The stock effect of initial public offerings on industry rivals.

Transcription

1 The stock effect of initial public offerings on industry rivals. Master Thesis MSc Finance TiU, Tilburg University Supervisor: Dr. M. van Bremen Second reader: Dr. D. Hollanders Author: Lotte Brands, Date of completion: 28 th of November, 2014

2 Abstract This thesis examines whether an initial public offering (IPO) affects the stock price of listed rivals in the same four-digit SIC industry and if the level of homogeneity in an industry as well as the number of days between the announcement and completion date of the IPO significantly influence this valuation effect. Companies that go public often underprice their shares guaranteeing a positive firstday return for investors which makes their shares more attractive than those of competitors. Using a sample of 385 IPOs and 1321 accompanying rivals from , a significant average cumulative abnormal return of -0.68% is found for industry rivals over the 21-day period surrounding the IPO. The relations between the level of homogeneity and the cumulative abnormal return and between the lag between the announcement and completion date of an IPO and the cumulative abnormal return are tested using an OLS regression. There is no evidence found for both relations. 1

3 Table of contents Abstract Introduction Literature review and hypothesis development What are the most common reasons for a company to go public? How do IPOs perform after listing? What are the competitive effects of an IPO? Hypotheses Methods Sample construction Variables Dependent variable Independent variables Control variables Analyses Univariate regression model Multivariate regression model Results Descriptive statistics Univariate regression analysis Multivariate regression analysis Conclusion Discussion Limitations and recommendations Conclusion References Appendices

4 1. Introduction Every year, many companies worldwide decide to go public through an initial public offering. An initial public offering is hereinafter defined as: IPO. An IPO is the first time that shares in a company are sold to public investors and subsequently traded on the stock market (Draho, 2004, p. 1). IPOs are widely researched. A large part of the existing literature has focused on the underpricing of IPOs, which measures the difference between the offer price of an IPO and the closing price on the first day that the IPO trades. Furthermore the short and long run performance of the IPO firm are topics researched often, in particular since it provides valuable information to investors. This thesis focuses on another phenomenon relating IPOs. IPOs could raise billions of dollars to the stock market in one day. An example is Facebook, which went public in May 2012 and sold over 421 million shares for over $16 billion in total. 1 Such enormous amounts can lead to changes on the stock market other than the availability of new shares. The arrival of Facebook on NASDAQ probably affected the stock prices of other social networking sites that were already listed on NASDAQ, or even on NYSE or AMEX. If investors at investment companies or private investors could anticipate to such price changes by making investment decisions based on reliable predictions this would furthermore add value. The reason to suggest that IPOs might have a valuation effect on existing rivals in the industry is because similar research has already been conducted outside the IPO setting. A significant effect for industry rivals is shown for instance for bankruptcy announcements by Lang & Stulz (1992) and for bond rating downgrades by Akhigbe, Madura & Whyte (1997). In this study it is checked whether an IPO affects the stock price of industry rivals on the stock market. Many traders are instantly engaged in the buying and selling of shares on the stock exchange in order to realize positive returns. For them, it is interesting to know how listed firms react to an IPO in their industry. Although some researchers (e.g. Akhigbe, Borde & Whyte, 2003 and McGilvery, Faff & Pathand, 2012) already investigated this topic partly, there still exist elements that could be related to this valuation effect about which investors are still in the dark. This thesis aims to contribute to the theory concerning some of these factors and also tries to gain additional evidence regarding some of the patterns that already have been documented. In specific, this thesis provides an answer to the following question: do industry rivals experience a decrease in their stock price in response to an 1 Retrieved from in November

5 initial public offering and how is this influenced by the degree of homogeneity in the industry and by the lag between the announcement date and the completion date of the initial public offering? To provide answers to this question a sample of IPOs from occurring on the NYSE, AMEX or NASDAQ is taken. All rival firms trading in the same four-digit SIC code industry as the IPO are downloaded in order to calculate the valuation effect. This valuation effect is determined by calculating the cumulative abnormal return (CAR) for each rival firm, which is done by using the OLS market model on the 21-day period surrounding the IPO completion date. The mean CAR is negative and significant. To give insight in the effect of the time lag between the announcement date and the completion date of each IPO and the type of industry the IPO operates in, an OLS regression is conducted. The results for the level of homogeneity in an industry as well as the results for the time lag are insignificant. The remainder of this thesis is structured as follows. In chapter 2, a review of existing literature regarding IPOs is presented after which three hypotheses are developed. The methods chapter describes the sample construction, outlines the different variables and demonstrates how the univariate and multivariate analyses are built up. The empirical results of the analyses are presented in chapter 4 and finally in the last chapter these results are discussed and compared to the existing literature and expectations. Thereafter are some limitations of this thesis and recommendations for further research documented. The last section briefly summarizes the findings of this thesis. 4

