Master Thesis
Finance
Underpricing and long-run performance of
Initial Public Offerings in the UK
Reputation of Underwriters and Uncertainty
By Gigl Brouwer
Abstract
This study investigates the potential effect of underwriters’ reputation and uncertainty about the value of the firm on (initial returns) underpricing and long-run performance of IPOs by using a sample of 229 UK Main Market IPOs between 1995 and 2013. We perform both a cross-sectional and multi-factor regression analysis. Our findings show that reputation of underwriters is negatively related to initial underpricing, but fails to find a robust positive effect on long-run performance. Furthermore, we show significant evidence for positive relations between uncertainty and underpricing and between uncertainty and long-run underperformance. Based on portfolios of combinations of underwriters’ reputation and uncertainty we see that post-IPO, the underwriter loses influence. The effect of uncertainty is persistent through time and mainly explains the differences in performance in the long-term. The analysis furthermore did not find an interaction between underwriters’ reputation and uncertainty.
Student number: s2016796
Program: MSc Finance
Supervisor: Prof.dr. W. Bessler
Date: 11-12-2016
Email: Giglbrwr@gmail.com
Word count: 17,785
Field Key Words: IPO, long-run underperformance, underwriters’ reputation,
2
I. Introduction of research topic
On May 11th, 2011 the social networks company LinkedIn made its debut on the NASDAQ Stock Exchange. The company hired Morgan Stanley and Bank of America’s Merrill Lynch to manage the initial public offering (hereafter ‘IPO’) process. The IPO raised 353 million dollars for LinkedIn. However, after the first trading day, the share price had seen a staggering increase of 109%. This first-day price jump left more than 370 million dollars in the hands of the investors instead of in the hands of LinkedIn.1 The situation portrayed in the LinkedIn anecdote is a classic case of IPO underpricing.2 It seems at first paradoxical that firms that are presumably in need of capital are willing to leave significant amounts of money ‘on the table’. However, this phenomenon has been from all times and is keeping academics busy for over 40 years.
Another anomaly that is keeping academic busy for quite some time is the subsequent long-run underperformance of IPOs. This phenomenon was first discovered and documented by Ritter (1991), who find that issuing firms in the US substantially underperform a sample of matching firms over a period of three years. For companies planning to go public, such as LinkedIn at the time, the underpricing and long-run underperformance anomalies are highly relevant.
When a firm plans to go public, a wide variety of issues have to be considered. One of these issues is the choice of the investment bank that will underwrite the issue. In the case of LinkedIn, Morgan Stanley and Bank of America’s Merrill Lynch were chosen as the lead underwriters3 in the IPO process. The important role underwriters have in this process does not stop at the IPO date. Investment banks provide a wide variety of so-called aftermarket services such as market-making, providing analyst coverage and price support services. The importance of the role of underwriters is also acknowledged by academics. A vast body of literature suggests that the initial underpricing and long-run underperformance anomalies can be mitigated when the choice of the underwriter is taken into consideration (Jain & Kini, 1999; Carter, Dark, & Singh, 1998). In other words, selecting an underwriter with above
1Linkedin Retains Most Gains Second Day After Surging in IPO. (2011). Retrieved August 15, 2016 from
Bloomberg:
http://www.bloomberg.com/news/articles/2011-05-18/linkedin-raises-352-8-million-in-ipo-as-shares-priced-at-top-end-of-range.
