Where have all the small-firm IPOs gone?
During the period 1980 to 2010, an average of 380 companies per year went public in the United States. Since 2000, the average has been only 190 initial public offering (IPOs) per year and this drop was particularly severe among small firms. A recent proposed reason for this phenomenon is the Economies of Scope hypothesis. This theory asserts that globalization and technological change has put increasing pressure on small firms and that selling out to a larger organization has therefore become more attractive than operating and going public as an independent firm. This study finds weak empirical evidence for this hypothesis. The findings do not indicate a gradual decline in small-firm IPO volume and the evidence also suggests that the recent introduction of the JOBS Act has significantly affected IPO activity. Lastly, small firms outperform large firms in recent years. Continued efforts are needed to find the exact cause for the drop in small-firm IPO volume over the last years to enhance policy attempts to stimulate the IPO market.
Else Herrebout 10219633
Supervisor: Dhr. dr. J.E. Ligterink
University of Amsterdam MSc Business Economics, Finance Track
Statement of Originality
This document is written by Student Else Herrebout who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Acknowledgements
First and foremost, I would like to thank my supervisor dr. J.E. Ligterink for his guidance and feedback during the process of my thesis writing. His insightful comments always gave me more confidence about my work and motivated me to continuously improve my thesis. Moreover, I would like to thank all my fellow classmates who I have shared this process with on the third floor of Roeterseiland and my friends to whom I could always go when I got stuck. A special thanks to Taco, Nina and Pieter who have made my master Finance a year to remember. Last, but certainly not least, I would like to thank my parents for the support they have always given me.
Table of Contents
1. Introduction ... 4
2. Literature review ... 6
2.1. Initial Public Offerings (IPOs) ... 6
2.2. The IPO decision ... 7
2.3. Cold IPO markets ... 8
2.3.1. United States ... 9
2.3.2. Small firms ... 10
2.4. Economies of Scope hypothesis ... 11
2.4.1. Evidence and criticism Economies of Scope hypothesis ... 12
2.5. Development of hypotheses ... 13 3. Empirical Design ... 15 3.1. Data ... 15 3.2. Method ... 18 4. Empirical analysis ... 20 4.1. Descriptive statistics ... 20 4.2. Results ... 28 4.2.1. Hypothesis 1 ... 28 4.2.3. Hypothesis 2 ... 35 4.2.4. Hypothesis 3 ... 37 4.2.5. Hypothesis 4 ... 38 5. Robustness analysis ... 40
5.1. Correlation independent variables ... 40
5.2. Modifying definition ‘small firm’... 41
5.3. Technology firms ... 42
5.4. Short-term BHARs ... 42
6. Conclusion ... 42
References ... 45
1. Introduction
Since 2000, initial public offering (IPO) volume has been well below historical levels (Dambra et al., 2013). As real GDP more than doubled over the past 35 years, this drop is even more remarkable (Gao et al., 2013). Moreover, the decline in IPO volume has been particularly severe for small firms (Ritter, 2012). While the small-firm IPO decline is most extreme in the US, this phenomenon is occurring in Europe as well (Ritter et al., 2013).
The lack of a dynamic IPO market matters because of two reasons. First of all, it could limit the growth in gross domestic product (GDP) (Weild & Newport, 2013). Secondly, it could have impact on job creation (Weild & Kim, 2009). Generally, companies that go public create many jobs. Kenney et al. (2010) show that in the ten years after going public, the average company increases employment by 60%. Weild & Kim (2009) calculate that the number of jobs lost because of the downfall of US IPOs over the period 2000 to 2011 lies between 11 and 22 million jobs in the US. US congress has expressed their concerns about the IPO market as well and this has resulted in the Jumpstart Our Business Startups (JOBS) Act in 2012. One of the goals of this law was to streamline the IPO process for emerging growth companies (Dambra et al., 2015). Although IPO activity has slightly increased over the last years, we can still question whether this is the result of the JOBS Act (Dambra et al., 2015). Moreover, as long as we don’t know the exact cause(s) of the drop in IPO volume among small firms is, it will be hard to make effective regulatory changes in the future.
One of the main explanations for the drop in small-firm IPO volume for which the JOBS Act has been created, is the “regulatory overreach hypothesis” (Ritter, 2012). This is an overarching designation for all regulatory costs borne by publicly traded firms that are particularly high for small firms. In addition, according to this theory, “the IPO market is broken” (Gao et al., 2013). However, researchers such as Ritter et al. (2013) and Gao et al. (2013) have questioned this explanation, and others that have been proposed, with empirical evidence. Besides this, they have introduced a new explanation which is fundamentally different. Gao et al. (2013, p. 1664) posit “that there is an ongoing change in the economy that has reduced the profitability of small companies, whether public or private”. Therefore, being a small independent company and growing organically by doing an IPO is increasingly an inferior business strategy compared to an alternative strategy of growing big fast, which is often accompanied through mergers and acquisitions (Ritter, 2012, p. 8). They call this explanation the “Economies of Scope hypothesis”. Although there is some first evidence in favor of the Economies of Scope hypothesis, results are not well established in academic research yet. Because of this, our study will empirically test whether the Economies of Scope hypothesis is the main explanation for the drop in IPOs of small firms.
The essence of this paper is especially relevant for government and businesses. If the Economies of Scope hypothesis is correct, policies designed to encourage small companies to go public independently could even harm the economy, rather than boost it (Ritter, 2012, p.21). In contrast, policies should focus on strengthening innovation and the efficient allocation of capital and labor (Ritter, 2012, p.22). Moreover, the decrease in IPO activity (and the increase in M&A activity) would not necessarily be alarming (Ritter, 2012, p. 22).
The purpose of this study is to investigate the following research question: “is the Economies of Scope hypothesis the reason for the decline in small-firm IPO volume?”. The paper will contribute to existing literature through four different aspects. First of all, we will control for more possible explanations in our time-series regressions than Gao et al. (2013) did. Furthermore, we will use a longer time horizon than Gao et al. (2013) used in their paper. Third, this study will be the first to investigate whether the results of the time-series regressions are more severe for highly concentrated industries, by calculating the Herfindahl-Hirschmann index (HHI) for all industries, and for a sample of technology firms. Finally, researchers have not related the introduction of the JOBS Act to the Economies of Scope hypothesis yet. Dambra et al. (2015) have found that the JOBS Act has significantly affected the number of IPOs in the first year after the introduction. In our study we will use their method with a longer time horizon to test whether this result is still applicable. This will enhance our understanding of the Economies of Scope hypothesis, since the theory predicts that such policies will not have a significant effect on small-firm IPO activity.
We will test the research question for a US sample of operating firms over the period of 1985 to 2015 by investigating four hypotheses. The first hypothesis is that the volume of small-firm IPOs has gradually declined over time. We will test this by regressing a quarterly time trend variable on the number of small-firm IPOs. The second hypothesis is that this gradual decline has been particularly severe in industries with a high product market concentration. For testing this hypothesis, we will run the time-series regressions with a sample of small-firm IPOs in highly concentrated industries. The third hypothesis states that the number of small-firm IPOs has not significantly increased after the JOBS Act. We will research this by doing a difference-in-difference regression with an international control sample. The fourth and final hypothesis is developed to support the evidence that small firms underperform on the long-run and is as follows: small public firms underperform large firms on the long run.We will test this hypothesis by computing the 3-year buy-and-hold returns (BHAR) on IPOs of small companies and analyze whether small firms show a lower BHAR than large firms.
The outline of this thesis is as follows. First a literature review will focus on the explanations for fluctuations in IPO volume, the background of the Economies of Scope hypothesis and the JOBS Act.
