aftermarket performance and observes within that period underperformance. Furthermore he finds that age of the firms tends to amplify the underperformance when they went public in high volume years. Lastly he documents a variation in performance based on the industry the IPO is designated too. He argues that this could be, because of possible industry fads. Although Schultz (2003) does not base his market-timing of going IPO on industry fads, he does observe similar underperformance results as Ritter (1991). He provides evidence of IPOs that underperform in the long-run are clustered around peaks of heavy IPO issuances.
Purnandam & Swaminathan (2004) apply own developed P/V ratios3 to analyze the long-run performance. They conclude that firms which underperform on the long-run have lower profitability, higher accruals, and higher analyst forecasts. They interpret the changing ratios as the change of risk. However their research cannot provide any conclusive evidence that the underpricing and long-run underperformance can be captured by their ratios. They do observe
3 Purnandam & Swaminathan (2004) develop 3 ratios, which are based on multiples (sales,
10
that overpriced IPOs at the first day had received higher growth forecasts compared to undervalued IPOs. Brav et al. (2000) also use ratios to match IPO firms with nonissuing firms. They conclude that the IPO returns are similar to nonissuing firms when matched on size and book-to-market ratios. However they do find underperformance amongst small issuing firms with low book-to-market ratios.
Chalk & Peavy (1987) study the IPO performance through segmenting IPO prices. They observe that stocks priced $1.00 or less generate most 1st day abnormal returns. Also the long-run (positive) performance seems to be concentrated amongst stocks worth $1.00 or less.
The methods used to measure the performance in the previous articles are as diverse as there are articles. This can be due to personal preference or data availability. In general there is no consensus on how to detect long-run abnormal stock returns. The first step in any method is to choose ‘’with what’’ to match IPO firms with. This can be done with control firms, which are matched through similar ratios or industries. Another possibility is to develop matching portfolios and the final possibility is to match with market indices. The second issue is ‘’how’’ to compare the IPO firm and peer performer. This can be done through calculating Cumulative Abnormal Returns (CARs), Buy and Hold Abnormal Returns (BHARs) or an own developed variable, like the wealth relative that Ritter (1991) proposes.
Barber & Lyon (1997) analyze the properties of long-run abnormal returns. They argue that misspecification is reduced when firms are matched with control firms4. These control firms should be matched on size and book-to-market criteria. Furthermore they argue that CARs yield positively biased test statistics.
The previous reflect a brief summary of the main published and researched themes. Some of the referred articles date back more than 40 years and since then a lot has changed. Events in our recent past like the before mentioned dotcom bubble and the financial crisis have rekindled the interest of researchers. By example the role of the investment banker has been under great scrutiny because of the financial crisis. These same investment bankers also play a role in the listing of new issued stocks. Thus in contrast to the idea of the facilitating role and equilibrium enforcers proposed by Beatty & Ritter (1986) and Beneveniste & Spindt (1986) there might be actually more to it. This is later also underlined by Loughran & Ritter (2002), they observe that this underpricing mechanism enforced by underwriters can also be exploited and thus be beneficial for underwriters.
In short the valuation of IPOs can relate to any perspective you take. Whether it is the long-run performance or about initial return. This is of importance to the issuer for getting a fair price and finding the optimal point between issuing an attractive stock and not leaving to much money on the table. In regards to the underwriter, valuation is of importance not to underprice too much,
4 Barber & Lyon (1997) argue that the matching method that involves market indices are subject
11
because of possible reputation loss and not too little, because of losing interest of investors and being stuck with the risk of having unsold IPO stocks in your inventory. For the initial investor it is of importance to be able to buy an underpriced stock, but on the other hand the valuation is of importance because it is a signal for future performance of the IPO. Thus it tells the investor to take a long-position, sell at the first day or not to buy at all. Underpricing itself is one of the subtopics in the IPO literature. Researchers view this as a trade-off between the triad of parties involved. In the more dated literature the role of the underwriter is perceived as neutral and or facilitator of IPO listings and thus an unbiased operator of the underpricing mechanism.
However Loughran & Ritter (2004) argue that the role of underwriters has shifted. Analyst coverage for IPOs becoming increasingly important during the run-up towards the Dotcom bubble. They find evidence that issuers choose top-tier underwriters, because these underwriters were better able to influence analysts and positively the value of the IPO. Hence, increasing the underpricing. Secondly they observe that the degree of CEO ownership explained why issuers would choose for an underwriter with a reputation of larger underpricing. This was because former owners would generate an increase of wealth to their personal brokerage accounts. So according to Loughran & Ritter (2004), underpricing changed and should not be perceived anymore as a reflection of the quality firm.
