THE EFFECTS OF LIQUIDITY LEVEL, COMMONALITY AND
RISKS ON IPO TIMING.
Master thesis Student Tim Waaijer
Number 6036759
Study Business Economics Track Finance
Supervisor Patrick Tuijp, MPhil Date 7 August 2015 Abstract
This paper contributes to both IPO and liquidity literature by examining a liquidity based explanation of IPO fluctuations. IPOs tend to cluster in time and are correlated with up-‐markets. Ex-‐ante IPO liquidity is proxied by sector liquidity. Liquidity level and its three characteristics (i) liquidity commonality, (ii) return sensitivity to market liquidity and (iii) liquidity sensitivity to market have mixed influences on the number of IPOs. Liquidity level of the market portfolio appears to be significant and a 1% increase in illiquidity decreases the number of IPOs in three months with 6.8%. Liquidity level on the sector level is not significant and that also counts for the three characteristics. Only market wide liquidity is dominant in explaining the fluctuating number of IPOs, corresponding with other literature that a liquid after-‐ market is favorable for IPOs.
Keywords: Liquidity, IPO, liquidity commonality, Amihud
STATEMENT OF ORIGINALITY
This document is written by student Tim Waaijer who declares to take full responsibility for the content 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 of the University of Amsterdam is responsible solely for the supervision of completion of the work, not for the contents.
1. INTRODUCTION
IPO volume tends to fluctuate over time (Lowry, 2003). Volume is higher in up-‐markets than in down-‐ markets (Derrien, 2005; Pastor and Veronesi, 2005). In times of crisis, few companies are newly listed on exchanges. While market returns are not a dominant factor in explaining these fluctuations (Ritter and Welch, 2002), returns are lower in down-‐markets (French, Schwert, and Stambaugh, 1987) and illiquidity is correlated with down-‐markets (Brunnermeier and Pedersen, 2009). This indicates a certain correlation between liquidity and IPO fluctuations.
This paper combines research done on IPOs and liquidity and liquidity commonality. Chordia, Roll and Subrahmanyam (2000), Amihud (2002), Pastor and Stambaugh (2003), Acharya and Pedersen (2005), Brunnermeier and Pedersen (2009) established findings for liquidity and liquidity commonality as did Ritter and Welch (2002), Eckbo and Norli (2005) and Pastor and Veronesi (2005) for IPOs, among others. Most research done in the combined field is focused on after-‐market liquidity for IPOs and the development of liquidity on the amount of ‘money left on the table’ (Corwin, Harris and Lipson, 2004; Ritter and Welch, 2002). Eckbo and Norli (2005) look at liquidity and long-‐run performance of IPOs. How liquidity exerts influence on the future number of IPOs has not yet been researched. Besides liquidity level, liquidity has three more characteristics: liquidity commonality, return sensitivity to market liquidity and liquidity sensitivity to market return (Acharya and Pedersen, 2005). This paper aims to clarify the determinants for a company to go public regarding liquidity and its characteristics. The main question to answer is how do liquidity and its characteristics influence the decision of firms to go public? Liquidity of a company cannot be observed ex-‐ante IPO date and that makes research on this specific topic difficult. However, the contribution of this paper is the development of a proxy: for each company I use its sector to create a portfolio of which I calculate the returns and liquidity. These values then serve as proxy for the IPO company.
Acharya and Pedersen (2005) develop in their paper a liquidity-‐adjusted CAPM based on the Amihud illiquidity measure (Amihud, 2002). They provide evidence that not only market risk and liquidity commonality are priced but also risks between returns and liquidity. These risks are return sensitivity to market liquidity and liquidity sensitivity to market liquidity. As return sensitivity to market liquidity is documented by Pastor and Stambaugh (2003), liquidity sensitivity to market liquidity is undocumented in academic literature before the paper Acharya and Pedersen (2005). It is the most important factor in their model due to the highest estimated premium.
The amount of ‘money left on the table’, i.e. the underpricing -‐ which is the difference between the offer price and the closing price of the first day in percentages, must compensate for the risk that
investors have by buying IPO shares. Underpricing is higher in “hot issue” markets and is linked to expected stock returns. Stocks that are expected to be more risky than others have higher underpricing (Derrien, 2005). According to Acharya and Pedersen (2005) liquidity, liquidity commonality and its two risks are priced in expected stock returns. Therefore liquidity and its characteristics might exert influence on the fluctuating number of IPOs.
For companies it is important to understand the reaction of the market when they go public and have accurate expectations. For investors it is important to understand the motives of a company to go public and value it accordingly. The dynamic field between companies and investors is interesting because both parties want to maximize returns. With more information about significant factors both companies and investors can time their IPO and investment in IPOs more strategically. Also it is interesting for academic literature to understand why IPO volume fluctuates and how the decisions of companies and investors are influenced.