The section afterwards will create the hypotheses. The next part will describe the data used and will explain the methodology. The results section will discuss the performance outcomes. The conclusion will summarize the findings and the final section also provides recommendation and a discussion.
2. Literature review
2.1. Initial Public Offerings (IPOs)
The recent drop in IPO volume is not a unique phenomenon. Previous literature has documented that IPOs generally tend to come in waves, which you can see in figure 1. Several researchers have studied this well-known pattern and has searched for the underlying causes (see, for example, Lowry (2003) and Pastor & Veronesi (2005)). In our attempt to find out the exact cause underlying the recent lack of a vibrant IPO market, we will give an extensive review of relevant literature in order to obtain an in-depth understanding of all possible explanations.
The first subsection of our literature review will analyze the different benefits and costs of an IPO, since the decision to go public is associated with considering many pros and cons. The subsection after discusses the possible reasons for periods in which IPO markets experience low volume, which are called cold IPO markets (Brau & Fawcett, 2006). In this subsection, we will first sum up some general explanations for this and thereafter we will describe some more arguments related to the recent drop in IPO volume, particularly for the US and small firms. In the last subsection, we will go in depth with one of the most recent and possible important explanations for the drop in small-firm IPO volume; the Economies of Scope hypothesis.
Figure 1. Number of yearly US IPOs
Source: Compustat database
0 2 00 4 00 6 00 8 00 10 00 To ta l IP O s 1980 1990 2000 2010 2020 Year of IPO
2.2. The IPO decision
Although there is a lot of literature on the motivation of firms to go public, the IPO decision is so complex that no single model can capture all of the relevant costs and benefits (Pagano et al., 1998, p. 36). Conventional wisdom says that going public is simply a stage in the growth of a company (Pagano et al., 1998). However, observed patterns in listing cannot completely be explained by this theory and this suggests that the decision of going public is a choice in which several benefits and costs have to be considered (Pagano et al., 1998). This section will give a broad framework of related benefits and costs associated with the choice of going public.
Academic theory poses four factors that contribute to a decision to go public for a firm (Brau & Fawcett, 2006). The first argument is the possibility of raising funds to finance investments (Pastor & Veronesi, 2005). Mikkelson et al. (1997) report that 64% of the firms going public state in their offering prospectus that the reason for their IPO is to finance capital expenditures. Later we will see that the need of funds can change over time and that this contributes to the fluctuation in IPO volume. Second, an IPO allows insiders to cash out (see, for example, Zingales (1995) and Mello & Parsons (2000)). Furthermore, IPOs facilitate M&A transactions by putting a price on the company (Brau & Fawcett, 2006). Lastly, an IPO can be a strategic move because of several reasons (Brau & Fawcett, 2006). First of all, an IPO expands the ownership base of a firm (Brau & Fawcett, 2006). This results in a more diversified portfolio of investors and less investors with real decision-making power (Chemmanur & Fulghieri, 1999). Moreover, according to Maksimovic and Pichler (2001), an IPO can increase the reputation and/or publicity of the firm. These authors reason that an IPO can create buzz in the business community and that this can create a first-mover advantage in the IPO’s niche as well (Maksimovic & Pichler, 2001). In 2006, Brau and Fawcett surveyed 336 Chief Financial Officers (CFOs) to compare practice to theory in the area of IPO motivation. With their survey they found that the benefit related to facilitating takeover activity is the most important motivation for firms to go public (Brau & Fawcett, 2006). Also, moderate support among CFOs is found for the arguments that an IPO functions as a tool to cash out and that it increases the ownership base and that it increases the firm’s reputation.
In contrast, many firms choose to remain private (Brau & Fawcett, 2006). Brau & Fawcett (2006) also question CFOs on this aspect and receive three explanations. They find that losing decision-making control is the most important reasons for firms to maintain private (Brau & Fawcet, 2006). Secondly, companies see the dilution of ownership as a disadvantage (Brau & Fawcett, 2006). Lastly, poor market and industry conditions play a crucial role in their decision of not going public (Brau & Fawcett, 2006). Besides the reasons that Brau & Fawcett (2006) received, Pagano et al.
(1998) explain that there are also considerable direct costs, such as an underwriter, external auditor and legal and financial reporting advisor fees, which can be a disadvantage (PwC’s Deals practice, 2012). Since many of these expenses do not increase proportionally with the size of the IPO, they weigh relatively more on small companies (Pagano et al., 1998). Therefore, these costs can convince firms, and particularly small firms, to remain private.
2.3. Cold IPO markets
As we have mentioned before, research has shown that IPOs tend to come in waves, characterized by periods of hot and cold markets (Brau & Fawcett, 2006, p.400). According to the low number of IPOs over the past years, we can argue that we currently experience a cold IPO market. Researchers such as Lowry (2003), Pastor & Veronesi (2005), Chemmanur (2009) and Baker et al. (2000) have analyzed the different reasons for fluctuations in IPO volume and why cold IPO markets can arise. Basically, there are six general explanations for cold IPO markets. As these are possibly related to the recent drop in IPO volume, we will take all of these explanations into account in our study.
First of all, as we have already described, a firm’s capital demand is one of the reasons to go public. Therefore, when the demand for funds is low, firms will be less likely to go public (Lowry, 2003). This demand for funds is related to economic conditions. When economic conditions are low and economic growth is low, companies tend to have less demand for capital (Lowry, 2003). This low demand translates into less firms seeking financing and less firms going public.
Furthermore, private companies seek this financing through a variety of sources like venture capital funds, angel investors, corporate investors, crowdfunding and bank loans (Smith & Smith, 2000). Since there are so many options, companies don’t necessarily have to raise funds with an IPO when their capital demand increases. A firm will only choose for public equity with an IPO when this form of financing provides the greatest net benefits (Lowry, 2003). The pecking order theory predicts that private companies are more likely to choose for private financing when this is (easily) available (Myers and Majluf, 1984). Consequently, Bharath et al. (2010) show that firms that have much access to private capital are more likely to stay private or go private after being public. That is, when firms have easy access to private financing there is less need to go public to raise funds. When market conditions worsen and stock prices decline, raising capital in the public market will be harder compared to private financing and companies will be less likely to go public (Pastor & Veronesi, 2005).
Third, undervaluation of stocks can be an explanatory factor for a cold IPO market. In periods in which stocks are mispriced, firms know that they will lose proceeds by issuing new shares (Pagano et al., 1998). Pagano et al. (1998) use market-to-book ratio as a proxy for misevaluation. They find that when industry market-to-book ratio is low, firms are less likely to go public (Pagano et
al., 1998). As a result of this, part of the high volume of IPOs in the late 1990s could be explained by the high market valuations on technology stock (Gao et al., 2013).
Furthermore, Chammanur et al. (2009) show that firms in industries and periods characterized by high information asymmetry between firm insiders and outsiders have less chance to go public. Firm insiders, managers, have an incentive to issue equity when the company is overvalued (Lowry, 2003). Therefore, outsiders, the investors, lower their estimate of the firm value when a firm announces an equity offering. On average, this mechanism makes sure that firms will go public when prices are correctly priced (Lowry, 2003). In conclusion, when information asymmetry is very high, firms will postpone an IPO and will find it optimal to obtain alternative types of financing (Lowry, 2003). Also, investor sentiment may drive valuations downward, to which firms respond by postponing an IPO (Baker et al., 2000). Therefore, this is the fifth explanation for cold IPO markets. It has empirically been shown that there is a relation between the level of investor optimism and the IPO volume (Lowry, 2003). When investors are pessimistic, they are less willing to pay for firms than they are worth. During these times, it is unfavorable to go public for firms as the costs are especially high (Lowry, 2003).