Rock (1986) and Chalk & Peavy (1987) argue that if the IPO market is efficient, we would not expect daily aftermarket to be significantly different from zero: the initial return should resemble the average of expected returns at that moment. Therefore most researchers like Ljungvist & Ljunqvist & Wilhelm (2003) measure the degree of underpricing as the closing price on 1st day minus the offer price. However for the more casual investor, it will be very hard to acquire an IPO stock at the first trading moment. And actually benefit from underpricing. It is therefore far more interesting what the performance of the stock in the long-run is and if the underpricing sustains over a longer period.
IV. Hypothesis Development
I make use of the previously described literature in exploring the dynamics that can identify and possibly explain the patterns of IPO stock behavior.
I will look into the limitations of arbitrage (float, price) and sentiment (range). The proxy, deal value is used for similar reasons as Ritter (1991). He argues that smaller offerings have the worst underperformance. Region is a rough proxy, because it can imply a lot, but is used as an indicative proxy for further research. Lastly I do not wish to frame my hypotheses in double negatives. Hence, I state the dependent variable as ‘performance’. Implying that a negative
12
to outperformance. Where effectively the performance is measured as the long-run Cumulative Abnormal Return (CAR).
Hypotheses:
1) IPO proceeds have a negative impact on long-run performance
I include deal value as a proxy for a myriad of reasons. Dorn (2009) uses it as a general proxy for IPO deal characteristics. He observes two patterns. Firstly, investors like ‘glamour stocks’, thus the larger the IPO deal value, the more interested an investor would be to invest in. Secondly, his findings show that IPO proceeds negatively influence the underpricing on the long-run. Thus the larger the IPO proceeds, the larger the underpricing and the stronger the reversal effect. One could observe a near similar reversal in the case of the IPO of Facebook, where abnormal returns in the first days were gained, but underperformance in the long-run was observed.
2) Price has a negative impact on long-run performance
Chalk & Peavy (1987) find that the offer price for stocks of $1.00 or less shows significantly abnormal returns over the long-run of 190-day market. They argue that possible reasons for the anomomalous aftermarket returns, can be due to high transaction cost for low-priced stocks. From a behavioral standpoint this could indicate that the margin in selling and buying low-priced stocks is lowered, due to the transaction costs. Therefore some sort of limitation of arbitrage could be visible.
3A) Percentage of shares offered has a negative impact on long-run performance
Bartov et al. (2002) test for the influence of ownership. They argue that the former owner decides the percentage of shares he is willing to issue. Therefore the amount of shares offered is a signal to the market. They find that a lower percentage of shares offered indicate a positive signal to investors, because more of the shares are retained by the owner and thus less likely that the owner is to bail out. I measure this through the variable IPO_Percent.
3B) Amount of shares offered have a negative impact on the long-run performance
The float is another variable related to measure the impact of the amount of shares offered. The argument is that it reflects the trade-off between supply and demand. Which implies that a lower float increases the price of the stock. So a higher price should impact the long-run performance on the long-run. Although they do not make a distinction on who holds the remaining shares, it can contribute in explaining the performance for an initial indicating proxy.
13
4) Region impacts the variation of underperformance
Because of the international scope of this research I wish to identify regional differences. I apply dummies to see whether long-run performance differs amongst these regions. The regions used are Northern America, Europe, Asia and Other. Engelen & van Essen (2010) conduct a research in explaining underpricing of IPOs through the use of firm-, issue- and country-specific characteristics. They argue that country specific variables, like the legal framework in place to protect investors can explain the degree of underpricing. On the other hand country-specific characteristics may be relevant like having a Silicon Valley, which can possibly stimulate technology. This implies that some regions are more technological advanced and or able to exploit technology better compared to other regions. However it remains an indicative and prospective variable in explaining the variation of underperformance amongst regions.
5) Markup negatively impacts long-run performance.
According to Bartov et al.(2002) the markup might indicate that offer prices that are located to or near the upperbound of a price range are more sensitive to underpricing and on the long-run more sensitive to underperformance.