6 2. Literature review and hypothesis development In this chapter the theoretical background related to initial public offerings (IPOs) is described. First, in section 2.1, the reasons for why companies go public are explained. After that, section 2.2 will focus on the effects of an IPO s own performance after listing, where both short term and long term performance are discussed. Thereafter, the competitive effects of IPOs are explained and finally in section 2.4 the hypotheses are developed based on the discussed literature. 2.1 What are the most common reasons for a company to go public? When management of a company is deciding about whether to go public or not they weigh a lot of advantages and disadvantages, which will be discussed in this section. Pagano, Panetta & Zingales (1998) stated that going public is often just regarded as the next step of growth for a company. However, looking at existing companies and previous listing-waves they immediately made clear that there is much more to consider when analyzing a company s decision to go public. They pointed at some very large companies in the United States that were still private, proving that an IPO is not necessarily needed for growth. The first advantage of going public is that a new source of finance becomes available for the company (Röell, 1996). When a company enters the stock market it sells stock to public investors, which means the company accesses an alternative source of finance. According to Röell (1996), recently listed companies can use extra finance to acquire other companies, repay current loans, make capital investments or assure stable future growth. Pagano et al. (1998) explored which of the finance advantages were most important for Italian firms that decided to go public and it appeared that Italian firms did not go public to finance investments and growth but mostly to rebalance their accounts after a period of high investment and growth. Secondly, going public is often used as a marketing tool. Röell (1996) stated that going public improves a company s image and yields immediate publicity. Once a company is listed, its stock price signals information to suppliers, workforces and customers. This can make the company for instance a more reliable employer or give customers confidence about product quality. A third benefit of going public is the decrease in borrowing costs that follows. With the fact that it is a company s duty to share financial information publicly it follows that the bank does not have 5

7 privileged information anymore since every possible lender can access the information (Pegano et al., 1998). As a result, competition between lenders of money is maximal which decreases the cost of borrowing. Diversification is another advantage of going public. Rock (1986) mentioned diversification as one of the main reasons to go public and stated that the reason for this is that the owners, pensioners and financial backers of the firm are risk averse and try to spread risks. As Pagano (1993) pointed out, sometimes going public is the only possible way for a company to diversify, for instance because the company faces a lack of liquidity or borrowing constraints. Lastly, an advantage of going public is that agency costs can be minimized by motivating employees and management. When ownership and control are separated for a company, so-called agency costs can arise (Jensen & Meckling, 1976). Management might take decisions to maximize the wealth of the company instead of maximizing shareholder value. Shareholders of a public company can therefore decide to give management a managerial performance-linked compensation based on the stock price, to increase their motivation and decrease agency costs (Röell, 1996). Furthermore Röell (1996) stated that an IPO motivates employees simply because employees recognize the growth intentions of the company. On the contrary, a negative consequence of going public is the fact that it entails direct and indirect costs. Ritter (1987) came up with proof about the direct expenses of going public and the indirect costs of underpricing. He displayed the direct expenses of going public as a percentage of gross proceeds and found that the evidence indicates that the direct costs of going public are equal to approximately $250,000 plus 7% of the gross proceeds (Ritter, 1987, p. 273). Furthermore he calculated underpricing based upon the first day s closing bid price and found an average underpricing of 14.80% respectively 47.78% for firm commitment offers and best offers. Loss of confidentiality can be regarded as a negative aspect of going public. When a company goes public it has to publish its annual figures. Pegano et al. (1998) acknowledge the loss of confidentiality as an expense of going public because companies can possibly lose their competitive advantage when they are obliged to present certain information publicly, for instance on research and development (R&D). Another example is that public companies are more visible for tax authorities which makes it harder to evade tax. Another problem that companies face when they are willing to sell their shares on the stock market is the decrease in outside shareholder monitoring. The true value of a project is not always clear 6

8 because moral hazard prevents parties from the direct transfer of information (Leland & Pyle, 1977). Chemmanur & Fulghieri (1999) found that after an IPO, outside shareholder monitoring decreases. This is because there exists a free rider problem with monitoring. The company sold shares to a lot of individual investors. Only a small part of the shareholders share in the costs of monitoring, while all shareholders benefit. Therefore the incentives to monitor often disappears (partly) after the IPO because of dispersed ownership, which causes operating performance to decline. Sometimes one of the advantages is not the reason per se for the decision to go public but a tool the company uses necessary to achieve another goal, which is the underlying reason for the company to go public. An example is the acquisition motive of going public researched by Celikyurt, Sevilir & Shivdasani (2010) who showed that more IPOs occur in mergers and acquisitions intensive industries and concluded that a company often goes public to be able to acquire another company. Some of the disadvantages of going public are more disadvantageous in particular industries. De Jong, Huijgen, Marra & Roosenboom (2012) found that in a competitive industry, companies are not very likely to go public because they lose private information to competitors and this outweighs the benefits. Concluding this section, it can be said that a company only goes public when management considered all options and decided that that the advantages of going public will probably outweigh the disadvantages in the near future, which increases the company s performance. Therefore, in the context of this research, it is expected that industry rivals listed on the stock market face an increase in competition when a company in their industry goes public, which decreases their stock price. This will be explained more closely in section 2.3, where the question: What are the competitive effects of an IPO? will be answered. 2.2 How do IPOs perform after listing? Companies that go public often underprice their shares in order to guarantee positive initial returns for the new shareholders. A large part of the existing literature has focused on the underpricing of IPOs, which measures the difference between the offer price of an IPO and the closing price on the first day the IPO trades. Ibbotson (1975) found evidence for underpricing, after combining the positive initial performance for IPOs during the 1960s with the aspect of aftermarket efficiency. The underwriter plays a major role in the underpricing of IPOs since the underwriter regulates the 7