Lastly, large costs related to information disclosure can be particularly high for pioneering firms in industries in which innovation is very important (Maksimovic et al., 2001). When disclosing information, pioneering firms face the risk that firms copy their technology, which reduces their competitive positioning in the market (Maksimovic et al., 2001). Therefore, cold IPO markets can arise in period in which these costs are especially high.
2.3.1. United States
Besides general explanations for a lack of a dynamic IPO market that could explain the recent drop, there are three arguments why especially the US is losing in number of IPOs. Weild and Kim (2009) blame the “Casino Capitalism” for this. Over the last years, not only commission and trading costs have gradually decreased but also new products and venues have arisen. These developments accommodate day traders and high-frequency traders and venues such as black pools make the market less transparent (Weild & Kim, 2009). According to Weild and Kim (2009), the understanding that higher transaction costs actually subsidize services that support investors gets forgotten and lower commissions are therefore harming investors (Weild & Kim, 2009). Moreover, in this “Casino Capitalism” firms find themselves in a market environment with a lack of research support, greater systematic risk and volatility and structural obstacles that block them from going public (Weild & Kim, 2009).
Second, the US has high direct and indirect costs in general compared to other countries (Ritter, 2012). For example, Abrahamson et al. (2011) noted that all moderate-size IPOs in the US pay
7% underwriting fees, whereas in Europe the fees are typically around 4%. It is possible that these high costs in the US have pushed firms to go public in other markets (Ritter, 2012).
Finally, whereas the US was the biggest financial market before, other markets are gaining trust as well and modern technology facilitates going public abroad (Zingales, 2007).
2.3.2. Small firms
The recent drop in IPO volume is particularly severe among small firms, which figure 2 shows. This phenomenon has been noticed by several researchers and some arguments have been proposed by different researchers. First of all, a series of regulatory changes, with the Sarbanes-Oxley (SOX) Act as the most important, has been blamed (Ritter, 2012). SOX was signed into law in 2002 after many corporate scandals. The law’s main goal was to improve the quality of financial reporting and to increase investor confidence (Lliev, 2010). Section 404 of the law imposed additional compliance costs on public firms. As a percentage of revenue, these costs have been excessively high for small firms (Gao et al., 2013). Because of this, small firms were exempted from many of the requirements in 2007 (Ritter, 2012). Furthermore, a decline in analyst coverage could have contributed to the drop in IPO volume among small firms. In general, analyst coverage results in greater awareness of a stock’s existence (Womack, 1996). This results in increasing demand among investors and a higher stock price (Ritter, 2012). Explanations for why this analyst coverage for small firms has declined over time point at the drop in the bid-ask spreads that began in 1994 in the US. This drop had a negative effect on the incentives for analysts to cover small firms (Gao et al., 2013). The decline in spreads has also been facilitated by technological changes, the U.S. Securities and Exchange Commission’s (S.E.C.) Order Handling Rules in 1997, Regulation ATS in 1998, the move to decimalization in 2001, and Regulation NMS in 2005 (Ritter, 2012). The combination of the introduction of SOX, regulatory changes and analyst coverage as an explanation for the low small-firm IPO volume has been called the “regulatory overreach” hypothesis (Ritter et al., 2013).
Figure 2. Number of yearly US total IPOs and small-firm IPOs
Source: Compustat Database
2.4. Economies of Scope hypothesis
A totally different argument for the drop in small-firm IPO volume is posit in the paper “Where have all the IPOs gone?” (Gao et al., 2013). Xiaohui Gao, Zhongyan Zhu and Jay Ritter (2013) argue that there has been a gradual and structural decline in small-firm IPOs over the last decades. They hypothesize that the increasing importance of economies of scale and scope is the explanation for this. Because of this ongoing change, companies find it more convenient to get big fast through mergers and acquisition (M&As) rather than going public and remaining independent (Ritter et al., 2013, p.2). Moreover, in contrast to other theories, this theory does not believe that the decline in IPO activity is related to the firm’s choice between being a private or public company (Gao et al., 2013). This theory views the decline as a consequence of a change in attractiveness of being a big rather than a small company (Ritter et al., 2013, p.2). The authors call this reasoning the “Economies of Scope hypothesis”.
The reasons for this change towards increasing importance of economies of scale and scope are related to changes in technology and globalization (Gao et al., 2013). We will outline three of these changes and explain what the relationship is with economies of scale and scope and therefore the size of a firm. First of all, according to research of Sood and Tellis (2005), the speed of technological change has increased over time. Products and processes are becoming increasingly complex and often have a multi-technology nature (Huissinger, 2010). For companies it is necessary to adapt to these changes quickly but this costs time and money (Gao et al., 2013). Therefore, scaling the organization and obtaining economies of scale and scope can be an option to overcome these costs. Secondly, technology, like the Internet, has made comparison shopping easier for consumers (Goldmanis et al., 2010). Because of these reduced search costs, a “winner take all” tendency has arisen (Gao et al., 2013, p.1671). Therefore, only larger firms will be able to compete in this new
0 2 00 4 00 6 00 8 00 10 00 1980 1990 2000 2010 2020 ipoyear
Total small-firm IPOs IPOsTotal
T ot al n um be r o f IP O s
market. Finally, globalization and the easing of market restrictions has provided firms to operate more easily in foreign markets and has created large growth opportunities (Julien et al., 1994). Because of this, competition has become fiercer and opportunities will be lost when they are not quickly seized by (big) companies (Goa et al., 2013).
2.4.1. Evidence and criticism Economies of Scope hypothesis
In their paper, Gao et al. (2013) find some first evidence in favor of the Economies of Scope hypothesis. Also, they show that some of the arguments proposed before cannot explain the drop in IPO volume among small firms. First we will show which arguments are tackled and then we will continue on the first proof of the Economies of Scope hypothesis. We will end with the first argument against the theory.
To begin with, Gao et al. (2013) show that SOX is not a significant explanation (Gao et al., 2013). They show that the effect on the profitability for small firms of paying the compliance cost is limited (Gao et al., 2013). Also, they show that the drop in analyst coverage has not been as large as predicted by literature (Gao et al., 2013). Furthermore, when direct and indirect costs are higher compared to other countries, we would expect firms to have listed in foreign countries. However, the researchers show that foreign listing have not significantly increased during the sample period (Gao et al., 2013 and Ritter, 2012).
Furthermore, Goa et al. (2013) find that it has become increasingly unprofitable to be a small firm; the percentage of small public firms with negative EPS has increased over time. This could be the first evidence in favor of the Economies of Scope hypothesis. Besides the findings of Goa et al. (2013), Bayar and Chemmanur (2011) show that the percentage of firms being acquired within three years after going public has increased. Gao et al. (2013) and Ritter (2012) find that, in line with this result, there is an uptrend in the frequency of acquisitions of firms that have recently gone public. This suggests that growing organically is getting less popular and growing fast by doing acquisitions is taking the overhand. Furthermore, they show that small firms underperform large firms in the long run (Ritter, 2012 and Gao et al., 2013). The final test of Gao et al. (2013) is a time-series regression and with this test they show that there has actually been a gradual decline in IPO volume among small firms, which is in line with predictions of the Economies of Scope hypothesis.
Although this is a lot of evidence in favor of the Economies of Scope hypothesis, the increase in IPO activity since 2012 among small firms, see figure 3, contradicts the theory. Dambra et al. (2015) are the first researchers to test whether this is the result of the Jumpstart Our Business Startups (JOBS) Act, enacted on April 5th, 2012. One of the goals of this law was to streamline the IPO process for emerging growth companies, primarily by easing various securities regulations (Ritter, 2012). Dambra et al. (2015) find that in the period April 2012 to March 2014 the number of IPOs
have significantly increased, possibly as a consequence of the JOBS Act. If this is true on the long run, the arguments in favor of the Economies of Scope hypothesis will get less strong. That is, the Economies of Scope hypothesis predicts that policy changes to make IPOs more attractive will not have a significant effect on the number of small-firm IPOs since growing organically will remain an inferior strategy (Gao et al., 2013).