2 Underpricing and first-day returns are used interchangeably in the remainder of this thesis. 3
3 average skills to optimally price and support the issue reduces underpricing and long-run underperformance. Besides the superior skill, selecting a high reputable underwriter in itself can already reduce underpricing and long-run underperformance. The reasoning behind this line of thought is related to the principles of asymmetric information. The literature mainly attributes underpricing and long-run underperformance to information asymmetry between investors (Rock, 1986) or between underwriter(s) and investors (Baron & Holmström, 1980). In order to alleviate this information imbalance, issuing firms hire more reputable investment banks to signal the quality of the firm to the market and thereby playing a ‘certification’ role. When this signal is picked up by investors, it reduces the uncertainty about the true value of the IPO firm. Beatty & Ritter (1986) call this uncertainty about the offering value ‘ex-ante uncertainty’ (p.213). They find that higher ex-ante uncertainty (hereafter ‘uncertainty’) leads to greater underpricing. A negative relation between uncertainty and long-run performance is found by, among others, Loo, Lee & Yi (1999). They argue that if the reputational capital is valuable to the underwriter, high reputable underwriters are more likely to avoid risky IPO firms and only underwrite the best IPOs associated with firms with lower risk characteristics. The empirical literature provides interesting insights into the performance of IPOs. Most literature implies however that stock returns are systematically linked to specific variables, such as underwriter reputation. As stated above, the literature assumes that underwriter prestige alleviates uncertainty, which will result in less underpricing and better long-run performance. However, the interaction effects between uncertainty and underwriter reputation have never been tested. For instance, the positive effect investment bankers’ reputation may have on IPO performance might diminish once accounting for uncertainty. In a similar vein, if the disclosed information is being perceived as trustworthy and hence uncertainty about the value is already low, does it matter which underwriter you select? A similar study is done by Doukas & Gonenc (2005). The authors jointly test underwriters’ reputation and venture capital effects on long-run performance and find that the effects of underwriters’ reputation on long-run performance only matters in the absence of venture capital. Apparently, underwriters’ reputation is not as systematically related to long-run performance as is suggested in most literature.
4 investigate different IPO combinations of underwriter reputation and uncertainty in an effort to discern if such characteristics are related to underpricing and long-run performance of IPO firms. Eventually, this paper answers the following research question:
What is the (simultaneous) effect of underwriter reputation and uncertainty on an IPOs first-day return and long-run performance?
In order to answer the research question, we use a sample of 229 UK IPOs. The UK IPO market is interesting, because reputational capital is particularly important for UK underwriters as they trade repeatedly with the same small group of institutional investors (Barnes & Walker, 2006). All IPOs are issued on the London Stock Exchange Main Market (hereafter ‘Main Market’) between January 1995 and November 2013. Most UK studies do not differentiate between the AIM4 and Main Market even though there are important differences that have to be taken into account.5 Two approaches will be used to test the effects on long-run performance. Firstly, we perform a three-year buy-and-hold abnormal return approach. Secondly, to improve the robustness of this research, a variation on the calendar approach is constructed and employed in the form of Carhart’s (1997) four-factor model.
This paper contributes to the existing literature by testing uncertainty and underwriters’ reputation simultaneously. The main findings are mixed. We find significant results for a positive relationship between uncertainty and initial underpricing and a negative relation between uncertainty and long-run performance. Underwriters’ reputation, on the other hand, does not show convincing results. The effects of underwriters’ reputation on underpricing are confirmed, however, the effects of underwriters’ reputation on long-run performance are highly mixed and found to be inconclusive. Furthermore, we do not find significant evidence for a relationship between underwriters’ reputation and uncertainty.
The remainder of this thesis is organized as follows. Section II provides a review of the literature related to IPO performance, the effects of underwriters’ reputation and the effects of uncertainty. Section III describes the methodology and the various variables of interest. Section IV describes and provides the data collection procedure and key statistics. Section V presents and discusses the empirical results and section VI concludes the paper.
4 ‘AIM’ stands for Alternative Investment Market and is the second market in the UK.
5 The Main Market and AIM are not equal in terms of regulations and corporate governance rules. The AIM has
5
II. Theory
2.1. IPO performance
The post-IPO performance of issuing firms has received great attention from the academic world. Initially, initial underpricing was the main focus point. Underpricing represents a discount from its presumed intrinsic fair value and is measured as the difference between the first-day price compared to the offering price (Ritter & Welch, 2002). However, after the pioneering research of Ritter (1991), a new strand of research with regard to IPO performance – long-run underperformance – was born.
2.1.1. Initial underpricing
One of the first to document on initial underpricing are McDonald & Fisher (1972) who find a mean excess first-day return of 28.5% in the United States. A more comprehensive study is conducted by Ritter (1984) who finds that for approximately 5,000 IPO firms in the US, the average first-day return is 18.8%. Numerous subsequent studies have followed and confirmed this empirical result in almost every capital market across the globe. Including a more recent study of Coakle, Hadass & Wood (2009), who find an excess first-day return of 10.50% in the UK.