9 market. If the offer price was purely determined by the market it would be as high as investors would pay with the consequence that investors do not expect to earn any abnormal returns, which makes the stock unattractive. The underwriter sets a price before the offering which means that if the value of the new issue is greater, investors can earn immediate profits. So, in order to compensate investors for the risk they take, IPOs are often underpriced resulting in positive initial performance. Because of the efficiency of the market this is usually temporary. Similar to Ibbotson (1975), Rock (1986) found that investors are less informed than issuers and therefore require returns to compensate the risk. Ritter (1984) pointed out particular industries as the determinant of underpricing. During his sample period there were a few moments that the monthly average initial returns were extremely high which, in turn, led to periods with more IPOs. Ritter researched the hot issue market of 1980 and found that the average initial return was 48.4%, while the remainder of his sample period gave an average initial return of only 16.3%. He stated that this was probably related to the oil and gas boom that arose during In later research, Ritter (1991) showed that the performance of IPOs was lower than the performance of the benchmark group that consisted of similar firms that did not go public. He found that IPOs appear to be overpriced in the long run and show underperformance, at least for the first three years after the IPO. Ritter showed a 3 year holding period return of 34.47% for a sample of IPOs versus a 3 year holding period return of 61.86% for the control sample. The poor performance has also been found by others for other sample periods and markets (e.g. Lee, Taylor & Walter, 1996). Since companies underprice their stock to guarantee positive initial returns, rival firms will suffer from an IPO at least temporarily. As this study focuses on the short term effect of IPOs on industry rivals, namely around the IPO s completion date, the findings of long term underperformance are not important. By underpricing shares, an IPO firm makes their stock immediately attractive for investors. It is therefore plausible that investors buy shares from the company that goes public rather than from the listed rival firms, at least around the IPO completion date. This implies that IPOs have a negative effect on the stock price of rival firms. 2.3 What are the competitive effects of an IPO? Earlier researchers have shown rival valuation effects for capital market transactions other than IPOs and also the effect of IPOs on industry rivals has already been examined. Significance is shown for 8

10 different variables tested in relation with the rival valuation effect of IPOs and outlined in this section to form a framework for this thesis. Listed firms react on information proceeded by other firms on the stock market. For instance, Lang & Stulz (1992) found that bankruptcy announcements have a significant effect on rival firms and Akhigbe et al. (1997) showed significant negative effects for rival firms after bond rating downgrades. An industry effect is dependent on the type of information the capital market transaction of interest releases, which is either firm-specific or industry-wide (Slovin, Sushka & Polonchek, 1992). Slovin et al. (1992) found that announcements of common stock issuances of individual banks are perceived as a negative signal by other commercial and investment banking firms whereas they did not found a negative stock price reaction for rivals of industrial firms announcing a common stock issuance. This implicates that an issue of stock releases only firm-specific information. If the announcement or completion of an IPO solely releases firm-specific information, it is expected no significant industry effect will be demonstrated. Such a thing was found by Howe & Shen (1998) for dividend initiations. However, as already explained in paragraph 2.1, some of the advantages and disadvantages of going public do have industry-wide implications and therefore this study hypothesizes that a valuation effect for industry rivals exists. Slovin, Sushka & Ferraro (1995) did show a negative stock price reaction for rival firms to conventional IPOs. They examined the valuation effects of carve-outs, spin-offs and sell-offs for firms operating in the same industry and found evidence that equity carve-outs and conventional IPOs caused a negative reaction of about 1% on rivals stock returns. For spin-offs and asset sell-offs no negative stock returns were measured for rivals. An industry effect was shown but not tested as the dependent variable yet. IPOs can induce two things: signal information to the whole industry or actually change the competitive balance in an industry (Akhigbe et al., 2003). An example can be taken from the study of Stoughton, Wong & Zechner (2001) who focused on the link between IPOs and product quality and showed that consumers retrieve product quality information from stock prices. Their model suggested that an IPO signals higher product quality when the initial return is high, which is bad news for competitors since they have to lower their prices to compete. Akhigbe et al. (2003) focused their research on the industry effect of IPOs by distinguishing two sets of independent variables. On the one hand by testing for information effects, which they described as relevant information that an IPO can convey to the industry as a whole and on the other hand by testing for competitive effects, which 9