Figure 3. Number of yearly US IPOs between 2000 and 2014
Source: Compustat database
2.5. Development of hypotheses
In the paragraphs before we have discussed all possible explanations for a drop in IPO volume that could be of explanation for the recent drop in (small-firm) IPO volume. However, it is still not clear yet which factor is mostly responsible for the declining IPO volume of small firms over the last decades. The Economies of Scope hypothesis recently proposed by Gao et al. (2013) seems to be a possible good explanation, but more research and evidence for this theory is needed. Therefore, we will continue their work and test whether the Economies of Scope hypothesis is the main explanation underlying the drop in small-firm IPO volume. To test this, we will first develop four hypotheses. According to the Economies of Scope hypothesis there has been a gradual increase in importance of economies of scale and scope over the last 35 years (Gao et al., 2013). Therefore, this theory suggests a gradual decrease in the number of small-firm IPOs, rather than sudden drops after regulatory changes (Gao et al., 2013). This results in our first hypothesis:
(H1) The volume in small-firm IPOs has gradually declined over time
Secondly, in the paper of Gao et al. (2013, p. 1675) the authors describe that the ultimate analysis to test the Economies of Scope hypothesis would be to research the industries in which economies of scale and scope are the most important. They argue that in industries to which this applies, there
0 1 00 2 00 3 00 4 00 5 00 To ta l IP O s 2000 2005 2010 2015 Year of IPO
will be more M&A activity and less small-firm IPOs (Gao et al., 2013). Since it is it is hard to define and measure in what industries economies and scale and scope are big, we cannot test this in our study. However, a proxy for this could be concentration in the market. That is, when there are only a few big players active in a product market this could indicate that it is hard to enter this market as a small firm and that it is likely that economies of scale and scope are important. Therefore, we will test whether the gradual decline in small-firm IPO has been particularly severe in industries with a high concentration in the market.
(H2) The gradual decline in small-firm IPO volume has been particularly severe in industries with a high product market concentration
Since going public autonomously as a small company has become an inferior strategy according to the Economies of Scope hypothesis, policy changes to reduce the costs of going public or to streamline the IPO process will not have a significant effect on the number of small-firm IPOs (Gao et al., 2013). As the JOBS Act has the goal to encourage small firms to go public, we will continue the study of Dambra et al. (2015) by testing whether the JOBS Act has actually effected the IPO activity among small firms significantly on the long run. According to the Economies of Scope hypothesis the JOBS Act would only have had a very limited positive effect on the number of IPOs. Therefore, our third hypothesis is the following:
(H3) The number of small-firm IPOs has not significantly increased after the JOBS Act
Although the theory of Economies of Scope applies to other countries as well, we will conduct our research only on the US, since this was the only country where the JOBS Act has been introduced (Dambra et al., 2015).
Finally, Gao et al. (2013) showed that small public firms underperform their style-matched larger counterparts. In the literature described before, we have seen that economic conditions partly contribute to fluctuation in IPO volume. Furthermore, Goa et al. (2013) used data until 2011 for their study. Ritter et al. (2012) argue that it is possible that part of the results of Gao et al. (2013) can be attributed to the bear market years at the final years of their data sample. That is, both GDP and market returns decreased during this period. In figure 4 we can see that real GDP even shrinked between 2008 and 2010 during the financial crisis. Also, figure 5 shows that there were negative market returns in most quarters between 2007 and 2011. Therefore, we will test whether this underperformance of small firms is still appearing if we include the bull-period of 2012-2015. According to the Economies of Scope hypothesis, we would expect that smaller firms still
underperform larger firms as it is getting harder to survive as a small firm. Because of this, our fourth hypothesis is as follows:
(H4) Small public firms underperform large firms on the long run Figure 4. Real GDP
Source: FRED Economic Data
Figure 5. Stock market return
Source: CRSP database
3. Empirical Design
3.1. Data
The initial sample that we use in this research consists of all IPOs of operating firms between 1980 and 2015. The analysis starts in 1980 because Gao et al. (2013) propose this has been the starting point of the change in speed in technological change and decline in small-firm IPO volume. Also, we use a sample period up until 2015 to include the introduction of the JOBS Act. In this study we focus on operating firms because these can potentially create jobs (Gao et al., 2013). Therefore, when we obtain IPO-data from Compustat, we exclude closed-end funds, real estate investment trusts (REITs), special purpose acquisition companies (SPACs) and bank, savings and loan (S&L) IPOs. Furthermore, for hypothesis 3 we use an international sample of IPOs which we will gather from Thomson One. We use a different database for this since Thomson One has more data about international IPOs, whereas Compustat has more data about revenues in the IPO-year available for US firms. We define a small firm as a firm with USD 50 million in revenues in the year of their IPO but we will adjust this measure several times in our robustness section to make sure our results do not depend on this proxy. Since researchers generally agree that the number of IPOs per year should grow at around the same pace as a country’s economy, we will normalize the number of IPOs with real GDP (Weild & Newport, 2013). Quarterly real GDP will be gathered from FRED Economic Data. Moreover, for our
12 13 14 15 16 17 R ea l G D P 2000q1 2005q1 2010q1 2015q1 Year -. 1 -. 05 0 .05 Q u a rt e rl y Ma rke t R e tu rn 2005q3 2008q1 2010q3 2013q1 2015q3 Year
international sample, we aggregate IPO activity to the nation-quarter level in the following way: we divide the number of IPOs by the number of publicly listed firms. We obtain aggregate public firm data from the World Bank’s World Development Indicators database (Dambra, 2015). Furthermore, in hypothesis 1 we will use a quarterly time trend that equals 0 for the first quarter of 1980 and increases by 0.01 for each quarter onward until the fourth quarter of 2015 (Gao et al., 2013).
In hypothesis 2, we will only use IPOs of companies in highly concentrated or highly dispersed industries. The main measure of concentration in the product market is the Herfindahl-Hirschman index (HHI), which is a commonly used measure in the empirical literature (Tirole, 1988, pp. 221-223). In the bachelor thesis of Herrebout (2014) HHIs are computed for all industries over the years 2005 to 2012. She uses the 48 industry classifications of Fama and French (2013) and a global sample of firms to calculate the sum of squared market shares based on firms’ sales. We will use the average of the HHI over these seven years and select the three industries with the highest HHI and those with the lowest HHI (see Appendix B). That is, the industries with the highest HHI represent the most concentrated product markets, and those with the lowest HHI the most dispersed.
In our regressions for testing hypothesis 1 and 2 we control for possible other explanations for the fluctuation in (small-firm) IPO volume. As other researchers proposed that the SOX regulation is one of the main reasons for the drop in US small-firm IPO volume, we will include a dummy variable that equals 1 from the third quarter of 2002 to the fourth quarter of 2007, and 0 otherwise (Gao et al., 2013). Also, it is possible that the introduction of the JOBS Act has had influence on the increase in IPO volume over the recent years, as we have discussed before (Dambra et al., 2015). Therefore, we control for the JOBS Act in our regressions by creating a dummy variable that equals 1 from the introduction of the JOBS Act, the second quarter of 2012, to the last quarter of 2015, and 0 otherwise (Dambra et al., 2015).
Moreover, we will use two different variables to control for differences in market conditions. First of all, we will use past market returns. We define this variable as the 6-month value weighted return on the NYSE, AMEX, NASDAQ and ARCA excluding dividends in quarters t-2 and t-1 in decimals from CRSP (Gao et al., 2013).