I use Amihud’s (2002) measure for illiquidity and Acharya and Pedersen’s (2005) normalized version of Amihud as illiquidity measures. Data for these measures is provided by CRSP and included only listed stocks on NYSE and Amex exchanges. CRSP has few recorded volume entries before 1983, which limits the ability to calculate reliable illiquidity proxies. Since the number of IPOs tends to fluctuate with the same rhythm as business cycles it is important to take a large timeframe to produce accurate results. Therefore the window of this research lies between January 1983 and March 2015.
Thomson One provides the necessary data for the number of IPOs for the sector portfolios and market portfolio. Only US IPOs on Nasdaq, NYSE and Amex are included in this research. The number of IPOs is not solely dependent on market conditions but also relies on the quality of the company itself. To control for these factors accounting data is required around the IPO issue date. Compustat provides this data.
Rolling betas indicate the monthly updated systematic risk, liquidity commonality, return sensitivity to market liquidity and liquidity sensitivity to market return. First step analysis is to provide evidence for the influence of liquidity level on the number of future IPOs. According to this research liquidity is significant for the market portfolio. Less market liquidity results in fewer future IPOs and corresponds with literature of Ritter and Welch (2002) and Brunnermeier and Pedersen (2009). On average a 1% increase in illiquidity according to Amihud’s (2002) measure estimates a decrease of 9.8% in the number of IPOs but it dependents on the number of forwarded months. Six-‐month forwarded estimates are impacted more by current the liquidity level than three-‐month forwarded estimates.
The second step is to test if sector liquidity also significantly explains the number of IPOs for the sector portfolio. I was not able to establish enough evidence to accept this hypothesis. The same holds for the liquidity betas. Adding controls and time-‐ and sector-‐fixed effects did not make a difference. The controls have large explanatory power while the betas do not add much to the explanatory power when they are significant. Based on the evidence in this paper, companies do not let their decision for an IPO be influenced by sector liquidity and the sensitivity to market return and liquidity.
Section 2 elaborates on important literature that explains studies in the fields of liquidity and IPOs and combines them. It shows that papers researching the combined field hardly exist. Section 3 clarifies the data sources and displays descriptive statistics of the variables. Further, the methodology and model is explained in Section 4 with summary statistics about the calculated betas. The results are presented in Section 5 for the two proxies for liquidity. Robustness results for fixed effects are presented in Subsection 5.2. Finally, I provide concluding comments in Section 6 together with a discussion.
2. LITERATURE REVIEW
This literature review elaborates on papers that are related to this paper. The determinants of the decision to go public can be inferred both from the ex-‐ante characteristics of the companies that go public and from the ex-‐post consequences of this decision. First, I will explain about IPOs, why firms would like to get listed and how one is structured. Furthermore, Section 2.1 elaborates on the fluctuating number of IPOs, as it seems that IPOs are clustered in time. The second subsection goes into further detail about liquidity after an IPO and what variables influence after-‐market liquidity. Third, Section 2.3 and 2.4 explain liquidity and liquidity commonality.
The main contribution of this paper is researching how liquidity commonality and liquidity risks exert influence on the number of IPOs. A general note for the equations in this paper: many equations are used to calculate variables for the market portfolio, sector portfolio and the portfolio containing only IPO firms. The difference is indicated by superscripts 𝑚, 𝑠, 𝑖 for market, sector and firm values respectively. Some equations might be presented in a different notation than in their original paper for a better understanding and comparison.
2.1 IPOs
An Initial public offering is one of the most important events on the capital markets for a company. The desire to raise capital and create a market where shareholders can convert their shares into cash are the main reasons for companies to go public (Ritter and Welch, 2002; Ellis, Michaely, and O'Hara, 2000). To guide the IPO process a company usually needs an investment firm as underwriter. Theoretically a firm does not need an underwriter but the process it needs to go through is extensive and complicated. Many requirements by the SEC need to be fulfilled and several reports with qualitative and quantitative analyses have to be handed in. Underwriters have the knowledge and experience to guide the process. An IPO can be structured in multiple ways. In a firm-‐commitment the underwriter guarantees that a certain amount will be raised by buying the entire offer and selling it to the pubic. A best-‐effort agreement does not give such a guarantee. In that case the firm takes the risk of not raising as much as it would like. An investment firm can diversify its risk by establishing a syndicate with multiple investment firms. One firm will lead the IPO while others sell a part of the shares and take on part of the risk. This initial agreement protects the underwriter against uncovered expenses (Ellis, Michaely, and O'Hara, 2000).