For this variable I needed a minimum of two prices described in the deal news dataset. It is approximation of the pre-IPO period. The range is constructed as follows:
Markup =
Where is either the opening price or offerprice and the price resembles an the lowest earlier price point during the pre-IPO period. This generates a ratio to which we can see how much the price differs from earlier mentioned variables. Hence a smaller ratio indicates that did not vary to much from the first . Thus following Bartov et al. (2002) a larger ratio implies that the values are more drifted apart during the pre-IPO period and the offer price. And a larger ratio should generate larger underpricing and thus increase the long-run underperformance.
V. Methodology
The initial sample is comprised of 167 initial public offerings. For the sample the following criteria were applied. 1) All IPOs for the period between 2001-2015 .2) Completed IPO deal values (proceeds) of $1.000.000, 3) SIC5 codes 6209 (other information technology and computer service activities) and 7311 (advertising agencies), 4) firms with a minimum of 2 year
14
listing since IPO6. A sample with an international span was constructed. This generated a list of ISIN7 codes I used.
In accordance to the research of Ritter (1991), I use 1) Cumulative Abnormal Returns are used. Later on I will also apply 2) Log Returns for the robustness tests to measure the long-run performance. Graph 1 provides an overview of the performance. Secondly I have adjusted these performances accordance to their own listed index, a regional index and a global index.
I measure the performance over a period for two years. This is for two reasons, firstly it provides an insight in the aftermarket performance, rather than solely focusing around the IPO date. Secondly the two-year period might allow for corrections in reaction to the initial response around the IPO date. In contrast to Ritter (1991) I have chosen for a two-year period instead of a 3-year period. The is because of the novelty of the industry, which implies that the firms do not have a long track record in terms of being publicly listed. Choosing for a longer aftermarket period would cause me to eliminate more firms. Hence, firms that have been covered in de media recently like Alibaba and Twitter are not in the sample, because they have not been listed for two-years or longer.
The returns have been calculated by using the closing prices of the individual stocks. Both the simple returns and log returns have been calculated for this research. Since the initial returns of the first day data could not be retrieved, a month 0 could not be calculated.. Therefore the measurement periods spans from month 1-24. Every month contains 22 trading days, which sums up to a total of 528 trading days, none of the firms delisted after two-years. The IPO portfolio constructed, reflects an equally weighted portfolio. Also the months all have an equal weight.
The monthly benchmark returns are calculated in a similar fashion. The raw returns of the different indices are matched per firm per day. The benchmarks used include 1) the listed index returns, 2) a regional and industry related index (regions: Northern America8, Europe9, Asia10 and Global11) and a global industry related index12. In case a firm was not classified as being located in America, Europe or Asia, it was matched to a global index. Thus matched twice, but firstly with the S&P global 1200 index and the entire sample of all firms were matched with the
6 The data showed that some firms had an IPO date within the period 2013-2015, but had
acquired a listing before. There are various reasons for this occurence. One is the already listed holding company would spinn-off his subsidiairy. In this case the IPO date of the holding company was used.
7 ISIN are Equity (in this case firm) identification tags 8 S&P 500 Information Technology Index
15
MSCI global information and technology index. The other used regional indices are all industry related benchmarks.
The following formulas have been used to calculate the variety of returns, for the benchmark-adjusted return for stock i for day t is:
Where is the abnormal return of firm i for day t. is the return of the firm and is the return of the market. As noted the day returns are summed up, to construct months holding 22 daily returns.
The average benchmark-adjusted return is the summation of all returns per month t, divided n stocks. Which results in an equally weighted portfolio of aftermarket monthly returns.
The Cumulative Abnormal Return is summation from event month q to event month s of average benchmark-adjusted returns.
Later on for the robustness tests I will use log returns denoted as . Therefore I transformed the simple returns . The formula used for this transformation:
Ritter (1991) also uses a sample of matched firms based on industry to adjust the returns for IPOs. Barber & Lyon (1996) and Gur-Gershoren et al. (2008) argue that when the sample is matched with control firms based on similar size and book-to-market ratios actually provide better results.
16
look at Alibaba, which exemplifies the issue with bandwidths. Alibaba has an average market cap of USD$ 225 billion. Amazon can be used as a peer, disregarding the fact that they operate in different markets. However Amazon has an average marketcap of USD$ 125 billion. Hence it becomes clear that firms like Google or Alibaba are not just unique in the services they provide, but also in their size.