11 they described as an IPO changing the competitive balance within the industry. Although Akhigbe et al. (2003) did not show significance for the industry effect, they did show significance for information and competitive effects in their cross-sectional model. According to Hsu, Reed & Rocholl (2010), who performed a similar research, Akhigbe et al. (2003) did not find a significant valuation effect because of their choice of sample selection. Although Akhigbe et al. (2003) did not show significance for the industry effect, other researchers reexamined this topic and showed significant abnormal returns for rivals in the industry after a company went public (e.g. Hsu et al., 2010). Hsu et al. (2010) expanded their research compared to Akhigbe et al. (2003) and also looked at rivals stock prices around withdrawn IPOs, IPOs that were announced but withdrawn rather than completed, and, as expected, industry competitors experience a positive stock price reaction to the withdrawal of an IPO. For completed IPOs they showed a significant negative stock price reaction. The opposite, the existence of an industry effect for firms going private, has also been investigated. Slovin, Sushka & Bendeck (1991) found that rival firms' stock price reactions were positive to a firm going private. Reverse thinking supports the finding that an IPO causes negative stock price reactions for rival firms in the industry. After the research of Hsu et al. (2010), the industry effect for IPOs has been investigated repeatedly for different industries and alternated with different areas of interest as the main angle of incidence. Lee, Bach & Baik (2011) highlighted, for instance, the valuation effects an IPO can have in a growing industry. The computer-related service industry (SICcodes 737) was therefore included as their sample. They showed IPOs in uncertain markets are not always beneficial for incumbent firms, depending on the closeness of competition and the level of R&D investments of a company. If a market segment is more concentrated, it is more likely that directly competing incumbents will suffer when another firm announces an IPO. A year later, McGilvery et al. (2012) researched the valuation effects of Australian IPOs by integrating the corporate governance profile of the IPO and the intended use of the offer proceeds in their model. They also found a negative stock price reaction to the completion of an IPO. Braun & Larrain (2009) researched the industry effect for IPOs in the adjusted setting of an emerging market. Since emerging markets are not (fully) integrated with the international market, IPOs are relatively big and therefore cause bigger shocks. They found a decline in prices when there is a high covariance between the IPO and the portfolio researched. The negative price effect is stronger when the market is less integrated internationally and when the IPO is bigger. This last section showed that many factors, like the type of industry, market conditions or company specifications, influence the strength of the industry effect. The degree of homogeneity in an industry, 10

12 which is researched by Kohers (1999) in relation with dividend initiations and omissions is not researched for IPOs yet and therefore of main interest in this study. The last section will shortly go through important parts of the literature to develop the hypotheses that will be tested in this thesis. 2.4 Hypotheses This research examines the valuation effect that initial public offerings (IPOs) cause for listed industry rivals and to what extent this can be explained by the degree of homogeneity in the industry. The focus will lie on rival returns around completed IPOs. More specifically, this thesis tries to answer the following research question: Do industry rivals experience a decrease in their stock price in response to an initial public offering and how is this influenced by the degree of homogeneity in the industry and by the time lag between the announcement date and the completion date of the initial public offering? Two different information effects are described in earlier literature, by for instance McGilvery et al. (2012), which are interesting when researching the effect of IPOs. On the one hand a contagion effect is possible, an IPO would then signal positive information to investors which likely causes competitors to benefit from the entry of a new company. On the other hand a competitive effect is possible, in that case an IPO signals negative information to investors with the consequence that competitors stock prices decrease. In section 2.3 some examples of earlier research are given and it is shown that in the case of an IPO as the corporate event of interest, generally stock prices of rival firms declined. As explained in sections 2.1 and 2.2, this is because a company only goes public when management believes going public increases company performance and because the IPO company underprices their shares the first days of trading to make them more attractive for investors. This entails that an IPO is usually perceived as a bad news announcement by its competitors. This study will touch upon a new era and focusses on the last decade, starting after the internet bubble from The American market is of interest and since there are no reasons to believe that the American market is significantly different from most markets used in earlier research the following can be hypothesized: H1: IPO completion results in a negative cumulative abnormal return for industry rivals. This study builds further on the research done by Akhigbe et al. (2003) which examined the effects of IPOs on rivals by making a distinction between two types of signals an IPO could convey. On the 11