When the terms of the initial agreement between the underwriter and issuer are set, the registration statement is filed at the SEC with a S-‐1 form according to the Securities Act of 1933 (Loughran and McDonald, 2013). Information about financial statement, insider holdings, management background and the offering is disclosed. During the ensuing required cooling period by the SEC the company and the underwriter go on a road show with a preliminary prospectus (“Red Herring”) to create interest in the company among investors (Jenkinson and Ljungqvist, 2001). When interest in the issue is build up, both parties decide on the price. Two of the most important determinants of a successful IPO are the current market conditions and the success of the roadshow besides the potential success of the company. After the issue date the underwriter ensures stabilization by buying shares when order imbalances arise (Ellis, Michaely, and O'Hara, 2000). Table 1, a simplified version of table 1 on page 1044 in the paper of Ellis, Michaely, and O'Hara (2000), provides a clear overview of the IPO process.
The path of an IPO offers opportunities for behavioral factors but other, more rational, factors appear to be more dominant (Ritter and Welch, 2002). By going public, CEOs help facilitate the acquisition of their company for a higher value than they would get by remaining private (Brau, Francis, and Kohers, 2003). The paper of Ritter and Welch (2002) further names being publicly listed itself adds
to a bigger public role and better visibility. Chemmanur and Fulghieri (1999) develop the more conventional wisdom that listed firms offer more dispersion of ownership and hence try to reduce the principal-‐agent problem.
Table 1: IPO process
Steps of the IPO process with the main events displayed. Credits for this table go to Ellis, Michaely and O’Hara (2000) who provide an extended version of this table in their paper on page 1044.
Step Main event
1. Initial step Select book-‐running manager and co-‐manager
Letter of intent
2. Registration process Registration statement and due diligence
Red Herring
3. Marketing Distribute prospectus; road show
4. Pricing and allocation Pricing; allocation
5. Aftermarket activities Stabilization; overallotment; option
Research coverage
What is interesting for this paper to know is which factors determine the timing of an IPO when the decision to go public has already been made. Korajczyk, Lucas and McDonald (1991) argue that adverse selection arises when a company issues shares. Insiders with superior information have an incentive to issue shares when the firm is overvalued. Investors will lower their valuations of the firm’s equity consequentially. Since firms disclose information on a regularly basis (ea. earning announcements, annual reports) the price decline after an issue will be lower when information between insiders and outsiders is more symmetric. However, according to Ritter and Welch (2002) IPO fluctuations are unlikely to be explained by asymmetric information theories because entrepreneurs are more inclined to sell shares after valuations in the public markets have increased. This contradicts with the theory of Korajczyk, Lucas, and McDonald (1991). What they do agree on is the finding that firms postpone their equity issue if they know they are currently undervalued. More IPOs are seen in up markets than in down markets. According to research done by Ritter (2015), the number of IPOs is correlated with up markets and the average first-‐day underpricing.
A difficulty with research on IPO timing is that you only observe the number of firms that are actually going public. The number of private firms that could have gone public is not observed. Pagano, Panetta, and Zingales (1998) found a unique private data set of Italian firms that overcomes this problem. They observe that larger firms and companies in industries with high market-‐to-‐book ratios are more likely to go public. Lerner (1994) finds a similar result for biotech companies in the US. Industry
market-‐to-‐book ratios have a substantial effect on the decision for an IPO. Investor sentiment (measured by the discount on closed-‐end funds), growth opportunities and adverse selection considerations all are determinants of aggregate IPO volume according to Lowry (2003).
Lowry (2003) includes a lagged value of the number of IPOS. She does this consistently in her papers to capture the apparent seasonality of IPO fluctuations. I follow her treatment.
2.2 IPOs and liquidity
IPO firms are different than their already listed peers (Eckbo and Norli, 2005). They underperform in the long run (3-‐5 years) (Ritter and Welch, 2002), have different volatility characteristics (Pastor and Veronesi, 2005; Loughran and McDonald, 2013) and are different in liquidity (Ellul and Pagano, 2006). The study done by Ellul and Pagano (2006) shows that IPO underpricing is affected by after-‐market liquidity. Investors want to be compensated not only for fundamental risk and adverse selection costs but also for the expected liquidity of the shares they buy and for the risk of an illiquid secondary market. Their article concludes that greater aftermarket liquidity results in lower IPO underpricing. They come to this conclusion after controlling for firm specific factors as the age, size and leverage of a firm. Their results are robust for market conditions such as the number and proceeds of recent IPOs. One disadvantage of their predictions is the small time window of 337 British IPOs between 1998 and 2000. Their frame covers exactly the Dot-‐com bubble. Though Ellul and Pagano (2006) only focus on British IPOs, Ljungqvist and Wilhelm Jr (2003) find evidence that US IPOs during the Dot-‐com bubble are significantly different from IPOs before and after the bubble. Especially Nasdaq was highly attractive for technology firms to have their IPO during the Dot-‐com boom (Ljungqvist and Wilhelm Jr, 2003).