Secondly, following Gao et al. (2013), we will take the initial returns on IPOs of small firms into account as a proxy for market conditions. We will obtain monthly data for this on the website of Jay R. Ritter1 about IPO data. Here, day returns are defined as the difference between the first-day closing price and the offer price divided by the offer price (Gao et al., 2013). We will calculate average quarterly initial IPO returns for small firms in quarter t-1.
Third, as a proxy for market conditions of small firms in particular, we will calculate the yearly percentage of small public firms with negative earnings per share (EPS) in quarter t. For public firms, we will use 150 million USD as a cutoff for small firms. We will construct the variable EPS by subtracting preferred dividend from net income and dividing this by the number of shares outstanding.
Next to market conditions, we have described that market inefficiency is also an explanatory variable for fluctuation in IPO volume. Pastor and Veonessi (2005) explain that market-to-book ratio is a good proxy for this. Because of this, we will construct a variable for market inefficiency by taking the log of the sum of the market value of small firms and divide these by the sum of book value of small firms (Gao et al., 2013). Market value will be calculated by taking the close price at the end of a quarter multiplied by the number of shares outstanding. Total assets minus total liabilities is the definition for book value of equity in a quarter. We will construct this measure at the end of quarter t-2 (Gao et al., 2013).
To control for differences in aggregate demand for capital over a period, we will use real GDP growth as this is the best proxy according to Lowry (2003). This variable will be the percentage of growth in real GDP from quarter t to quarter t+3 in percentages obtained from FRED Economic Data.
Also, we will use a variable to measure the easiness of access to private financing among private firms in a period. We will measure the access to credit by calculating the growth in total quarterly credit raised by private companies from quarter t-2 to t-1. We will gather data about total domestic credit raised by private companies from the World bank database.
Another important explanation for IPO fluctuation that needs to be controlled for is investor sentiment. Like Gao et al. (2013), we will create two variables for this. Following Lowry (2003), future market returns are a first good proxy for investor sentiment. We will construct this variable by gathering data from CRSP and obtaining the monthly value weighted market returns. With this data we then calculate the market returns from quarter t+1 to t+4. The second variable will be the discount on closed-end funds. According to Lee et al. (1991), if investor sentiment is higher, investors pay relatively more for closed-end funds and the discount is smaller. Data for this variable can be found on the website of Jeffrey Wurgler of Stern NYU2.
Furthermore, we will follow Lowry (2003) with constructing a proxy for information asymmetry. We will use the dispersion of analyst forecasts of public firms’ earnings as a measure for this. Moreover, the standard deviation of analyst forecasts for EPS represents the uncertainty in the market (Lowry, 2003). To construct this control variable we will obtain the standard deviation of EPS
forecast among analysts from IBES and we will calculate the average standard deviation across companies in quarter t.
As a final control variable in our regressions for hypothesis 1 and 2, we will include a quarter 1 dummy, which equals one if it is the first calendar quarter and zero otherwise. Lowry (2003) describes that IPO volume is highly seasonal and that there are significantly fewer IPOs in the first quarter of the year. A possible explanation for this is Wall Street’s practice of effectively shutting down between Christmas and New Year’s, which results in a lower number of new IPO registrations (Lowry, 2003).
For hypothesis 3 we will follow Dambra et al. (2015) in their approach of the controls in the difference-in-difference regressions. Besides the dummy for the JOBS Acts that we have already defined, we will include variables for past market return and growth in GDP for each country in our sample. We will obtain the total market return index of a country from Datastream. In our regressions, we will use the return in quarter t-1 per country. The annual GDP growth per country will be gathered from the World Bank. Finally, like Dambra et al. (2015), we will include the log of the average US market-to-book ratio of small firms using the same approach we have described before.
To test hypothesis 4, we will use the same sample of IPOs as for hypothesis 1. We will use monthly stock prices of the firms of these IPOs from the CRSP/Compustat Merge database. Only firms with the stock price of their IPO-month available will end up in our sample. Since we will compare and analyze the average buy-and-hold return (BHARi) for small and large companies, we first need the 3-year buy-and-hold return of a company, which is calculated as follows: (Price end period – Price IPO) / (IPO price). Furthermore, monthly value-weighted index excluding dividends are obtained from CRSP and returns over a specific period are calculated as follows: (1+return month of IPO) x (1+return month 2) x (1+return month 3)…. x (1+return end period).
3.2. Method
With the obtained data, we will test the first hypothesis with a time-series regression following Gao et al. (2013). In this regression, we use quarterly IPO activity as the dependent variable, scaled by real GDP. To compare and relate the number of small-firm IPOs to large firms and total number of IPOs respectively we will use three measures for IPO activity as a dependent variable: (1) Total IPOs / Real GDP, (2) Small-firm IPOs / Real GDP, and (3) Small-firm IPOs / total IPOs. Furthermore, we will use a quarterly time trend variable to capture the impact of a gradual change in the importance of economies of scale and scope on scaled IPO volume (Gao et al., 2013). That is, a negative coefficient on the time trend would suggest that IPO experienced a gradual decline over the last decades. Moreover, we will incorporate controls for all possible explanations of lower or higher IPO volume
that we have discussed in the literature review. For hypothesis 2 we will use the same method, but then on a different sample of IPOs. As a result, the regression for our first and second hypothesis is as follows:
(H1 & H2) IPO Volumet = α + β1Time trend + β2 SOX dummy + β3 Analyst Coverage + Β4 JOBS dummy + β5 Real GDP growth + β6 IPO Initial return + β7 % small firms with negative
EPS + β8 Past market return + β9 Log M/B small firms + β10 Credit to private sector + β11 Dispersion in analysts’ forecast + β12 Future market return + β13 Closed-end fund Discount +
β14 Quarter 1 dummy + εt
We will test hypothesis 3 using a difference-in-difference regression. Following Dambra et al. (2015) we will use an international control sample in which we include IPOs from developed countries with the largest stock exchanges. These countries are Australia, Canada, Hong Kong, Japan and the United Kingdom. Furthermore, we will use the same control variables as Dambra et al. (2015), who use some country-specific controls and a US-specific control variable for the market-to-book ratio. If this hypothesis is correct, we expect that the dummy variable US times the after JOBS Act is not significantly different from zero. Taking all of this into account, the regression model for hypothesis 3 looks as follows:
(H2) IPO Volumeit = α + β1 USi ×Post-JOBSt + β2 Country Real GDP growthit + β3 Lag country stock returnit + β4 US Log M/B small firmst + εit
For testing the fourth hypothesis, we will follow Gao et al. (2013). We will measure and compare the post-issue stock return performance of small- and large-company IPOs. Compared to Gao et al. (2013) we will use a longer time horizon, that consists of IPOs between 1980 and 2012. We will measure 3-year buy-and-hold returns with the closing market price on the first month of trading until the earlier of either their 3-year existence on the market or their delisting month. Furthermore,
we will measure the average buy-and-hold abnormal return (BHARi) with respect to the value-weighted market return (Gao et al., 2013). Finally, the formula for the BHARi looks as follows:
𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝑖𝑖,𝑇𝑇 = � �1 + 𝐵𝐵𝑖𝑖,𝑡𝑡� − � �1 + 𝐵𝐵𝑀𝑀,𝑡𝑡� min (𝑇𝑇,𝑑𝑑𝑑𝑑𝑑𝑑𝑖𝑖𝑑𝑑𝑡𝑡)
𝑡𝑡=1 min (𝑇𝑇,𝑑𝑑𝑑𝑑𝑑𝑑𝑖𝑖𝑑𝑑𝑡𝑡)
4. Empirical analysis
4.1. Descriptive statistics
Table 1 shows the total number of firms in the sample. Firms in Compustat for which their IPO date is not available or for which revenues are not available in the year of their IPO are deleted from the sample. After removing the non-operating firms, our sample consists of 9,886 firms. Table 2 provides the yearly number of total IPOs, small-firm IPOs and the percentage of small-firm IPOs of total IPOs. Up until 1984 there have only been a few IPOs. This could partly be attributed to the fact that Compustat had less information about companies than that they have now. Therefore, we will run our regressions only from 1985 onwards as we believe our data is truly reliable from this point. Although the yearly numbers are not completely in line with the numbers Gao et al. (2013) use in their sample, see appendix A. we can see the same trends over the years in our sample. Table 3 represents the average number of total and small-firm IPOs over the periods 1980 to 2000 and 2000 to 2016. For total IPOs the average is 378 and 189 respectively for these two periods. For firms with total revenues below 50 mln USD in the first year of their IPO, these numbers are 208 and 61. These big differences can also be seen in figure 2, where we can see a downward trend in both total IPOs and small-firm IPOs. On top of that, figure 4 shows that this trend has been particularly severe among small firms, since the percentage of small firms of total IPOs has decreased as well.