13 one hand an IPO could signal something to the industry as a whole, while on the other hand it is possible that the IPO changes something within the industry, which changes the relative competitiveness of rival firms. When searching for the effect of IPOs on existing rivals, firms are identified as rivals when their SIC code matches the IPO s, which means this study focuses on changes within the industry rather than on the stock market as a whole. As explained in section 2.3, it is not likely that significance can be demonstrated when an IPO only signals firm-specific information. However, according to Kohers (1999) the degree of homogeneity in an industry, which is high when firms in an industry are relatively similar to each other, influences the industry effect. When a company in an industry with a high degree of homogeneity makes an announcement (in the case of Kohers about dividend omissions and initiation), investors may interpret a firm-specific announcement as information that has industry-wide implications (p. 140). A consequence would be that returns in industries with a high degree of homogeneity are stronger related to the announcement firm than returns in industries with a low degree of homogeneity. Although Kohers (1999) did not show significance for the level of homogeneity in an industry, an effect was showed. This thesis will test if the level of homogeneity in the industry of the IPO affects the stock price reaction of industry rivals. Since investors in an industry with a high degree of homogeneity will react faster (more) to firm-specific information it is expected that the decrease in industry rival s stock price is larger in industries with a degree of homogeneity and therefore the following is hypothesized: H2: A high level of homogeneity in an industry is negatively related to the cumulative abnormal return of industry rivals. McGilvery et al. (2012) and Lee et al. (2011) researched, like many others, the effect of IPOs on existing rivals on two dates, the announcement and the completion and the announcement and the filing date, respectively. When the announcement date, which is the first day the market receives information about the IPO and the completion date, which is the actual IPO date, lie close together the effects on industry rivals around the completion of an IPO might differ compared to when a big time lag between the two dates exists. According to Lucey & Dowling (2005) investors are partly led by feelings when making investment decisions. An example they showed is that investors buy shares more easily when they like a company. When a reasonable time lag exists between the announcement and completion date of an IPO investors feelings concerning the IPO firm could change, e.g. by news articles published in the media. So when investors have more time to decide whether they want to 12

14 buy the shares of an IPO or not, simply because the time lag is bigger, their investment decision could change. Therefore it is hypothesized that: H3: The time lag between the announcement date and the completion date of an IPO is related to the cumulative abnormal return of industry rivals. 13

15 3. Methods This chapter describes the selection and construction of the sample used in this research and provides the regression models used to test the valuation effect of initial public offerings on industry rivals. Throughout the paragraphs in section 3.1, the sample omissions are described resulting in a complete overview of the sample. Section 3.2 describes the different variables used in this study and explains the event study methodology which is used to derive the cumulative abnormal return of industry rivals, which is included as the dependent variable in the multivariate analysis. The third section outlines the regression models that are included to test the different hypotheses. Zephyr 3.1 Sample construction To construct a sample with all completed IPOs necessary to conduct this research, Zephyr 2 is used. Zephyr is chosen because of the database s high quality data that is updated hourly and covers over ten years of history around the world. The second hypothesis needs information on both the announcement date and the completion date of every IPO which Zephyr both provides. Compared to the SDC Platinum database of Thomson Financial and Mergerstat, which is used in many other IPO studies, Zephyr also covers information on deals of smaller value. Zephyr is used to obtain 1158 IPOs for the years 2001 to 2012 occurring on the NASDAQ national market, NADAQ/NMS (Global Market), NYSE, or AMEX. Only the IPO deals for which the deal value is known by Zephyr are downloaded. According to Loughran & Ritter (2004), who investigated the difference in underpricing of IPOs over years, in the years of the internet bubble period, , underpricing jumped to another level. Therefore it is chosen to start the sample from In addition, the rival valuation effect for American firms is not yet researched for this last decade and therefore interesting. In this study, the primary US Standard Industrial Classification (SIC) code of each IPO is used as identifier to look for rival companies operating in the same industry. Some of the IPOs in the Zephyr output do not have a four-digit SIC code (referring to 1 of the 1005 industries) but only a two-digit SIC code (referring to 1 of the 83 major groups) or a three-digit SIC code (referring to 1 of the 416 industry groups) 3. Since this study will focus on rival firms in the same four-digit industry, IPOs with 2 A database from Bureau van Dijk that provides comprehensive deal data with integrated detailed company information for M&A, IPO, private equity, venture capital deals and rumours. 3 Retrieved from on