Based on the previous, it is in my belief that the best peer to use are the returns of related market indices. Other researchers like Ritter (1991) and Dorn (2009) also make use of indices even though they do not have a global sample. Therefore I also use 3 benchmarks, which generate three adjusted returns. Table I in the data collection section provides the patterns.
V.A Data collection
Several authors of the reviewed articles mentioned the problems they encountered in creating a database. This mainly concerned in filling in the blindspots. For this research I did not have acces to IPO prospectuses. Most of the described literature focus on firms listed in the United States, for which you can use EDGAR to extract IPO prospectuses from. However my sample has a global span. Which raises issues with accessing and degree of disclosure. Therefore I used ORBIS and the M&A information records they gathered.
17
Graph 1 The blue line presents the cumulative average raw return. The red, green, and purple are the adjustments made to the raw return. The benchmarks deducted from the raw return are also cumulative average returns.
A last note on these inconsistencies is that ORBIS occasionally provided the IPO price and or opening price, which was used for the construction of our dataset. The degree of documentation ORBIS provided roughly depended on the size, year and country of the firm. In total the sample consists out of 167 firms and 30+ countries. Ranging from Chili to Japan and from New Zealand to Kenya. Besides the global impact these firms have with their services, the diversified sample also justifies the use of a global benchmark.
The services that these firms provide range from search engines (Google, Yandex), internet portals (Facebook) to digital marketing firms. Firms like Groupon were actually classified as advertising agencies. This makes it also hard to isolate the sample. However all firms in the sample have been manually checked, whether they perform and or contribute to the digital environment.
V.B Regressions
18 1)
2)
3)
= the Cumulative Abnormal Return of company “i”;; it can either be the one using the local market index (CARLI in this case), the regional market index (CARREG) or the global market index (CARGLO)
= the percentage of shares offered at the IPO moment by company “i” = the total deal value of the IPO by company “i”
= the price in dollars offered for a share at the IPO moment by company “i” = the number of shares offered by company “i” at the IPO moment
= the markup range of the company “i” at the IPO moment
= dummy variable, taking value 1 if firm “i” is a company located in the US, and 0 if it is located somewhere else
= dummy variable, taking value 1 if firm “i” is a company located in Asia, and 0 if it is located somewhere else
= dummy variable, taking value 1 if firm “i” is a company located in Europe, and 0 if it is located somewhere else
19
Because the data is built up of approximitizations and therefore might contain some errors, it was winsorized. This implies that extreme outliers at the the top and bottom 5% of the values were eliminated.
VI. Descriptive statistics
Table 1 represents the descriptive statistics of the variables. Where the CAR is subdivided in the CARLI, CARREG and CARGLO. The first are the adjusted returns matched with the index it is listed on, The second are the adjusted returns matched with regional index and the third are the adjusted returns matched with the global index. The values of the CARLI are clearly positive. However the other two CARs differ significantly and show a negative pattern. In other words benchmarking the firms with the local index shows a positive performance, while benchmarking it with the regional and global index identify an underperformance. Comparing the values between the CARLI, CARREG and CARGLO show that the minimum, maximum and standard deviation do not vary too much from each other. Looking at mean and median values of the CARREG and CARGLO it depicts that the global index shows a somewhat larger underperformance. An explanation for this variation of positive performance and underperformance, is that the average performance is more sensitive to external factors when matched with markets besides their own. And that his average underperformance is amplified when it is matched with the global index.
20
Variable Mean Median Min. Max Std. Dev. No. of Obs.
CARLI 1.648 6.975 -167.300 150.198 79.599 167
CARREG -9.217 -13.178 -177.790 146.378 81.748 167
CARGLO -17.826 -17.788 -180.118 136.278 81.518 167
IPO Percent1 .253 .250 .110 .473 0.091 113
IPO Deal Value2 82744840 35493600 2074538 3,87E+10 1,08E+09 135 IPO Price Dollar3 387.816 6.974 0.140 3.772.650 1.020.969 131
IPO Shares4 28.040 10.000 0.003 198.255 49.451 131
Markup5 0.752 0.398 0.047 2.924 0.850 80
Table 1: reports descriptive statistics of the sample from 2001-2015. 1 are the % of shares offered where 1.00 would be equal to 100% shares offered. 2 are the IPO proceeds measured in USD. 3 are the prices per share measured in USD. 4 are the amount of shares divided by a 1000000. 5 are the pre-IPO offer range.