The paper of Ellul and Pagano (2006) provides also evidence that liquidity is priced as the liquidity-‐adjusted CAPM of Acharya and Pedersen (2005) predicts. They find results that show lower underpricing for stocks with higher after-‐market liquidity. Eckbo and Norli (2005) find that IPO stocks exhibit significantly greater stock turnover when compared to non-‐IPO stocks matched on stock exchange, equity size and book-‐to-‐market ratio. This offers a liquidity-‐based explanation for lower expected returns to IPO stocks on the long run. In a zero-‐investment portfolio with a long position in IPO firms and short positions in size-‐matched firms Eckbo and Norli (2005) find evidence of positive alpha. According to the efficient market hypothesis a positive alpha should not be possible (Basu, 1977) and this again underlines the legitimacy of the Acharya and Pedersen (2005) model.
2.3 Liquidity
Pastor and Stambaugh (2003) explain liquidity as follows: “Liquidity is a broad and elusive concept that generally denotes the ability to trade large quantities quickly, at low cost, and without moving the price”. Liquidity is measured by the bid-‐ask spread of a stock. The quoted spread:
𝑄!!= 1 𝑛!! 𝐴𝑠𝑘!!− 𝐵𝑖𝑑 ! ! 𝑚!! !!! !!! , (1)
and the effective spread: 𝐸!! = 1 𝑛!! 𝑝!!− 𝑚!! 𝑚!! !!! !!! , (2)
are the most common bid-‐ask spread measures, where 𝑝!! denotes the actual transaction price of
company 𝑖 at time 𝑡, and 𝑚!! denotes the midpoint of the bid-‐ask spread, expressed in percentages
(Chordia, Roll, and Subrahmanyam, 2000). Illiquid stocks have higher expected returns since there is risk in being not able to sell shares for a fair price (Acharya and Pedersen, 2005). Liquidity is identified as a risk factor, which is furthermore supported by Amihud and Mendelson (1986), Chordia, Roll and Subrahmanyam (2001) and Avramov, Chordia, and Goyal (2006). All studies find that less liquid stocks have higher expected returns due to increased transaction cost. It is more costly for an investor to trade when the bid-‐ask spread is larger. More liquid stocks generally have a lower premium to compensate investors.
In extreme situations market liquidity can dry up. The model of Brunnermeier and Pedersen (2009) sheds light on flights to quality. This can occur in periods of high market uncertainty. Risky securities become highly illiquid because investors choose to hold more liquid assets. Market makers must hold margins for their positions. These positions are financed by their capital and their margin cannot be larger than the amount of capital on the balance. When it is difficult to find capital for their balance, funding liquidity, this can have a spillover effect on market liquidity. In case of a market wide phenomena, this decrease in market liquidity affects on its turn funding liquidity. This spiral effect can cause market liquidity to dry up.
Furthermore, liquidity affects market volatility (Brunnermeier and Pedersen, 2009). In times of uncertainty in the market, market makers widen their spread between the bid-‐ and ask price which
immediately results in less liquid stocks. This enlarged bid-‐ask spread has the effect to increase volatility in stock prices (Brunnermeier and Pedersen, 2009). Volatility in its turn affects expected returns (French, Schwert, and Stambaugh, 1987). This demonstrates the importance of liquidity in the stock market. The best measure for liquidity is the bid-‐ask spread (Amihud, 2002). Unfortunately, data to calculate the spread is not always available. CRSP records bid and ask data for NYSE only since December 28, 1992 (CRSP, 2015). Amihud’s measure (2002) is highly correlated with the bid-‐ask spread and data to produce the measure is better available before 1992 than the spread data. Besides the Amihud (2002) measure there are more liquidity proxies (Goyenko, Holden and Trzcinka, 2009). Roll’s measure, which is a low frequency measure, is an estimator of the effective spread based on the serial covariance of the change in price (Roll, 1984). Holden (2009) and Goyenko, Holden, and Trzcinka (2009) develop a proxy of the effective spread based on observable price clustering, named the effective tick. Holden’s measure (2009) uses both serial correlation and price clustering and combines Roll’s (1984) measure and the effective tick (2009), which are special cases of Holden’s measure. These three measures are better at explaining liquidity transaction costs than at estimating price impact. Amihud’s (2002) illiquidity, Amivest liquidity and Pastor and Stambaugh’s measure (2003) are constructed to measure price impact of liquidity. Goyenko, Holden and Trzcinka (2009) find evidence in favor of the Amihud’s (2002) illiquidity measure in order to estimate price impact. Amihud’s (2002) illiquidity predicts price impact more accurately than the other two price impact proxies.
Proxies for liquidity also differ per geographical region (Lesmond, 2005; Bekaert, Harvey, and Lundblad, 2007; Loughran, Ritter, and Rydqvist, 1994). Countries that are less accurate and disciplined in recording data from their exchange require a measure that corrects for the frequency of this data. Lesmond (2005) finds that price-‐based liquidity measures such as Roll’s (1984) measure perform better at representing cross-‐country liquidity effects than do volume based liquidity measures. Within-‐country liquidity is best measured with the liquidity estimates of Lesmond, Ogden, and Trzcinka (1999) or Amihud (2002). Amihud’s (2002) illiquidity is downward biased in cross-‐country analysis in for low liquidity markets. This bias is manifested by zero returns that affect Amihud’s (2002) illiquidity. This downward bias does not exist in within-‐country analysis where Amihud’s (2002) illiquidity is on average 50% correlated with bid-‐ask spread (Lesmond, 2005).