6 specific circumstances quality firms signal through underpricing. It is expected that only good firms can recoup this loss with subsequent offerings once more information about the issuing firm becomes available. In the late 1990s, initial returns substantially increased. This resulted in a new strand of literature questioning whether the aforementioned theories are able to explain these abnormal levels of underpricing. These studies turn to behavioral explanations, such as irrationality of investors. They claim that investors hold (too) optimistic beliefs about the companies future prospects in the late 1990s. Issuers particularly sell shares to ‘book building’ investors. Book building investors on their turn sell the shares to small, irrational investors. Bookbuilding investors require underpricing to gain on the subsequent sale. Ljungqvist, Nanda & Singh (2006) argue that the issuer brings the optimal amount of stocks to the market to capture the ‘surplus’ from overoptimistic investors, but also to ensure profits for the book building investors. Most of these - and other empirical studies assume that underpricing is undertaken intentionally. Gouldey (2006), however, distinguishes between intentional and unintentional incentives to underprice. He shows that unintentional underpricing exists and is a consequence of investors’ heterogeneous expectations about the intrinsic value of equity.
Even though literature is consistent in confirming the presence of underpricing, financial economists still offer many competing theories as to why this phenomenon is so prevalent. Implicit in the theories is the assumption that the secondary market price is eventually corrected to the IPO stock’s intrinsic value (Gouldey, 2006).
2.2.2. Long-run underperformance
7 underperformance that persists for between 36 to 60 months after floatation and their results support the market timing hypothesis of Loughran and Ritter (1995, 2000).
Accordingly, raising new equity generally leads to long-run underperformance suggesting information asymmetry or agency problems. However, not all firms underperform an appropriate benchmark. For instance, Brav & Gompers (1997) find a significant cross-sectional difference in performance of venture capital-backed IPOs versus nonventure-backed IPOs. They suggest that venture capitalist involvement better signals the quality of an IPO. VCs also have a direct stake in the issuing firm. With a direct stake, VCs can exhibit more active involvement in corporate governance of the issuing firm to better align operations with the interest of investors (Krishnan, Ivanov & Masulis, 2011). Moreover, Kuntara & Nikhil (2007) find that IPOs with low block sales significantly outperform IPOs with high block sales from the lockup expiration until the third year. Other comparable cross-sectional outperformance of IPOs is found for firms not active in post-IPO acquisitions (Amor & Kooli, 2016), for firm age, where older firms outperform younger firms (Ritter, 1991), size (Carter, Dark, & Singh, 1998), ownership structure (Jain & Kini, 1994) and an important concept in cross-sectional difference in performance is the role of underwriters’ reputation. These differences in performance are in most cases related to the degree of uncertainty about the value of the issuing firm that is decreased. Less uncertainty generally results in better performance.
2.2. Underwriters’ reputation
As mentioned earlier, IPO underpricing and long-run underperformance are often attributed to information asymmetry between parties in the IPO process. One mechanism that has been developed to help bridge this gap is signaling. Similar to venture capitalists, underwriters can play an important role in alleviating this information imbalance. Investment banks (and venture capitalists) can have a certification role in which they certify the quality of the issuing firm.
8 firms’ performance first surfaced in the early 70s. Logue (1973) examined 250 IPOs and argues that the degree of prestige of an investment bank signals the riskiness of the issuing firm to the market. Carter & Manaster (1990) approach underwriters’ reputation in a similar vein and find a significant negative relation between underwriters’ reputation and underpricing. Price run-ups compensate uninformed investors for the risk of trading against superior information. Price run-ups harm the issuing firm, as it decreases the potential capital gains from going public. Nimalendran, Ritter & Zhang (2007) show that in a sample of 3,499 IPO firms in the US, more than $93.5 billion is left on the table for the investors. Consequently, higher quality firms will attempt to reveal their identity to the market by selecting prestigious underwriters. At the same time, reputational underwriters attempt to maintain their reputation by market only high-quality firms. This is where the certification role comes into play. Pre-IPO, investment banks screen issuing firms in order to make sure the firms they underwrite are not harmful to their reputation. Since their reputation is linked to the succes of the IPO, high-reptable investment banks that underwrite firm certifies the quality to investors.