I have provided a correlation matrix in the Appendix section. The correlation matrix can help is assessing whether there is possible multicollinearity between the variables. According to Belsey et al. (2005) a value equal or larger than 0.7 or – 0.7 can indicate multicollinearity. I observe a possible correlation between Region and Global (0.836), therefore Global was dropped from the model.
VI.A Regressions
Model 1 Model 2 Model 3
Coeff. p-value Coeff. p-value Coeff. p-value Intercept 100.153 0.828 -32.899 .449 -34.561 0.413 IPO Percent 279.003 .023** 242.160 .0567* 228.653 0.071* IPO Shares -0.352 .079* Markup 23.796 0.051* 18.951 0.100* 19.342 0.100* US -64.089 0.092* ASIA -76.521 .049** -50.332 0.041** -51.455 0.036** EU -82.504 .016** -40.086 0.075* -53.327 0.016** Adj. 0.094 0.104 0.105 R-Squared
21
Table 2 shows the basic regression results with the different dependent variables. As beforementioned the values are winsorised at 5%. The Jarque-Bera values indicated that all variables are all normally distributed after winsorization of 5%. A White-test was performed to test for possible signs autocorrelation and heteroscedasticity of the errors. The White-test did not indicate any issues with autocorrelation and heteroscedasticity. I applied HAC Newey-West to correct the standard errors and to completely rule it out.
Model 1:
Out of the first model some of the variables were dropped to test whether it increases the significance and influences the coefficient. Model 1 represents the optimal results derived from the initial model. The significant variables explain the performance to a degree of 9.4% (Adjusted R-Squared). Looking at the variables, not all impact is as expected. The percentage of ownership issued at IPO has a positive influence on the performance of the firm. The percentage of shares was measured from 0.00 - 1.00, thus an increase of 1% of shares issued, leads to an increase of 2.79% performance. This goes against Bartov et al. (2002), who suggested that it is a sign to investors, that the owner does not want to bail out, thus a lower % of shares offered would increase the price.
The number of shares offered have a negative impact on the performance, thus lead to underperformance. Since shares are measured in millions, an increase of share by 1 million leads in my sample to an increase in underperformance of 0.352%. The reason for this could be as Chalk & Peavy (1987) suggest, that an increase in shares available to trade positively influences the arbitrage opportunities. Thus investors are better able to drive prices to their true (lower) value, when there are more shares available.
An increase of 1% of the markup range leads to an increase of 23% long-term performance. Here I find again results opposed to Bartov et al. (2002). They suggest that when a firm is offered at the upperbound of pre-IPO valuations it faces of previous prices in the long-run underperformance. While my results suggest offering at the upperbound actually positively influences the performance.
Furthermore the dummy variables all show a negative impact and thus lead to underperformance. However variation per region was found. A firm located in the EU underperforms worse compared to US and Asia located firms.
The second model shows similar positive influences, however IPO Shares is not significant compared to the first model. Also the coefficients are smaller in size for both the positive and negative signs. The last model shows similar results as the second model. However the regional effect is somewhat stronger.
22
variation in strength, but not in signs. A possible explanation could be that these firms all go through the same maturity process after the IPO.
VI.B Robustness
As beforementioned I also apply log returns for the robustness tests. In general the reason to use log returns, is that it brings the values closer. Instead of eliminating outliers. For the robustness I transformed the simple returns into log returns. Hudson and Gregoriou (2010) find that the interpretation of log returns and the statistics derived from log returns is not as easy as with simple returns. They argue that the results, can therefore only be used as an approximitizational method. Because they find that it is not a 1 on1 relationship as with simple returns. They state the log returns complicate the risk and return relationship. This is because of the logarithmic values become positively biased, but are closer together, while simple returns can have a higher spread. Therefore the risk and return relationship can possibly be violated. Also the interpretation of the p-value is different, since it is now the probability and significance level of the log probability.
Variable Mean Median Min. Max Std. Dev.
No. of Obs. CARLI -0.176 -0.129 -0.934 0.482 0.374 167 CARREG -0.211 -0.192 -1.022 0.426 0.382 167 CARGLO -0.269 -0.242 -1.085 0.402 0.390 167 Table 3: are the descriptive statistics of the cumulative log returns.
23
Graph 2: The blue line presents the cumulative average log raw return. The red, green, and purple are the log adjustments made to the log raw return. The benchmarks deducted from the log raw return are also cumulative average returns.
All cumulative average log returns patterns are clearly negative and in contrast to graph 1, the raw return does not outperform the local index anymore.