Focusing on Amihud’s (2002) illiquidity measure: he finds evidence that over time expected market illiquidity affects the ex-‐ante stock excess return. This liquidity effect appears to be stronger for stocks of small firms. Amihud’s (2002) proxy for illiquidity:
𝐴!! = 1 𝐷!! 𝑅!! 𝑑𝑣𝑜𝑙!! !!! !!! , (3)
further named Amihud’s (2002) illiquidity (3). This is the daily ratio of absolute stock return to its dollar volume, averaged over some period, where 𝑅!! is the price change and not the percentage change and
𝑑𝑣𝑜𝑙!! is the traded volume multiplied by the closing stock price of the same day. The intuition behind Amihud’s measure is as follows: when 𝐴!! is high, meaning a stock is illiquid, the price of a stock moves a
lot in response to little volume.
The reason why Amihud (2002) uses this proposed measure is partly due to the non-‐availability of spreads to measure liquidity and other proxies. Only daily prices and volumes are required to calculate the Amihud measure and this is averaged over a span of time. Therefore, missing data will not induce errors as long as there are enough data points per period to give a reliable average. The more complex proxies such as Holden’s measure (Goyenko, Holden, and Trzcinka, 2009) require more data, which reduces the amount of observations
2.4 Liquidity commonality
Besides liquidity level there are more characteristics of liquidity (Acharya and Pedersen, 2005). Acharya and Pedersen (2005) solve a simple equilibrium model with liquidity risk. They modify the CAPM and add liquidity, which results in a liquidity-‐adjusted CAPM:
𝐸! 𝑟!!!! = 𝑟!+ 𝐸! 𝑐!!!! + 𝜆! 𝑐𝑜𝑣! 𝑟!!!! , 𝑟 !!!! 𝑣𝑎𝑟! 𝑟!!!! − 𝑐!!!! + 𝜆! 𝑐𝑜𝑣! 𝑐!!!! , 𝑐 !!!! 𝑣𝑎𝑟! 𝑟!!!! − 𝑐!!!! − 𝜆! 𝑐𝑜𝑣! 𝑟!!!! , 𝑐 !!!! 𝑣𝑎𝑟! 𝑟!!!! − 𝑐!!!! − 𝜆! 𝑐𝑜𝑣! 𝑐!!! ! , 𝑟 !!!! 𝑣𝑎𝑟! 𝑟!!!! − 𝑐 !!!! , (4)
where 𝑐!!!! is Amihud’s (2002) measure scaled to match the mean and standard deviation of returns and
𝜆! = 𝐸! 𝑟!!!! − 𝑐!!!! − 𝑟! is the risk premium. Equation (4) states that the required excess return is the
expected relative illiquidity cost 𝐸! 𝑐!!!! plus four betas multiplied by the risk premium. The Acharya
and Pedersen model (2005) identifies liquidity commonality and two liquidity risks:
1. 𝐶𝑜𝑣! 𝑐!!!! , 𝑐!!!! (liquidity commonality): when a security’s illiquidity co-‐moves with market
illiquidity its expected return is higher. Ceterus paribus, in times of market illiquidity a stock is riskier if it gets more illiquid. Stocks that have a lower covariance are less sensitive to market illiquidity and stay
more liquid. Therefore, these stocks are less risky and should have a higher price than their less liquid peers. The presence of a time-‐varying common factor in stock and market liquidity is also documented by (Chordia, Roll, and Subrahmanyam, 2000) and (Hasbrouck and Seppi, 2001). Most stocks’ illiquidities are positively correlated with market illiquidity. This is in line with a flight to liquidity in times of down markets as the model of Brunnermeier and Pedersen (2009) proposes. Acharya and Pedersen (2005) add the effect of commonality on asset prices to existing literature and find that the commonality is priced.
2. 𝐶𝑜𝑣! 𝑟!!!! , 𝑐!!!! (return sensitivity to market liquidity): due to the negative sign in the model
investors are willing to accept a lower return for stocks that have higher returns in illiquid times. Pastor and Stambaugh (2003) find that the average return of highly sensitive stocks to market liquidity is 7.5% higher annually than low sensitive stocks.
3. 𝐶𝑜𝑣! 𝑐!!!! , 𝑟!!!! (liquidity sensitivity to market returns): the third effect can be interpreted as
the willingness to accept lower expected return when a security can easily be sold at low cost in a down market. This ability of a stock is valuable. When an investor has wealth issues in a down market it lowers his transaction costs when his security can be sold at a reasonable price.