16 two- or three-digit SIC codes and IPOs without a SIC code are deleted. Consequently, 17 IPOs are deleted from the sample. Also the IPOs without a ticker or with the ticker unlisted are removed. This removal is necessary because the ticker is the company s identifier and used to remove the IPOs that do not have data available in the rival sample (e.g. because they already delisted and Zephyr did not have this information available). Next to this, Following Hsu et al. (2003) and McGilvery et al. (2012) all financial firms are excluded from the sample which is supported by Slovin et al. (1992), who found a significant intra-industry effect for banks after a common stock issuance announcement but did not find this effect for rivals of industrial firms. All IPOs with an SIC code between are therefore removed as well. IPOs with a deal value below $50,000,000 are removed because it is expected that such small IPOs have a weak effect on rivals. Akhigbe et al. (2003) included all IPOs and did not find a significant rival valuation effect. IPOs that were delisted or not found in the Center for Research in Security Prices (CRSP) 4 are removed. This removal is necessary because only for the IPOs that occurred on one of the stock exchanges that CRSP covers, the industry effect can be calculated. To be able to calculate the industry effect also the IPOs that do not have rivals on the NYSE, AMEX or NASDAQ around the IPO date or firms that do not have enough days of stock data available to conduct an event study are excluded from the sample which leaves a final sample of 385 IPOs. This sample of IPOs is later used as input for the event study. An overview of the sample construction is given in table 1. Table 1: Sample Construction This table shows the sample selection consisting of 385 IPOs from until retrieved from Zephyr in September Number of firms Total Initial Public Offerings 1158 Less: not a four-digit SIC code 17 Less: financial firms (SIC ) 336 Less: IPOs with a deal value < $50,000, Less: ticker unlisted / missing 20 Less: IPOs not found in CRSP 200 Less: no rival stock information 90 Total CRSP preserves an extensive database with security prices, returns, and volume data for the NYSE, AMEX and NASDAQ 15

17 CRSP In this research is focused on the performance of listed industry rivals after an IPO rather than the performance of the IPO itself on which is focused in many earlier research. In order to conduct an event study, stock information on listed industry rivals is needed. Wharton Research Data Services (WRDS) 5 is used to consult the Center for Research in Security Prices (CRSP) to download daily returns for all existing companies with the same four-digit SIC code as the IPOs in the sample. Those firms in the same industry as the IPO are defined as industry rivals in this study. The rival firms will be matched with the IPO events based on their four-digit SIC codes and therefore all rivals that do not have a four-digit SIC code are removed. All IPOs are also saved in the industry rival sample, since frequently an IPO will function as a rival company to another IPO in the same industry later in the sample period. When conducting the event study, industry rivals that do not have enough observations available are dropped. Subsequently the inclusion of IPOs in the rival sample will not cause any problems. Compustat The accounting and market data needed to create the control variables, which will be explained later in section 3.4, is downloaded from Zephyr for the IPO firms and from CRSP and Compustat North America 6 for the industry rivals. Compustat North America contains information in fundamentals annually on many data items for the firms in the database and is updated monthly. After combining the accounting and market data, the final sample includes 1321 industry rivals. Some industry rivals are used several times to calculate abnormal returns around different IPOs because in the sample period up to 35 IPOs occurred in one industry. A total of 7274 industry rival observations is therefore included in this research. 5 WRDS is the leading data research platform from The Warton School of the University of Pennsylvania which is used worldwide in 33 countries. 6 S&P Capital IQ's Compustat North America is a database of U.S. and Canadian fundamental and market information on active and inactive publicly held companies. It provides more than 300 annual and 100 quarterly Income Statement, Balance Sheet, Statement of Cash Flows, and supplemental data items. Retrieved November 1 st 2014 from 16

18 3.2 Variables Dependent variable Cumulative abnormal return Since IPOs are unexpected events for the market and provide new information to it, it is likely that IPOs have a financial impact. The event study methodology is therefore an appropriate way to measure the effect (McWilliams & Siegel, 1997). To measure the effect of IPOs on listed rivals, the cumulative abnormal return (CAR) of each rival firm operating in the same 4-digit SIC code industry as the IPO is calculated and later used as the dependent variable in the cross-sectional analysis of the effect. In order to retrieve the CAR of industry rivals an event study following the method described by de Jong & de Goeij (2011) is conducted around the completion date of each IPO and presented in this section. Following Hsu et al. (2010) and McGilvery et al. (2012) a standard market model approach is used. In order to be able to calculate the abnormal returns (ARs), rival firms are matched to an IPO event if they share the same SIC code as the IPO. The abnormal returns (AR i,t ) are calculated by subtracting the expected return or normal return (NR i,t ) of the rival firm s stock from the actual return (R i,t ) of the rival firm s stock. This is represented in the following formula: AR i,t = R i,t NR i,t (1) Before it is possible to calculate the normal returns, a few rival firm parameters require estimation. For this purpose an estimation period of 180 days is taken, from 220 days until 40 days before the IPO s initial trading date, which is defined as the event date t=0. This estimation window is close to the window used by Hsu et al. (2010) [-212,-42] and the window used by Lee et al. (2011) who took the estimation window from 255 to 46 days prior to the event. Using the returns in the estimation window, the parameters α i and β i are obtained for each rival firm. This is incorporated in the following regression model: R it = α i + β i R mt + ε it (2) Where R it is the rival firm s actual stock return at time t and alpha and beta are firm-specific parameters. R mt is the daily return on a stock market index m at time t. For the stock market index the equally weighted market return without dividend on the S&P 500 is consulted in this study. With the IPO s first trading date as t=0, the initial event period is defined as the period starting 10 days before and ending 10 days after t=0. This is based on Hsu, et al. (2010) who found that until 20 days before the IPO filing or IPO completion there are about no significant deviations between the 17