For our international sample we present the same figures. For this sample we use Thomson One and therefore the number of total US IPOs differs in this sample, which you can see in Table 5. Since Thomson One has not all revenues for companies in their IPO-year available, we lose some data here. However, except for Canada, we can still include around 1/3 of the total IPOs in our sample. The total number of IPOs in Australia, United Kingdom, Canada, Japan and Hong Kong are relatively low compared to the US. Although these differences seem really big, Zingales (2007) has shown before that the US attracted 48% of all the global IPOs in the 1990s. Therefore, it is not strange that the US has relatively had so many IPOs. Also, from table 6 it is clear that data for countries other than the US is probably not complete up until 1995, since there are zero IPOs in some years. Therefore, we will only use data only 1995 in hypothesis 3. As we have stated before, Dambra et al. (2015) has shown that these five countries have experienced the same trend in IPOs over the years and because of this, it is a good sample to control for. To confirm this, table 7 shows us that that the average of yearly IPOs in all countries has decreased after 2006. Moreover, figure 7 shows the IPO trend over the years. Although inspection of this figure shows us that the US has experienced a steeper drop in IPO volume, there is a similar trend across all other five countries.
Table 1. Total number of IPOs in US sample
Sample Number of IPOs
Total firms Compustat 36.125
Minus IPO date not available 12.032
Minus companies without revenues available in IPO year 10.443
Minus non-operating firms 9.886
Table 2. Total yearly US IPOs (1980-2015)
The sample of 10,440 initial public offerings obtained from Compustat excludes closed-end funds, REITS, SPACs and S&L IPOs. Small-firm IPOs are categorized on the basis of the 12-month sales in the year of the IPO. Firms with less than 50 mln USD in the year of the IPO are defined as a small firm.
Period Total nr of IPOs Small-firm IPOs % Small-firms of total
1980 1 0 0% 1981 2 2 100% 1982 1 0 0% 1983 9 6 67% 1984 13 3 23% 1985 34 13 38% 1986 307 199 65% 1987 612 381 62% 1988 232 115 50% 1989 116 68 59% 1990 144 87 60% 1991 368 205 56% 1992 456 250 55% 1993 660 333 51% 1994 586 322 55% 1995 660 393 60% 1996 966 556 58% 1997 787 401 51% 1998 694 317 46% 1999 759 429 57% 2000 526 289 55% 2001 91 38 42% 2002 78 16 21% 2003 60 19 32% 2004 243 85 35% 2005 251 65 26% 2006 239 64 27% 2007 330 67 20% 2008 86 31 36% 2009 85 14 17% 2010 188 48 26% 2011 143 39 27% 2012 142 26 18% 2013 205 63 31% 2014 232 72 31% 2015 134 42 31% Total 10440 5058 48%
Table 3. Average of yearly US IPOs
Period Average nr yearly IPOs Average nr yearly small-firm IPOs Percentage small-firm IPOs of total IPOs
1980-2000 378 208 55%
2000-2015 189 61 32%
1980-2015 290 141 48%
Table 4. Average of yearly US IPOssample of Gao, X., Ritter, J. R., & Zhu, Z. (2013). Where have all the IPOs
gone? Journal of Financial and Quantitative Analysis, 48(06), 1663-1692.
Period Average nr yearly IPOs Average nr yearly small-firm IPOs Percentage small-firm IPOs of total IPOs
1980-2000 310 165 53%
2000-2012 99 28 28%
1980-2012 234 116 50%
Figure 6. Percentage small-firm IPOs of total IPOs US
Table 5. Total number of IPOs in international sample
US Australia UK Canada Japan Hong Kong
Total firms with IPO-date Thomson
One 17.921 3.016 3.424 6.293 3.076 1.925
Minus companies without revenues
available in IPO year 7.439 983 1.057 286 1.492 788
Minus non-operating firms 4.466 692 661 145 404 244
.2 .3 .4 .5 .6 .7 % Sma ll fi rms o f t o ta l f irms 1980 1990 2000 2010 2020 Year
Table 6. Total yearly IPOs international sample (1991-2015)
The sample of 9,092 initial public offerings obtained from Thomson One excludes closed-end funds, REITS, SPACs and S&L IPOs.
Year United States Australia Kingdom United Canada Japan Hong Kong
1991 254 2 0 4 8 0 1992 375 2 3 5 9 0 1993 432 2 4 7 40 2 1994 366 2 10 5 67 0 1995 367 4 5 11 68 0 1996 592 10 42 18 64 1 1997 414 11 67 16 47 17 1998 89 7 61 11 73 0 1999 217 25 34 15 116 8 2000 112 57 130 10 135 18 2001 51 27 70 5 148 39 2002 53 40 67 14 128 52 2003 44 66 23 9 99 41 2004 96 100 88 20 71 48 2005 78 98 82 17 67 51 2006 81 100 71 12 66 53 2007 82 138 65 23 63 74 2008 9 21 10 3 24 31 2009 40 20 7 2 8 52 2010 99 44 21 18 12 60 2011 103 33 20 10 20 57 2012 127 16 27 7 20 30 2013 159 35 45 10 33 48 2014 181 86 73 9 46 54 2015 103 35 27 10 38 48 Total 4524 981 1052 271 1470 784
Table 7. Annual average number of IPOs per country
Period United States Australia Kingdom United Canada Japan Hong Kong
1995-2004 581 103 188 252 160 72
Figure 7. Yearly IPOs per country
Table 8 provides summary statistics for all variables that could be of explanatory power for the drop in small-firm IPO volume. First of all, real GDP in billions of chained 2009 USD has been increasing since 1980, with only a slight dip around 2008, which you can see in figure 4. Secondly, market return and the initial return of IPOs are the variables related to market conditions. Looking at quarterly market returns, the percentage return lies between -8% and +7% and the average is 1%. In figure 5 we can also see that the values are widely spread and that there is no clear pattern over the years. Initial returns of IPOs are considered as the second proxy for market conditions. There is a big difference between the minimum of -5.5% and maximum of 97%. The minimum occurred during the financial crisis around 2008 and the maximum in the DOT-com bubble around 2000, which you can see the graph in appendix C1. Over the past 35 years, on average, firms had a positive initial return 17.5% when going public. On average, 43% of the small firms (public companies with revenues below 150 mln USD) had negative earnings per share (EPS). Moreover, the graph in appendix C2 presents an upward moving trend in small firms with a negative EPS. This indicates that it has become less profitable to be a small public firm and this is in line with results shown before by Ritter et al. (2013).