Model
1 Model 2 Model 3
Coeff. p-value Coeff. Coeff. Intercept -0.108 0.055** -.200 0.000 *** -0.222 0.000 *** IPO Deal Value -0.0001 0.003*** 0.0001 0.003 *** -0.0001 0.003 *** IPO Shares 0.001 0.041** 0.001 0.002 *** 0.001 0.003 *** EU -0.163 0.031 *** -0.140 0.078 * Adj. 0.100 0.100 0.114 R-Squared
Table 4: shows the significant results from the regression
equations. Three models have been used to test three different dependent variables (LogCARLI, LogCARREG andLog CARGLO. All significant results are presented at varying significance levels. *** is at 1% ** is at 5% and * at 10%
24
an increase of USD$10.000,- decreases the log return by 1%. Dorn (2009) also implied that there might be a tendency for investors to invest in ‘glamour stocks’. Besides the influence the log returns have on the values of the coefficient, it is also because the deal value is measured in US dollars. Based on these results the local and global benchmarks provide support for the first hypothesis.
The second observation is that for all the dependent variables, the amount of shares offered at the IPO have a positive influence on performance. The measurement unit for the float was per 1 million shares. Hence, an increase of 1billion shares increases the logreturns by 1%. Accordingly to hypothesis four I expected opposite results. An interpretation of this is something to ponder about, since the literature suggests that the more stocks available to trade the more arbitrage opportunities arise. Therefore the price of the underlying asset can more easily be driven to its true value. And since graph 2 depicts a clear underperformance it is somewhat confusing. Especially since the literature does not provide an answer. A plausible explanation could be, that because of the increased amount of shares, leads to a less concentrated ownership. However this only holds when investors positvely appreciate it, when a firm is less dominated by block holders. I only found significance dummy variable for the local and global index. Where Europe shows a clear negative relationship, thus a firm located in Europe, contributes to the underperformance.
VII. Conclusion & implications
25
European internetfirms show significantly more underperformance. This can be used for furture research and practice. Also the last hypothesis showed an opposite result. Apparently IPOs priced at the upperbound of the price range constructed through pre-IPO period have a positive influence on the performance. It is important to note, that this independent variable is not the same as underpricing. Secondly the range was constructed out of the values reported and not out of prospectuses.
What the paper does find is a clear underperformance for internet firms. Hence, the intention of the paper to identify the performance of this novel industry is fulfilled. Secondly I find evidence, that investors are attracted to glamour stocks. And that his actually contributes to the underperformance. Altough Dorn (2009) describes this as a clear behavioral action, of investors being influenced by media exposure and the size of stocks. I cannot provide a clear stand in this, due to the mixing results. Nor that it characterizes specifically internet firms. Applying the three different benchmarks contributed to the mixed results in explaining the underperformance. The literature used does not make a distinction in the preferences of investors, just in rational and or sentiment investors. It could also be that investors have a reluctance and or eagerness to invest in foreign stocks. This could explain the mixed regional results. The differences in performance per benchmark could also be due to maturity and business model of the firm. Many firms use the IPO proceeds to expand (globally). On a local level, the firm might be considered as unique and thus outperform the local index. But on a regional and global level it has to compete with more firms pursuing similar goals. These existing firms make up the regional and global index, and create a higher average return. So the benchmark return is higher and the IPO firm performance is lower, because the IPO firm is perceived less unique.
The benchmarks show different results, the conclusion should be that it is an attempt to show the different patterns and should lead to more research on whether there is an ideal method to measure long-run performance. Because the ability to define abnormality lies in the ability to define normality.
The implications of this research depend on the perspective. This paper mainly looked at the position of investors. Based on the simple returns, the long-run pattern shows a clear underperformance. Graph 1 suggests, that long positions at the start of the IPO are only positive after roughly 20 months. But only when you invest solely in the index of the IPO where the firm is listed, since the regional index and global index have a better performance than the IPO firm itself. However for an investor to constantly rebalance their portfolio based on a regional and or global scale, can be a tedious thing to do.
26
issuers are apart from what they say13. For this information asymmetry, legal technicalities are built in. Like lock-up periods which prohibit the issuer and or management to sell shares after IPO for a certain period.
For underwriters the main risk is the underperformance and a loss of reputation. They provide the valuations and forecasts of the IPO firms. A clear form of underperformance can increase the risk of legal cases against them, since investors base their investments on the valuations and forecasts the underwiter gives.