19 mean cumulative abnormal returns, but that from the period starting 10 days before the IPO the CARs change because there is uncertainty about whether or not an IPO will be completed. It is therefore chosen to start the event window 10 days before the event date. Contrarily Ahkigbe et al. (2003) did not start the event window before the event date but only from the event day on. This could be a reason that they did not find a significant industry effect. The initial event period is denoted as [-10, 10]. To provide insight in the CARs within a shorter time period, also the event windows [-5, 5] and [-1,1] are added in the univariate analysis presented in section With the rival firm specific parameters α i and β i estimated in equation 2, the normal returns can now be calculated. This is done by filling in the OLS estimates of the regression coefficients, and the returns on the stock market index over the event window in the following equation: NR it = α i + β i R mt (3) With the results of equation (2) and (3) the abnormal return in equation (1) can be calculated for each rival over the event window. The last step in getting the cumulative abnormal returns is the accumulation of all the rival abnormal returns in the event window. This reflects the industry rival abnormal reaction across several periods. How this is done is represented in the following formula: CAR T2 T1 = T2 t=t1 AR i,t (4) Independent variables The level of homogeneity in an industry Although the degree of homogeneity in an industry is not tested in relation to IPO completion yet, it has been researched by several others. Dooley, Fowler & Miller (1996) focused for instance on the relation between homogeneity or heterogeneity in an industry and the average profitability of that industry. They hypothesized that high levels of strategic homogeneity are associated with high levels of industry profitability and found support for it in their research. Kohers (1999) did include the degree of homogeneity in an industry in an event study concerning dividend omission and initiation announcements, but did not find a significant effect for it on industry rivals. However, as explained in paragraph 2.4, in this thesis the degree of homogeneity is also a variable of interest. If there indeed exists an effect caused by the degree of industry homogeneity, it is expected to be positively related 18

20 with the cumulative abnormal return of industry rivals. The industry degree of homogeneity will be included as a dummy variable which equals 1 for industries with a high degree of homogeneity and 0 if otherwise. Following Kohers (1999), who only marked an industry as homogeneous when it was distinctly similar in terms of the product/services or the general operating environment, the agriculture production and services industry (0109 < SIC < 0784), the metal mining, oil and gas extraction industry (1009 < SIC <1390) and the electric and gas utilities industry (4909 < SIC < 4992) are recognized as homogeneous. The time lag between the announcement and the completion date of an IPO To test for the third hypothesis the number of days between the announcement date and the completion date of an IPO is incorporated in the regression model. This can be easily calculated by subtracting the announcement date from the completion date, which are both available from Zephyr Control variables In order to find the actual impact of the independent variables on the cumulative abnormal return of industry rivals, control variables are added to the regression. Size of the IPO firm The size of the IPO is included as the natural log of the deal value as downloaded from Zephyr. As explained in section 3.1 only IPO deals for which the deal value is known are included. This is somewhat different from Akhigbe et al. (2003) since they included the natural log of the market value on the first trading day, calculated as the product of the number of common shares outstanding and the closing price on the first trading day. They provided evidence that the size of the IPO has a negative impact on the rival portfolio return, since a larger IPO can be a threat to the industry faster. Lee et al. (2011) did also include IPO size as a control variable and showed statistical significance for IPO size at the 5% level 7. Because Hsu et al. (2010) indicated that it is likely that Akhigbe et al. (2003) did not find a significant valuation effect because they included all IPOs, even the very small ones, for this study is chosen to exclude IPOs with a deal value below $50,000,000 from the sample. Age of the IPO firm 7 Lee et al. (2011) focused with their research on one particular growing industry: the computer related service industry. 19

21 Another statistic that might provide some interesting insight is the age of the IPO firm. Lee et al. (2011) did include age of the IPO firm as a control variable, however they did not show statistical significance for it. Following Akhigbe et al. (2013) and McGilvery et al. (2012), that did not include firm age as a control variable, it is chosen not to include IPO firm s age in this study. Exchange of the IPO firm Because the New York Stock Exchange (NYSE) and the American Stock Exchange (AMEX), which is an element of the NYSE, have some differences with the National Association of Securities Dealers Automated Quotations (NASDAQ) a control variable for the stock exchange is included. Kohers (1999) stated that NASDAQ firms face greater information asymmetries and that NYSE-traded firms may have a higher level of institutional ownership, may be more actively traded and more generally well known, thus causing investors of industry-related firms to attribute a high level of significance to information related to this type of firm (p. 142). The dummy variable representing the stock exchange equals 1 when the firm is traded on NASDAQ; 0 otherwise. Industry concentration Another control variable that is included by Akhigbe et al. (2003) as well as Hsu et al. (2010) is a measure for industry concentration, for which the Herfindahl Index was used. Following preceding research, McGilvery et al. (2012) also included the Herfindahl Index as a control variable and showed that when three-day CARs were used, the estimated coefficient on competition is negative and significant. Since statistical significance for the Herfindahl Index as a control variable is proven several times, the index is included in this study as well. The Herfindahl Index for each industry is computed as the squared sum of the fractions of industry market capitalization of rival firms in the industry that are included in the sample. As described earlier the market value is calculated as the product of shares outstanding and the stock price. The higher the index the more concentrated the market. Size of the rival firm The size of each rival is included as the natural log of the market value of the rival firm, where market value is calculated as the number of outstanding shares multiplied by the stock price at the day of the IPO. Hsu et al. (2010) suggested that larger firms are better able to deal with an increase in competition and Lee et al. (2011) did show statistical significance for the size of rivals at the 1% level. Therefore, the size of the rival firm is expected to have a positive impact on the rival firm CAR. 20