9 operations is unlikely to be great and underwriters generally do not bear large financial risks. Dong, Michel & Pandes (2011) however argue in line with Carter, Dark, & Singh (1998) and Chau (2014) that underwriters reputation and long-run performance are negatively related. By observing 7,407 US IPOs from 1980 to 2006 they find that IPOs with greater underwriter quality significantly outperform IPOs with low quality and showed positive abnormal returns. The aforementioned relations are formalized in the following hypotheses:
Hypothesis I: Underwriters’ reputation is negatively related to initial underpricing. Hypothesis II: Underwriters’ reputation is positively related to long-run aftermarket performance.
2.3. Uncertainty
IPOs involve more uncertainty than existing listed firms. Before a firm enters the market, determining the intrinsic value is inherently uncertain. A large strand of literature states that the level of uncertainty about the fair value of the firm determines the level of underpricing (Gouldey, 2006).
10 Recently, studies have defined another type of uncertainty. They state that despite underwriters’ efforts, a residual element of information asymmetry and uncertainty will filter into the secondary market and is more persistent than previously assumed. This type of uncertainty is termed ex-post value uncertainty (Falconieri, Murphy & Weaver, 2009). On a sample of 2,029 IPOs in the period 1993 to 1998, they find a significant positive relation between several proxies for ex-post value uncertainty and underpricing. Another research of Beneda & Zhang (2009) finds that both initial (ex-ante) idiosyncratic volatility and subsequent volatility changes in the first trading year are significantly negatively related to long-run performance.
Based on these theories, this study expects that uncertainty is positively related to underpricing and negatively related to long-run performance. When formalizing the aforementioned results, the following hypotheses are:
Hypothesis III: Uncertainty about the value of the issuing firm is positively related to initial underpricing.
Hypothesis IV: Uncertainty about the value of the issuing firm is negatively related to long-run aftermarket performance.
2.4. Joint effect underwriters’ reputation and uncertainty
This study expects that the foreseen positive effect of underwriters’ reputation is more pronounced for firms facing greater uncertainty. Furthermore, we expect that differences in performance between high- and low reputational underwriters diminish after accounting for uncertainty.
11 signaling role of underwriters. Certifying the quality of the issuing firm is most important for firms that are hard to value (i.e. high uncertainty).
Advocates of the signaling theory state that low-risk firms benefit from exposing their risk profile to the market. One way of doing that is by hiring a prestigious underwriter who is able to communicate the ‘identity’ of the issuing firm. This signal reduces the information imbalance and uncertainty with regard to the value of the firm (Carter & Manaster, 1990). However, this study argues that once uncertainty is negligible, the (if any) positive effects of signaling diminishes. If investors agree about the value of the firm, no difference in performance should exist between high- and low reputational underwriters. Both lines of reasoning lead to the following hypotheses:
Hypothesis V: The greater the uncertainty, the more positive the effect of underwriter reputation on long-run performance will be.
Hypothesis VI: Low uncertainty will lead to no significant difference in performance between firms with high-reputable underwriters and low-reputable underwriters.
III. Methodology
3.1. Dependent variableThe main objective of this research is to look at the (simultaneous) effect of underwriters’ reputation and uncertainty on the performance of UK IPOs. Both initial underpricing and long-run three-year performance are used as a measure of performance.
3.1.1. Underpricing
To measure underpricing, this study follows the common method used in literature, which is the simple percentage return between the offer price and the first-day closing price.
𝑟𝑖,𝑡 = ((𝐶𝑃𝑖,𝑡− 𝑂𝑃𝑖
𝑂𝑃𝑖 ) − 1 (1)
where 𝐶𝑃𝑖,𝑡 is the closing (bid) price of firm 𝑖 at time 𝑡 and 𝑂𝑃𝑖 is the offer price of the shares
of firm 𝑖.
12 average daily market returns, market movements are not able to explain the abnormally high first-day returns.