I can conclude with the remark of ‘caveat emptor’, meaning ‘buyer be on your guard’. Currently there are a lot of rumoured IPOs of firms that provide services, which any person can understand. Spotify and Uber are great examples of this. And even though they might be framed as the next big thing, the results show an opposite pattern. The next step for future research is to be able to isolate the winners from the losers.
VIII. Limitations & extensions
First of all this research has been about second generation listed internet companies. A clear definition of these firms is lacking. One of the reasons is that not just for investors but also for the general public, the firms operate in different industries but all rely on the same medium. Whether it is a music streaming firm like Spotify or a search engine called Google, they impact life in different ways. What can be noticed is that these second generation of internet firms turn conventional business models and industries upside down. Also most of the conventional firms did not take the first step to make use of the internet, hence these second generation internet firms are all new to the industry they want to change.
Secondly all the research in regard to the data relied heavily on the data retrieval programs ORBIS, Bloomberg and Datastream. These programs in turn impose restrictions and limitations. Not all data could be retrieved through internal errors and or unavailability. Secondly one is bound to the limited sampling tools Orbis provides. Therefore a clear distinction of social media firms could not be identified. The solution I used is to backtrack iconic firms and used their industry codes to identify the sample. The manualy filled in dataset I used was depending heavily on the records, besides a possible bias in the data. It was not an ideal set, due to the missing values. A more complete set from one source and possibly verified by another, could possibly have resulted in stronger and or (more) significant results.
13 The following url shows a video interview in which one of the former owners of Instagram
27
What does characterize these firms and the industry, is that a large part of them is not listed, but is privately financed. It could be possible that these firms eventually will pursue a public listing, but since most firms are not older than 5-10 years it will remain to be seen. Therefore it is also a suggestion for further research to focus on these privately backed firms. There is however a firm called Rocket Internet, which is only recently listed which invests in these not-listed second generation internet firms. Thus the performance of the firms can possibly be deduced from the performance of Rocket Internet.
Altough some possible extensions for further research were given throughout the paper, I wish to address some in specific. Engelen & Essen (2010) find that country characteristics are able to explain underpricing. My study shows that there is a significant variation of regional performance of IPOs. Defining these country and or regional characteristics for long-run performance of IPOs, could be of interest for further research.
I also indicated that the bookbuilding process is not transparent for most of the parties except the underwriter. Dorn (2009) was researched a dataset of bookbuilding of one underwriter. Acquiring this data appears to be exceptional, but can provide unique insights the underpricing mechanism.
Brav et al. (2000) indicates that the amount of shares offered could provide a signal for underpricing. However there is a large diversity in legal contracts involved with IPOs. As my results show, the amount of shares is significant. Lock-up periods for management, prohibiting them to sell for a certain period can be of influence on the long-run performance.
Lastly I did not discuss the influence of M&A activities of these internet firms. I cannot forgo on the acquisition spree that these internet firms are involved in. Microsoft acquiring Skype for USD$ 8.5 billion or Facebook buying WhatsApp for a whopping USD$ 19 billion. And than I have not even touched upon all the non-related firms Google is buying. Is it an excess of money and or a way to compete with others. Thus buying something so an other cannot have it? In any way they do this with the shareholder money, which impacts value creation and or value destruction and ultimately the long-run performance of internet firms.
IX. Bibliography
Barber, B., & Lyon, J. (1997). Detecting Long-run Abnormal Stock Returns: The Emperical Power and Specification of Test Statistics. Journal of Financial Economics , 341-372.
28
Beatty, R., & Ritter, J. (1986). Investment Banking, Reputation, and the Underpricing of Initial Public Offerings. Journal of Financial Economics , 213-232.
Belsey, D., Kuh, A. & Welsch, R. (2005). Regression diagnostics: Identifying influential data and sources of collinearity. book (vol. 571), John Wiley & Sons.
Benveniste, L., & Spindt, P. (1989). How Investment Bankers Determine the Offer Price and Allocation of New Issues. Journal of Financial Economics, 343-361.
Bondt, W. de. (2004). Introduction to the Special Issue on Behavioral Finance. Journal of
Empirical Finance , 423-427.
Brav, A., Geczy, C., & Gompers, P. (2000). Is the Abnormal Return following Equity Issuances Anomalous. Journal of Financial Economics , 209-249.