Moving on to the variable which serve as a proxy for access to private financing, the graph in appendix C3 shows the pattern in total domestic credit raised by private companies on top of the summary statistics report in table 8. We can see that the total amount of credit as a percentage of GDP has increased exponentially up until the financial crisis in 2008. Therefore, the minimum and maximum of 89% and 203% respectively are widely diverged.
The discount on closed-end funds is one of the proxies for investor sentiment and this variable shows the lowest discounts between 2001 and 2007, which indicates that investors were
0 2 00 4 00 6 00 8 00 10 00 1995 2000 2005 2010 2015 ipoyear
IPOs US IPOs Australia IPOs United Kingdom IPOs Canada IPOs Japan IPOs HongKong
T ot al n um be r o f IP O s
optimistic at that time and a peak can be seen around the financial crisis of 2008.
Furthermore, the average standard deviation in analysts’ forecast presents the uncertainty in the market. The average of this variable is 0.09 and the trend moves between 0 and 0.15, which you can also see in the graph of appendix C5.
Finally, the graph in appendix C6 shows that analyst coverage has been relatively low between 1996 and 2009. This is in line with the theories that propose that incentives to cover (small) firms have been low after the drop in the bid-ask spread. However, the minimum of 5.3 and maximum of 7.8 analysts that cover a firm are not so widely spread.
Table 8. Descriptive statistics
Descriptive statistics of all explanatory variables over the sample period 1980 to 2015. Small firms are defined as firms with annual sales below $50 million in the first year of their IPO. For seasoned public firms, we define a small firm as firms with annual sales below $150 million. Real GDP is US annual real GDP in trillion USD. Market return is the six-month value weighted return in percentage on the NYSE, AMEX, NASDAQ and ARCA excluding dividends. IPO initial return is the first-day return of small firms in the first year of their IPO in percentages. % small firms with negative EPS is the percentage of small public firms of total public firms that have negative EPS. Market-to-Book small firms is the log of the sum of the market value of small firms divided by the sum of book value of small firms. Credit to private sector is the total credit raised by private companies as percentage of GDP. Dispersion in analysts’ forecast represents the standard deviation of analyst forecasts for EPS. Closed-end fund discount is the discount on closed-end funds in percentages. Analyst coverage represents the number of estimates that cover an initial public offering.
Variable N Mean SD Median Min Max
Real GDP (trln USD) 145 11,36 3,18 11,32 6,38 16,49
Market Return % 148 0,01 0,03 0,01 -0,08 0,07
IPO Initial Return % 148 17,52 17,33 13,08 -5,5 97,07
% small firms with negative EPS 149 0,45 0,18 0,48 0 0,67
Market-to-Book small firms 149 0,59 0,19 0,59 0,05 1,02
Credit to private sector (% of GDP) 149 145,3 38,04 141,63 89,13 206,30 Dispersion in analysts' forecast (SD) 149 0,09 0,04 0,1 0 0,15
Closed-end fund discount (%) 145 7,88 5,44 8,47 -3,48 24,36
Table 9 presents correlations among the variables of interest that will be used in our empirical study. That is, the value of the variable in the quarter or time period that we will use in our regressions is used to check the correlation with the other variables.
First of all, the regulatory overreach hypothesis predicts a negative correlation between quarterly (small-firm) IPOs with SOX and analyst coverage. We can see that this is indeed the case in this sample. However, according to the regulatory overreach hypothesis JOBS should positively correlated with (small-firm) IPO volume, but this appears not to be true. Moreover, when market conditions would partly determine a decision to go public or not, we would expect that the correlation between (small-firm) IPOs and past market return is positive, the correlation between (small-firm) IPOs and percentage IPO return is positive and the correlation between (small-firm) IPOs and the percentage of firms with negative EPS is negative as well. We can see that these predictions are correct in practice. Furthermore, market-to-book ratio is highly correlated with both quarterly small-firm IPO and total IPOs. This in line with theory, which anticipates that when firms are overvalued, there is a higher chance of firms going public. When real GDP increases and therefore capital demand increases, we can see that the number of (small-firm) IPOs increases as well, since the correlation is positive. Also, when the dispersion of analysts’ forecast increases, there is more uncertainty in the market. Therefore, a negative correlation is expected and we can see this is actually the case in table 9. Finally, the Economies of Scope hypothesis believes that there has been a gradual decrease in the number of small-firm IPOs over the last 35 years. In our correlation matrix, we can see that there is indeed a negative correlation between the time trend and the number of IPOs. However, this correlation is not stronger for small firms than for total firms and this is not in line with the theory. Overall, the negative correlation with the time trend and the JOBS Act could both indicate that the Economies of Scope hypothesis is correct.
Finally, many of our independent variables (highly) correlate with each other. The highest correlation exists between our time trend and the percentage of small public firms with negative EPS. These two variables have a significant correlation of 0.81. Although there are some variables not significantly correlated with each other, we have to be cautious with running our regressions, since multicollinearity can arise. Because of this, we will also test our hypotheses with only the variables that are not significantly correlated with each other in our robustness analysis.
Table 9. Correlation matrix
Simple correlations of all dependent and explanatory variables over the sample period 1980 to 2015. Total IPOs per quarter represents quarterly number of IPOs in the US. Total small-firm IPOs represents quarterly number of IPOs of firms with less than 50 mln USD in revenues in the first year of their IPO. Time trend is the quarterly time trend that equals 0 for the first quarter of 1980 and increases by 0.1 for each quarter onward until the fourth quarter of 2015. SOX dummy is a past-Sarbanes-Oxley dummy that equals 1 from the third quarter of 2002 to the fourth quarter of 2007, and 0 otherwise. Analyst coverage in (t- 1) represents the number of estimates that cover an initial public offering in quarter t - 1. Real GDP (%) in [t; t + 3] is the percentage growth in real GDP in from quarter t to t + 3. IPO initial return in (t – 1) is the first-day return of small firms in percentages in quarter t - 1. % Small firms with negative EPS in (t) is the percentage of small public firms of total public firms that have negative EPS in quarter t. Market return [t - 2; t - 1] is the six-month value weighted return in percentage on the NYSE, AMEX, NASDAQ and ARCA excluding dividends in quarters t – 2 and t – 1. Log M/B for small firms in (t – 2 ) is the log of the sum of the market value of small firms divided by the sum of book value of small firms in quarter t – 2. Credit to private sector growth (%) in (t – 1) is the percentage growth in the average total credit raised by private companies in quarter t – 1. Dispersion in analysts’ forecast (SD) in (t) represents the standard deviation of analyst forecasts for EPS in quarter t. Future market return in [t + 1; t + 4] is the value weighted return in percentage on the NYSE, AMEX, NASDAQ and ARCA excluding dividends from quarter t + 1 to t + 4. Closed-end fund discount (%) in (t – 4) is the discount on closed-end funds in percentages in quarter t – 4.