3.1.2. Long-run performance
3.1.2.1. Buy-and-hold abnormal return
Aftermarket performance is measured in two ways. First, buy-and-hold abnormal returns (BHAR) are used. This event-time methodology is widely used to calculate long-run returns (see, e.g. Dong, Michel & Pandes, 2011; Carter, Dark, & Singh, 1998) and is said to be superior compared to cumulative abnormal returns (CAR) when longer time horizons are examined. (Gregory, Guermat & Al-Shawawreh, 2010). Buy-and-hold returns are defined as
𝐵𝐻𝐴𝑅 = 1 𝑁∑ [∏(1 + 𝑟𝑖,𝑡) 36 𝑡=1 − ∏(1 + 𝑟𝑏𝑒𝑛𝑐ℎ𝑚𝑎𝑟𝑘,𝑡) 36 𝑡=1 ] 𝑁 𝑡=1 (2) where 𝑟𝑖,𝑡 is the raw return for firm 𝑖 at month 𝑡 and 𝑟𝑏𝑒𝑛𝑐ℎ𝑚𝑎𝑟𝑘,𝑡 is the raw return of the selected benchmark at event month 𝑡. All return data is adjusted for dividends.
BHARs for each firm are calculated by compounding 36-monthly equally weighted returns. We use the closing price of the first trading day as a starting point. Consistent with Levis (2011), this study does not want the long-run returns to be influenced by potential initial under (over) pricing. Furthermore, if a fund dies before the three-year holding period, an equal amount is invested in the market index. Hence, missing sample firm returns are filled in with benchmark portfolio returns (Mitchell & Stafford, 2000).
As mentioned by Gregory, Guermat & Al-Shawawreh (2010), a benchmark should “match the characteristics of the event firm as closely as possible” (p. 615). The FTSE All-Share Index (hereafter ‘FTSE AS’) is the most appropriate and best matching benchmark because the index includes small-, medium- and large-cap stocks and represents approximately 98% of the total UK market capitalization.6 We therefore use the FTSE AS as the benchmark.
3.1.2.2. Multi-factor model
Recent studies have challenged the assumptions inherent in event-time methodologies such as BHAR (e.g. Mitchell & Stafford, 2001). Instead, this strand of literature advocates a
6
13 calendar-time approach as a better alternative.7 They argue that an event-time methodology potentially leads to event clustering and cross-correlation in IPO returns. Therefore, shocks in the economy will cause outcomes to be overstated. Event-time methods do not take this correlation into account, whereas cross-correlation and event clustering is accounted for in a calendar-time approach (Brav & Gompers, 1997).
To improve the robustness, We will employ a version of the calendar-time approach, as suggested by Brav & Gompers (1997).8 They build their model on the implications of Fama & French (1993) that a multi-factor model may explain the cross-section of stock returns. We will employ Carhart’s (1997) four-factor model. The four-factor model is an extension of the Fama-French three-factor model. This extended model is defined as
𝑟𝑖𝑡− 𝑟𝑓𝑡 = 𝛽1+ 𝛽2∗ 𝑀𝐾𝑇𝑡+ 𝛽3∗ 𝑆𝑀𝐵𝑡+ 𝛽4∗ 𝐻𝑀𝐿𝑡+ 𝛽4∗ 𝑈𝑀𝐷𝑡+ 𝜀𝑡 (3) where 𝑀𝐾𝑇𝑡 is the monthly return of the stock market minus the risk free rate.9 𝑆𝑀𝐵𝑡 is
Small (cap) stocks Minus Big stocks and refers to the premium on small stocks. 𝐻𝑀𝐿𝑡 is High (book/price) Minus Low and refers to the premium on value stocks. 𝑈𝑀𝐷𝑡 is Up Minus Down and refers to the momentum factor of Carhart (1997).10 Finally, the 𝛽1 relates to the
part of the return that the four factors in the model cannot ‘explain’.11
We use the intercept from time-series regression as an indication of risk-adjusted performance. We will build sub-portfolios to test for differences in performance between high- and low underwriters’ reputation and high- and low uncertainty. Furthermore, we will construct four sub-portfolios based on underwriters’ reputation and uncertainty.