Acharya and Pedersen (2005) test their theoretical model and find evidence that the estimators of their liquidity-‐adjusted CAPM are significant. Due to collinearity of the measures it is difficult to statistically distinguish the relative return impact of the individual measures. Therefore, they use all four estimates in their analysis. Chordia, Roll, and Subrahmanyam (2000) document the existence of liquidity commonality as they expect liquidity to be influenced by common, broader market factors. Hasbrouck and Seppi (2001) find a small effect of common factors that determine liquidity. The effect of return sensitivity to market liquidity is empirically supported by Pastor and Stambaugh (2003). Average return of stocks who are highly sensitive to market liquidity exceeds average returns on lower sensitive stocks by 7.5% annually, adjusted for size, value and momentum factors.
Literature as mentioned in the section above provides variables that influence the choice of going public, as well as factors that influence the timing of an IPO. Liquidity has its own determinants and IPOs have different liquidity characteristics than their already listed peers. However, the study by Acharya and Pedersen (2005) implies that liquidity commonality and its two risks are priced and it could influence IPO timing when controlled for aggregate industry and market variables. This gives a better understanding of IPO waves. In the following section I will explain and examine the data set and support this with descriptive statistics.
3. DATA
Thomson One provides the data for the number of IPOs with its corresponding offer price and the amount of raised capital, further named proceeds. Thomson One offers extensive data about IPOs and therefore enables it to properly clean the data. According to Ritter (2015) not every IPO is equally useful in research. Ritter has done extensive research in the IPO field and I will use his methodology regarding cleaning IPO data. In most studies he excludes ADRs, ADSs, Beneficial Interests, Capital shares, Income Depositary, Limited Liability/ Partnership Interest, Trust Receipts, Units, Master Limited Partnership, SPERs, REIT, financial firms and penny stocks (offer price of less than $1 per share). The main reason for this paper to exclude them is that those firms might have different motivators to go public and therefore cause noise in the data.
This research only focuses on US data from major exchanges. Therefore, I only include IPOs from Nasdaq and NYSE (which includes Amex). Firms that get a listing on the smaller exchanges in the US might do so for specific reasons. Again, this paper aims to shed light on the effect of liquidity, liquidity commonality and its risks on the number of IPOs for the general picture and not for exceptionalities. For that same general picture it is important to have a large time window to account for year specific externalities. Reliable data from Thomson One is available from 1972 until present. However I will only use IPO data from January 1983 until March 2015 because CRSP does not provide accurate data on volume entries before 1983. I am therefore unable to reliably calculate Amihud’s (2002) illiquidity ratio.
Amihud’s (2002) illiquidity is calculated following equation (3). I compute the illiquidity for each portfolio and market portfolio as
𝐴!! = 𝑤 !!𝐴!! ! !!! , (5)
where 𝐴!! is the daily Amihud (2002) illiquidity for stock 𝑖 and 𝐴!! is the monthly averaged Amihud
according equal weighted factor 𝑤!!. For market values, 𝑠 is replaced for 𝑀. Table A10 in the Appendix
lists the number of IPOs per year and per SIC sector for more detail. The improved version of SIC, NAICS, started in 1997 and therefore is not in line with the large time window this research needs. Panel C of Table A10 describes the major divisions. To have enough observations I drop sectors that only saw 10 IPOs or less between 1983 and March 2015.
Next to Amihud’s (2002) measure of illiquidity this paper also replicates the scaled version according to Acharya and Pedersen (2005). They normalize the illiquidity measure to match the same level and variance with the level and variance of the effective spread as mentioned in Section 2.4 and to make it stationary. Acharya and Pedersen (2005) define the normalized version of Amihud as follows:
𝑐!!= min 0.25 + 0.30𝐴
!!𝑃!!!! , 30.00 . (6)
𝐴!! is Amihud’s (2002) illiquidity where 𝑠 indicates the sector. For market values 𝑠 should be replaced by
𝑀. 𝐴!! is then multiplied by an index ratio of market capitalization 𝑃!!!! . It is the ratio of the one month
lagged total market capitalization divided by the first observation of the market cap. January 1983 is the first observation in this research.
Figure 1: IPOs and market liquidity
The Amihud (2002) illiquidity measure calculated as an equal weighted average of the available CRSP stocks between 1983 and March 2015. The number of IPOs are the available IPOs from the Thomson One database and only from Nasdaq, NYSE and Amex. These are selected according to the methodology of Ritter (2015). Illiquidity and IPOs are correlated with a value of 0.1005. In line with research of Ritter (2015), the number of IPOs is lower in down markets, shown as the grey area below following NBER recessions. Exact dates are July 1990 until March 1991, March 2001 until November 2001, and December 2007 until June 2009.