Campbell, C., Du, Y., Rhee, S., & Tang, N. (2008). Market Sentiment, IPO Underpricing, and Valuation. working paper .
Chalk, A., & Peavy, J. (1987). Initial Public Offerings: Daily Returns, Offering Types and the Price Effect. Financial Analysts Journal , 65-69.
Chemmanur, T., & He, J. (2011). IPO Waves, Product Market Competition, and the going Public Decision: Theory and Evidence. Journal of Financial Economics , 382-412.
Cook, D., Kieschnick, R., & Ness, R. V. (2006). On the Marketing of IPOs. Journal of Financial
Economics , 35-61.
Dorn, D. (2009). Does Sentiment Drive the Retail Demand for IPOs? Journal of Financial and
Quantitative Analysis , 85-108.
Engelen, P., & Essen, M. v. (2010). Underpricing of IPOs: Firm-, Issue- and Country-specific Characteristics. Journal of Banking & Finance , 1958-1969.
Franklin, A. & Faulhaber, G., (1989), Signalling by Underpricing in the IPO Market. Journal of
Financial Economics, 303-323
Gwilym, R. a. (2010). Can Behavioral Finance Models Account for Historical Asset Prices?
Economic Letters , 187-189.
Gur-Gershgoren, G., Hughson, E. & Zender, J. (2008) A Simple-But-Powerful Test for Long-Run Event Studies. Robert Day School of Economics and Finance, Research Paper.
29
Hand, J. (2000). The Role of Economic Fundamentals, Web Traffic, and Supply and Demand in the Pricing of U.S. Internet Stocks. working paper .
Higson, C., & Briginshaw, J. (2000). Valuing Internet Businesses. Business Strategy Review , 10-20.
Ibbotson, R. (1975). Price Performance of Common Stock New Issues. journal of financial
economics, 235-272
Ibbotson, R. & Jaffe, J., (1975), Hot issue” markets. The journal of finance, 1027-1042.
Ljungqvist, A., Nanda, V., & Singh, R. (2006). Hot Markets, Investor Sentiment, and IPO Pricing. Journal of Business, 1667-1702.
Ljunqvist, A. & Wilhelm, J. (2003). IPO Pricing in the Dot-com Bubble. The Journal of Finance, 723-752.
Loughran, T. & Ritter, J. (2004), Why has IPO Underpricing Changed over Time. Financial
Management, 5-37.
Mcdonald, J. & Fisher, A., (1972). New-Issue Stock Price Behavior. The Journal of Finance, 97-102.
Ofek, E. & Richardson, M. (2002). The Valuation and Market Rationality of Internet Stock Prices. Oxford Review of Economic Policy , 265-287.
Purnanandam, A., & Swaminathan, B. (2004). Are IPOs Really Underpriced? The Review of
Financial Studies , 811-847.
Ritter, J. (1984), The Hot Issue Market of 1980. Journal of Business, 215-240 Ritter, J. (2003). Behavioural Finance. Pacific-Basin Finance Journal , 429-437.
Ritter, J. (1991). The Long-run Performance of Initial Public Offerings. Journal of Finance , 3-27.
Rock. K. (1986). Why new issues are underpriced. Journal of Financial Economics, 187-212 Schultz, P. (2003). Pseudo Market Timing and the Long-run Underperformance of IPOs. Journal
of Finance , 483-517.
Stracca, L. (2004). Behavioral Finance and Asset Prices: Where do we Stand. Journal of
Economic Psychology , 373-405.
30
Trueman, B., Wong, F., & Zhang, X. (2000). The Eyeballs have it: Searching for the Value in Internet Stocks. Journal of Accounting Research , 137-162.
Trueman, B., Wong, F., & Zhang, X. (2001). Back to Basics: Forecasting the Revenues of Internet Firms. Review of Accounting Studies , 305-329.
Trueman, B., Wong, F., & Zhang, X. (2003). Anomalous Stock Returns around Internet Firms' Earnings Announcements. Journal of Accounting and Economics , 249-271.
Welch, I. (1989), Seasoned Offerings, Imitation Costs, and the Underpricing of Initial Public Offerings, The Journal of Finance, 421-449
Yung, C., Colak, G., & Wang, W. (2008). Cycles in the IPO Market. Journal of Financial
Economics , 192-208.
31
Appendix
Table 4 Correlation Matrix
Graph 3 IPO Distrubtion Per Year