3.2. Independent variables
3.2.1. Underwriters’ reputation
Several proxies have been developed to determine underwriters’ reputation. The three most appealing are the Carter & Manaster (CM), Johnson & Miller (JM) and Megginson & Weiss
7 Also known as Jensen’s alpha approach. For each calendar month the average abnormal return on a portfolio
consisting of all firms that conducted an IPO within the past T-t months is calculated. These returns are regressed on a corresponding benchmark return (Espenlaub & Tonks, 2000).
8 Disadvantage of the calendar-time approach is that this method has lower power to detect abnormal
performance compared to event-time methods (loughran and Ritter (2000).
9 As the risk free rate the 10-year UK government bond is taken. 10
A more detailed explanation about the construction of and the reasoning behind the factors can be found in the article of Gregory, Tharayan & Christidis (2013).
11 Datasets containing the monthly SMB, HML and momentum factors for the UK market are retrieved from the
14 (MW) measures. Of these three measures, the CM measure is the most widely used in literature. CM uses ‘tombstone announcements’ to create their ranking. However, the construction of such ranking requires a very tedious process. Carter, Dark, & Singh (1998) find that the MW measure is strongly correlated with CM and can therefore be seen as a good alternative. In this measure, the market share of the underwriter serves as a proxy for reputation. The authors argue that market share of the lead underwriter(s) is positively correlated with reputation. Market share is calculated as the percentage of total proceeds during the sample period. When multiple underwriters are involved with an IPO, the market share is calculated by using the equally weighted market share. Finally, underwriters are divided into two groups; high - or low reputation based on an average and median cut-off point. A list of the top 20 underwriters can be found in table XIII of appendix E.
3.2.2. Uncertainty
Uncertainty can be either ex-ante or ex-post. This study will use two proxies for uncertainty. Firstly, to estimate ex-ante uncertainty, we partially follow Ritter (1984). He uses the standard deviation of returns of the first trading month as a measure of ex-ante uncertainty.12 The advantage of the standard deviation as a proxy for risk stems from its proven record in financial research. Standard deviation encompasses all kinds of firm-specific risks, such as cash flow risk, earnings risk etcetera. It is plausible to assume that the more difficult it is for the market to value the IPO firm, the more volatile the stock price will be after flotation. Beneda & Zhang (2009) furthermore suggest that the level of return volatility provides market information about the firms’ idiosyncratic risk. However, this study argues that the standard deviation of returns does not entirely cover ex-ante firm-specific uncertainty. We use the excess monthly volatility, compared to the benchmark which is defined as
𝑒𝑥𝑐𝑒𝑠𝑠 𝜎𝑖= [{[ 1 𝑛 − 1] ∑(𝑟𝑖𝑡− 𝑟̅)𝑖 2 21 𝑡=2 } 1 2 − {[ 1 𝑛 − 1] ∑(𝑟𝑚𝑡− 𝑟̅̅̅)𝑚 2 21 𝑡=2 } 1 2 ] ∗ √21 (4) where 𝑟̅𝑖 is the average return of stock i from day two through day 21 and 𝑟̅̅̅𝑚 is the average return of the benchmark over the same period. The relative volatility is more relevant in this matter, as it provides more information on the firm-specific risk. If the market is highly volatile when a firm enters the market, the returns of that firm will most likely also be more volatile. Moreover, if that same firm would enter during a calm period, the return volatility of
12
15 the firm will be lower. The risk characteristics of the issuing firm, on the other hand, do not differ that much between periods of high – and low volatility. Therefore, excess volatility is assumed to be a more precise estimate of uncertainty. Furthermore, Beneda & Zhang (2009) find a significant relation between initial firm-specific volatility and underpricing, but are unable to find a significant relation between uncertainty and idiosyncratic volatility changes during the first year. The authors further argue that post-IPO performance is influenced by both initial idiosyncratic volatility and subsequent volatility changes in the first trading year. We therefore use the excess volatility 21-days proxy only for tests on underpricing.