The most important reason to normalize is that Amihud’s (2002) illiquidity is measured in percent per dollar whereas the Acharya and Pedersen model (2005) is specified in terms of dollar cost per dollar invested. Though my model does not necessarily require such normalization, it does offer a robust view on the model. A nice property of the normalization is that it made 𝑐!! stationary. This is
especially important for Acharya and Pedersen (2005) to match stationary expected returns. Stationary illiquidity is more reliable with forecasting than a non-‐stationary illiquidity (Stock and Watson, 2012) and historical relationships can be generalized for the future. As Figure 1 indicates, Amihud’s (2002) measure is not stationary. Figure A4 in the Appendix displays both 𝐴!! and its normalized version 𝑐!!. An
Augmented Dickey-‐Fuller test tests for stationarity of both variables. Table A11 in the Appendix reports the results of this test with the number of significant autocorrelated lags according to Bartlett's formula MA(q) with 95% confidence bands. 𝑐!! is significantly autocorrelated to 20 lags and is significantly
stationary with 0 and 30 lags specified. However, the Augmented Dickey-‐Fuller test is sensitive to the number of lags in a finite sample (Cheung and Lai, 1995). Though the results in Table A11 do no imply the normalization made Amihud’s (2002) measure stationary but it is reasonable to say that 𝑐!! is more
stationary than 𝐴!!.
The effect of illiquidity on the number of IPOs is studied for completed IPOs on NYSE and Amex exchanges between 1983 and March 2015. The data to calculate the illiquidity ratios is supplied by CRSP. Due to the market microstructure of the Nasdaq, that is different than that of NYSE and Amex, Nasdaq listed stocks are excluded (Gao and Ritter, 2010). Prior to 2001 Nasdaq recorded trade volume as trades with market makers and trades among market makers. This created double counting in most cases. Volume entries are needed to calculate Amihud’s (2002) measure. Nasdaq has altered its way of recording gradually since 2001 and since 2004 there are no important differences between recorded volume of Nasdaq and NYSE (Bennett and Wei, 2006). Although adjustment factors exist for the difference in recording (Gao and Ritter, 2010), these are estimates instead of real adjustment factors. Therefore I will not use stock data from Nasdaq from CRSP.
The equal-‐weighted market Amihud (2002) illiquidity is plotted in Figure 1 with the number of IPOs. In times of recession (according to NBER) there is more illiquidity than in the surrounding years as expected in theory and provided with evidence by Brunnermeier and Pedersen (2009) and Pastor and Stambaugh (2003). Though it appears that few firms are going public in illiquid times, illiquidity and the number of IPOs are positively correlated with 0.1005. Exact regions specified by NBER are July 1990 until March 1991, March 2001 until November 2001, and December 2007 until June 2009.
The monthly returns in the data set of this paper are daily stock returns from CRSP minus the three-‐month Treasury bill as risk free rate from the Federal Reserve Bank, averaged per month. To construct a portfolio return for each sector and a market return portfolio, the monthly returns are averaged per sector and market respectively:
𝑟!!− 𝑟 !! = 𝑤!! 1 𝐷!! (𝑟!! !!! !!! ! !!! − 𝑟!!), (7)
where, 𝑤!! is the equal for each stock, depending on the number of stocks in a sector month. 𝐷!! is the
number of traded days for stock 𝑖 in month 𝑡. 𝑟!! Is the daily return in percentages for stock 𝑖. For market
values, 𝑠 is replaced for 𝑀. Market return is plotted in Figure A5 in the Appendix.
Figure 2: Ratio of IPOs per exchange
The fractional distribution of IPOs between NYSE and Nasdaq. The percentages are the number of IPOs on each exchange divided by the total number of IPOs on NYSE and Nasdaq. Data from Thomson One.
The number of IPOs is different per exchange. Between January 1983 and March 2015 7,247 companies have gone public. 5,823 (80%) of those IPOs are completed on Nasdaq and 1,424 on NYSE. This is after cleaning the data according to the methodology of Ritter (2015). Especially during the Dot-‐ com bubble in 1999 and 2000 Nasdaq saw the majority of IPOs. The distribution of IPOs between NYSE and Nasdaq is plotted in Figure 2. In 1999 and 2000 NYSE had only 9% and 7% of the total IPOs
0% 20% 40% 60% 80% 100% 1983 1988 1993 1998 2003 2008 2013 NYSE Nasdaq
The Dot-‐com mania motivated many firms to raise capital on equity markets (Ljungqvist and Wilhelm Jr, 2003). After the bubble burst, the number of IPOs went down dramatically with 88% from 2000 until 2003 for Nasdaq.
In Table 2 we see the descriptive statistics of variables for sector portfolio (Panel A) and market portfolio (Panel B). Firm specific variables for IPOs are reported in Panel C of Table 2. Market-‐to-‐book (M/B) ratio is calculated by dividing the market value of equity by the book value of equity. The formula calculates the ratio for sector and market market-‐to-‐book ratio. For the IPO observations I use proceeds, the amount of capital raised, divided by the book value of equity.