22 Leverage of the rival firm Following Hsu et al. (2010) and Lee et al. (2011) the level of financial leverage of the rival firms is furthermore included. Hsu et al. (2010) stated that companies with less leverage endure less from an IPO since they are financially more flexible with their investments, i.e. a low leverage ratio makes the firm less vulnerable to changing market conditions. The leverage ratio is calculated by the end-ofyear total liability divided by the book value of the end-of-year total equity, by which the end of the fiscal year of the IPO is meant. When the fiscal year was not available from Compustat the end of year accounting data is used. Book equity is constructed from Compustat by downloading the book value of stockholders equity, the balance sheet deferred taxes and investment tax credit and the book value of preferred stock for which the redemption value is chosen. Subtracting the second from the first and adding the book value of preferred stock gives the book value of equity. Performance of the rival firm A measure of performance is included as a control variable to exclude that stock returns change due to changing performance rather than due to the IPO. The accounting performance measure of return on assets for firm i at the fiscal-year end of the IPO event e is used, following Lee et al. (2011). The return on assets is calculated by dividing the accounting earnings before interest and taxes by total assets which are both downloaded from Compustat. When the return on assets is higher and therefore the performance of a company, the expected cumulative abnormal return is higher. Therefore a positive relation is predicted. 3.3 Analyses Univariate regression model When the CARs for all industry rivals are calculated, the next necessity in order to be able to conclude about the correctness of hypothesis 1 is the implementation of a test to find out if the CARs are significantly different from zero. Both a parametric and a non-parametric test are performed to test the null hypothesis that the mean CAR equals zero. The most common test used for that purpose is the simple t-test (de Jong & de Goeij, 2011). However, according to de Jong & de Goeij (2011) it is not likely that the variance of the abnormal returns is equal for all firms, because some stocks are riskier than others. Therefore the simple t-test is an inadequate measure. To overcome this problem abnormal returns are standardized before the test statistic is computed. Standardization is a method 21

23 used in many studies but often ascribed to Patell (1976). To calculate the standardized abnormal returns, the time series average of the abnormal returns for each firm i over the event period [T 1,T 2 ] is used to calculate the time series standard deviation for each firm i, as described by de Jong & de Goeij (2011). The abnormal returns as estimated in the event study are divided by the time series standard deviations to generate the standardized abnormal returns. The parametric test that is used in this thesis is the t-test as proposed by Boehmer, Musumeci & Poulsen (1991) because it supplementary corrects for possible event-induced variance in the abnormal returns. Consequently, to test whether the mean CAR equals zero under the null the following test statistic is used: BMP = N CASAR S ~ N(0,1) (5) Where CASAR is the cumulative average of the standardized abnormal returns and S* represents CASAR s cross sectional standard deviation. Since the sample size used in this thesis is abundant, the quantiles of the normal distribution are used as critical values for the t-test. To account also for potential cross-sectional correlation of returns in the industry, portfolios are formed containing the equally-weighted CAR of the industry rivals of each IPO and also tested for significance with the test statistic presented in equation (5). In contrast to the parametric test, the non-parametric test does not require such strict assumptions about return distributions. Non-parametric tests handle outliers and imperfections better. As suggested by McWilliams & Siegel (1997) as a non-parametric test the Wilcoxon signed rank test is performed, which considers both the sign and the magnitude of cumulative abnormal returns. To give more insight in the distribution of the CARs furthermore the percentage of the CARs that are negative is given Multivariate regression model A regression model for the cross-sectional analysis of the valuation effect needs to be composed. For this purpose, a regression with the rival firm s CAR as the dependent variable, as calculated with the event study explained in section 3.2.1, is designed. As explained in the previous section, the variables for which this regression is controlling are IPO size, rival size, industry concentration, rival leverage, the stock exchange and rival performance. To test the second and third hypothesis the key variables, homogeneity in the industry and the lag between the announcement date and the completion date of an IPO are included as independent variables. This is incorporated in the following regression model: 22