We use a different proxy for tests on long-run performance. This proxy is more related to ex-post uncertainty and the riskiness of future cash flows. The measure furthermore encompasses both initial idiosyncratic volatility and the subsequent volatility changes during the first year. We partially follow Carter, Dark, & Singh (1998), who calculate the standard deviation of IPO stock raw returns using a time-series of 255 days, commencing on day six. We look at rolling 21-day volatility in excess of the volatility of the market from day six to day 261. We argue that in normal volatility proxies, large shocks are not sufficiently accounted for, especially when taking the one-year volatility. Furthermore, similar to the first measure of uncertainty, we assume that the standard deviation of a stock in excess of the market volatility is a better proxy for uncertainty. We calculate the excess rolling standard deviation measure in a similar way as is done in formula 3. The excess volatility for both measures of each IPO firm in the sample can be found in a scatter diagram in figure IV of appendix D.
3.3. Control variables
Several control variables are included to isolate investment bank reputation effects and uncertainty effects. The non-binary variables are based on natural logarithmic (hereafter “logarithmic”) values to reduce non-normality. Exact definitions and how they are measured can be found in table X in appendix A.
This study controls for age. Older firms are less risky and therefore the market should demand a lower return for IPO offerings (Ritter, 1984). We expect that age has a negative effect on underpricing and a positive effect on long-run performance. Age is measured as the logarithmic number of years between the date of incorporation and the IPO date.
16 underpricing and positively related to long-run performance. Offer proceeds are measured as the logarithmic pound value of proceeds raised with the IPO.
Ritter (1991) observes that smaller issuing firms tend to perform poorer. Furthermore, larger firms have better access to financial resources necessary for firm survival and profitability (Finkle, 1998). Firm size, measured as the logarithmic pound value of market capitalization, is expected to be negatively related to underpricing, and positively related to long-run performance.13
Furthermore, Ritter (1984) finds that in ‘hot issue’ periods initial underpricing is significantly higher than in cold periods. Later, Ritter (1991) found that IPO firms showing the most significant long-run underperformance are concentrated in these hot periods of high-volume IPOs. This study controls for ‘hot issue’ markets by using a dummy variable.
Finally, several studies find a positive relation between venture-backed IPO firms and performance. Brav & Gompers (1997) show that venture-backed IPOs are less subject to severe underperformance, compared to non-venture-backed IPOs. The authors claim that venture capital helps to overcome asymmetric information in financial markets. Therefore, a negative relation with underpricing and a positive relation with long-run performance are expected.
Table XII of appendix C shows the correlation between all variables. Correlations between the main variables are not disturbingly high, so multicollinearity does not seem to be an issue.
3.4. Regression models for cross-sectional regressions
This study uses cross-sectional regression analysis. Because BHARs and underpricing are often positively skewed and thus not normally distributed, we use White’s heteroskedasticity-robust standard errors and covariances. The regression model used to test for underwriters’ reputation – and uncertainty effects on initial underpricing has the following form:
𝐼𝑈𝑃𝑖 = 𝛽0+ 𝛽1∗ 𝑈𝑊𝑅𝐸𝑃𝑖𝑑𝑢𝑚𝑚𝑦 + 𝛽2∗ 𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑡𝑦𝑖 + 𝛽3∗ 𝐴𝑔𝑒𝑖𝑙𝑜𝑔+ 𝛽4
∗ 𝑀𝑎𝑟𝑘𝑒𝑡 𝑐𝑎𝑝𝑖𝑙𝑜𝑔+ 𝛽5∗ 𝑃𝑟𝑜𝑐𝑒𝑒𝑑𝑠𝑖𝑙𝑜𝑔+ 𝛽6∗ 𝐻𝑜𝑡 𝑖𝑠𝑠𝑢𝑒𝑖𝑑𝑢𝑚𝑚𝑦 + 𝛽7∗ 𝑉𝐶𝑏𝑎𝑐𝑘𝑒𝑑𝑖𝑑𝑢𝑚𝑚𝑦
(5)
13 Several studies use total assets of the issuing firm before going public as a proxy for size, however, we argue
17 where UWREP refers to underwriter reputation grouped on a median and average cut-off point. Uncertainty refers to the excess standard deviation of the first trading month.