Total IPOs per quarter Total small-firm IPOs per quarter Time
trend SOX dummy
Analyst coverage in (t-1) JOBS dummy Real GDP growth (%) in [t; t + 3] IPO initial return in (t - 1) % small firms with negative EPS in (t - 1) Market return in [t - 2; t - 1] Log M/B for small firms in (t - 2) Credit to private sector growth (%) in (t - 1) Dispersion in analysts' forecast (SD) in (t) Future market return in [t + 1; t + 4] Closed-end fund discount (%) in (t - 4)
Total IPOs per quarter 1.000
Total small-firm IPOs per quarter 0.9832* 1.000
Time trend -0.2217* -0.2997* 1.000
SOX dummy -0.2113* -0.2677* 0.2951* 1.000
Analyst coverage in (t-1) -0.2432* -0.2040* 0.106 -0.2217* 1.000
JOBS dummy -0.2094* -0.2261* 0.5134* -0.139 0.5736* 1.000
Real GDP growth (%) in [t; t + 3] 0.3052* 0.3179* -0.2676* -0.027 0.013 -0.084 1.000
IPO initial return in (t - 1) 0.139 0.136 -0.048 -0.136 -0.2518* 0.012 0.095 1.000
% small firms with negative EPS in
(t) -0.3007* -0.3511* 0.9131* 0.2380* 0.123 0.4323* -0.1775* -0.072 1.000
Market return in [t - 2; t - 1] 0.2263* 0.2170* -0.124 -0.027 0.119 0.059 0.2562* 0.2482* -0.2057* 1.000
Log M/B for small firms in (t - 2) 0.4350* 0.3598* 0.4549* 0.3308* -0.2948* -0.015 -0.063 0.097 0.4505* 0.155 1.000
Credit to private sector growth (%)
in (t - 1) 0.042 0.049 -0.2285* 0.2147* -0.115 -0.089 -0.031 0.119 -0.2565* -0.070 0.1644* 1.000
Dispersion in analysts' forecast (SD)
in (t) -0.029 0.038 -0.5228* -0.4594* 0.3652* -0.004 0.095 0.021 -0.4477* 0.017 -0.4423* 0.102 1.000
Future market return in [t + 1; t + 4] -0.015 0.003 -0.053 -0.038 -0.028 -0.055 -0.102 0.019 -0.033 0.086 0.012 0.084 0.073 1.000
Closed-end fund discount (%) in (t -
4.2. Results
The following section describes the results obtained by the empirical research. In the first subsection we will discuss the results for hypothesis 1. We will start this subsection by replicating the time-series regression of Gao et al. (2013). Then we move on to our main regression and after this we will run this regression several other times with some adjustments. The second subsections will provide the outcomes of hypothesis 2 and thereafter we will analyze hypothesis 3.
4.2.1. Hypothesis 1
4.2.1.1. Replication time-series regression Gao et al. (2013)
To test hypothesis 1, we will first replicate the time-series regression of Gao et al. (2013) in order to check whether the different dataset and small differentiation from the construction of the variables influence the results. In the paper of Gao et al. (2013) they also test whether the Economies of Scope hypothesis is the reason for the drop in small-firm IPO volume by regressing a quarterly time trend on the total (small-firm) IPOs.
Table 10 presents the results of these regressions. Just as in the paper of Gao et al. (2013), total IPOs scaled by GDP is the dependent variable in model 1, total small-firm IPOs scaled by GDP in model 2 and in model 3 the percentage of small-firm IPOs scaled by total IPOs is the dependent variable. The first and also the most important difference with the results of Gao et al. (2013) is that the coefficient on the time trend is only significant in our third model, whereas this coefficient is significantly negative in all models of Gao et al. (2013) (see appendix D). As the Economies of Scope hypothesis predicts a negative trend in all models, this is one of the reasons of Gao et al. (2013) to confirm the Economies of Scope hypothesis and it is interesting that we cannot support this in our regression. Another main difference is the significance on the SOX dummy. In the results of Gao et al. (2013) this coefficient is not statistically negative and therefore the authors conclude that the regulatory overreach hypothesis cannot be the explanation for the drop in small-firm IPO volume. However, in all our models this dummy is negative and significant. Because of this, we cannot ignore the regulatory overreach hypothesis as an explanation for the drop in small-firm IPO volume. Although we do not know the exact cause of these differences, a possibility is that the different sample compared to Gao et al. (2013) has a big influence on the results3.
Moreover, there are some control variables that are different in sign and significance for our
3Since Gao et al. (2013) uses Thomson One database and this study uses Compustat database, we have
compared the two samples of (small-firm) IPOs to find where differences arise. Unfortunately, we have only found that both of these databases are incomplete, but we have not been able to find the specific reasons for this.
models as well. For example, the percentage of small firms with negative EPS is significant in our first 2 models, where this is not the case in the study of Gao et al. (2013). For market return in [t – 2; t – 1], market return in [t +1; t + 4] and the quarter 1 dummy the opposite is applicable. These variables are tabulated with a significant coefficient in the table of Gao et al. (2013) (appendix D) in model 1 and 2, but in our 2 first models these coefficients are not significant. It is possible that these differences can be attributed to the differences in construction. In our analysis we use yearly percentages of small firms with negative EPS, while Gao et al. (2013) use quarterly data. Moreover, for market return Gao et al. (2013) only used NASDAQ returns, whereas we used different stock markets to calculate the market returns. Finally, all other variables that we have constructed in the exact same way as Gao et al. (2013) show the same sign and significance.
Table 10. Replication quarterly time-series regression on quarterly IPO volume of Gao, X., Ritter, J. R., & Zhu,
Z. (2013). Where have all the IPOs gone? Journal of Financial and Quantitative Analysis, 48(06), 1663-1692
Table 10 reports the results of the time-series regressions. The t-statistics are reported in parentheses below the coefficients. Small firms are defined as firms with annual sales below $50 million in the first year of their IPO. The dependent variables are the number of IPOs (model 1), the number of small-firm IPOs (model 2), all scaled by annualized quarterly real GDP. In model 3, the dependent variable is the fraction of IPOs from small firms. Time trend is the quarterly time trend that equals 0 for the first quarter of 1980 and increases by 0.1 for each quarter onward until the fourth quarter of 2015. SOX dummy is a past-Sarbanes-Oxley dummy that equals 1 from the third quarter of 2002 to the fourth quarter of 2007, and 0 otherwise. Real GDP (%) in [t; t + 3] is the percentage growth in real GDP in from quarter t to t + 3. IPO initial return in (t – 1) is the first-day return of small firms in percentages in quarter t - 1. % Small firms with negative EPS in the year of quarter (t) is the percentage of small public firms of total public firms that have negative EPS in quarter t. Market return [t - 2; t - 1] is the six-month value weighted return in percentage on the NYSE, AMEX, NASDAQ and ARCA excluding dividends in quarters t – 2 and t – 1. Log M/B for small firms in (t – 2) is the log of the sum of the market value of small firms divided by the sum of book value of small firms in quarter t – 2. Future market return in [t + 1; t + 4] is the value weighted return in percentage on the NYSE, AMEX, NASDAQ and ARCA excluding dividends from quarter t + 1 to t + 4. Closed-end fund discount (%) in (t – 4) is the discount on closed-end funds in percentages in quarter t – 4. Quarter 1 dummy is a first-quarter dummy that equals 1 in the first quarter of each year, and 0 otherwise. Newey-West standard errors are used. *** indicates significant at a 1 percent level, ** indicates significant at a 5 percent level and * indicates significant at a 10 percent level.
(1) (2) (3)
Total IPOs Total Small-firm IPOs
% Small-firm IPOs of total IPOs Time trend 0.0733 0.0250 -0.00325*** (1.70) (0.91) (-4.05) SOX dummy -7.884*** -4.756*** -0.0865* (-5.84) (-5.94) (-2.26) Real GDP growth (%) in [t; t + 3] 1.743*** 0.986*** -0.0118 (3.86) (3.60) (-1.76)
IPO initial return in (t - 1) 0.0160 0.0112 0.000730
(0.51) (0.56) (1.03)
% small firms with negative EPS in (t) -46.73*** -24.14** -0.213
(-3.66) (-3.00) (-0.83)
Market return in [t - 2; t - 1] -4.680 -3.292 0.483
(-0.21) (-0.23) (0.69)
Log M/B for small firms in (t - 2) 25.59*** 13.47*** 0.279**
(8.11) (6.80) (2.75)
Future market return in [t + 1; t + 4] -0.0164 -0.00715 -0.000337