The normalized Amihud illiquidity of Acharya and Pedersen (2005) has an average value of 0.0152% for the market portfolio. Monthly market return is averaged at 0.203%. The normalized value represents the cost of illiquidity according to Acharya and Pedersen (2005). These results correspond with the findings of Acharya and Pedersen (2005) since the net monthly return is on average positive 0.1878%. All important variables and controlling variables appear to have expected properties.
Market and sector optimism are calculated by the average price-‐earnings ratio (P/E) and market-‐to-‐book ratio (M/B). Derrien (2005) and Lowry (2003) use the same method for market optimism. The P/E ratio is calculated as the share price divided by the earnings per share. In a similar fashion as M/B, Cash and Leverage ratios are computed. The Cash ratio is the cash proportion of assets. The Leverage ratio is the proportion of long-‐term debt of total assets. Data needed to calculate these control variables comes from Compustat.
Although cash holdings are not suggested by academic literature, intuitively it can influence the choice of wanting your firm publicly listed. When a firm has the availability of highly liquid assets to make investments it is less inclined to go through the whole procedure of an IPO. Based on this simple intuition, the expectation is that cash holdings will be relatively low for certain IPOs. It differs per sector since some sectors are more prone to hold cash. There are more reasons for a firm to hold cash than only investment purposes and the correlation Table A12 in the Appendix does not indicate collinearity issues. Therefore it is added as control variable at firm-‐level.
Overall there is no danger for multicollinearity as few variables are highly correlated (Table A12). Most sector variables are highly correlated with its market equivalents but this is evident, the market portfolio is partly constructed by the same stocks that built the sector portfolios. But there are more variables highly correlated. Proceeds are correlated with market value. This is not completely anticipated by theory although it can be reasoned that market value represents a form of market optimism and the relative easiness of raising capital. Further analysis on this topic lies outside the scope
Table 2: Descriptive statistics of basic variables
Data is structured by the SIC classification and month-‐year. IPOs per sector are the number of IPOs for a specific SIC sector per month in a given year. That describes also the natural logarithm IPOs per sector. IPOs market is the number of IPOs for the total market per month in a given year. The time window starts at January 1983 and continues until March 2015, data is monthly. Table A10 in the Appendix provides more detail about the number of IPOs per sector and per month-‐year. Amihud per sector and market are the Amihud illiquidity measure following equation (3) and averaged for each SIC sector and the market according to (5), respectively. The normalized Amihud of Acharya and Pedersen (2005) follows the procedure of (6). Excess returns are reported as monthly values in percentages. Panel C displays observations for the IPO firms. Market-‐to-‐book ratio is calculated as the proceeds divided by the book value of equity.
Variable # of observations Mean Std. Dev. Min Max Skewness Kurtosis Panel A: Sector observations
Number of IPOs 4,356 1.315 0.901 1 14 5.343 43.574
Natural logarithm of
the number of IPOs 4,356 0.7986 0.2492 0.6931 2.7081 2.932 13.325
Amihud 44,848 0.0523 0.1051 0.0000 2.8005 6.573 82.860
Normalized Amihud 44,772 0.0158 0.0361 0.0000 1.2104 8.889 157.685
Excess sector return
in % 44,700 -‐0.196 10.416 -‐91.692 78.003 -‐0.847 17.354
Proceeds in $ mln 4,356 94.851 77.156 2.500 980.012 1.889 9.849 IPO underpricing in % 3,809 19.832 43.994 -‐99.400 636.360 5.316 46.241
MV mln $ 44,850 4,713 11,102 1.705 301,747 8.342 125.611
M/B ratio 44,459 1.4542 9.9647 0.0034 818.882 48.328 2966.67
P/E ratio 44,469 1.592 4.016 0.001 129.077 10.816 163.928
Panel B: Market observations
Number of IPOs 4,356 29.092 16.630 1 80 0.578 2.853
Natural logarithm of
the number of IPOs 4,356 3.2204 0.6553 0.6931 4.3944 -‐0.693 3.092
Amihud 387 0.0546 0.0330 0.0095 0.1866 0.742 3.313
Normalized Amihud 387 0.0152 0.0067 0.0039 0.0397 1.218 4.659
Excess market return
in % 386 0.203 4.050 -‐18.581 14.678 -‐0.807 6.390
Proceeds in $ mln 4,356 101.503 97.1745 12.2566 548.509 2.308 9.292 IPO underpricing in % 3,949 19.039 19.465 -‐19.920 183.470 3.032 14.589
MV mln $ 387 4,256 2,974 755.89 12,280 0.697 2.688
M/B ratio 387 0.4114 0.1161 0.0821 0.6287 5.276 54.238
P/E ratio 387 2.585 0.717 1.477 4.867 0.736 3.044
Panel C: IPO observations
M/B ratio 3,049 2.1476 23.3688 0.0006 1052.632 37.243 1530.49