To test for effects on long-run performance, the following regression is used: 𝐵𝐻𝐴𝑅𝑖 = 𝛽0+ 𝛽1∗ 𝑈𝑊𝑅𝐸𝑃𝑖+ 𝛽2∗ 𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑡𝑦𝑖 + 𝛽3∗ 𝐴𝑔𝑒𝑖𝑙𝑜𝑔+ 𝛽4
∗ 𝑀𝑎𝑟𝑘𝑒𝑡 𝑐𝑎𝑝𝑖𝑙𝑜𝑔+ 𝛽5∗ 𝑃𝑟𝑜𝑐𝑒𝑒𝑑𝑠𝑖𝑙𝑜𝑔+ 𝛽6∗ 𝐻𝑜𝑡 𝑖𝑠𝑠𝑢𝑒𝑖𝑑𝑢𝑚𝑚𝑦 + 𝛽7∗ 𝑉𝐶𝑏𝑎𝑐𝑘𝑒𝑑𝑖𝑑𝑢𝑚𝑚𝑦
(6)
where UWREP refers to underwriter reputation split on a median and average cut-off point. Uncertainty refers to rolling excess standard deviation of the first 255 days. Furthermore, to test for an interaction effect between underwriters’ reputation and uncertainty, we add an interaction term to both regressions which is UWREP times Uncertainty.
IV. Data and descriptive statistics
4.1. Sample selection and data sourcesWe use two primary databases in order to construct the dataset. The first sample of UK IPOs is created with Zephyr, a database that contains worldwide IPO and M&A data. Initially, firms with IPOs on the LSE Main Market between 1995 and 2013 are selected. As this study also looks at three-year abnormal returns, we select 2013 as end year. We choose 1995 as the starting year. Firstly, because the sample period now includes two complete stock market cycles and IPO waves. Secondly, data is more scarce prior to 1995. Furthermore, we include only firms with a known deal value and exclude firms active in the financial, insurance and utility industry. These industries are often government regulated, therefore pricing and returns of these firms can be biased (Peltzman, 1968). The initial sample from Zephyr consists of 332 IPOs. The second database we use to complement this sample is Thomson Research. A search on the same criteria results in 41 additional IPOs, leaving a total initial sample of 373 IPOs. However, despite as being recognized as IPOs, for 10 firms it is a secondary listing. Another 51 IPOs are excluded because these are identified as global depository receipt.14 One IPO appears to be an insurance company.
Next, to construct the underwriters’ reputation variable, information on the lead underwriters per IPO have to be derived. We use Thomson One Banker as the primary source to obtain
14
18 underwriter information. For several IPOs Thomson One Banker does not provide information. In these cases, we manually look up IPO prospectuses. Ultimately, we exclude 29 IPOs due to missing information on underwriters. Furthermore, information on other variables, such as gross proceeds raised with the IPO are also obtained from Thomson One Banker. Regarding the date of incorporation, for a number of firms only the year is available. For these firms, June 31st of that year is taken as the date of incorporation. For two firms that started floating in the year of incorporation, but before the 30th of June, January 1st is taken. Monthly returns, first-day returns and daily returns to calculate return volatility, are collected from Thomson DataStream. Due to insufficient or missing return data, another 53 firms drop from the sample. Initially, the three-year BHARs include several severe outliers. To reduce this effect, but not delete useful data, we winsorize monthly buy-and-hold returns at 98%. Because a calendar-time approach has fewer observations per month than an event-time approach, we winsorize calendar-time monthly returns at 95% in order to reduce non-normality. Furthermore, all monthly returns are in Pound Sterling.15 The final sample includes 229 IPOs. Table XI in appendix B contains a detailed overview of the above-mentioned search process.
4.2. Sample distribution
Table I presents the yearly distribution of the 229 IPOs in the sample, together with the mean and cumulative yearly distribution of offer proceeds raised with the IPO and market capitalization. Two IPO cycles can be identified with a strong increase and a subsequent drop in the annual number of firms going public. The first IPO cycle spans from 1996 to 2003, reaching a peak of 46 IPOs in 2000 and a subsequent drop to 4 firms going public in 2003. This ‘new economy’ period of high growth and low inflation ended with the burst of the dot-com bubble. The second cycle, spanning from 2004 to 2009, shows similar dynamics with a high of 21 IPOs and a low of only 1 IPO in 2007 and 2009. Interestingly, IPO waves and particularly declines in IPO activity are highly correlated with the market. We see in figure I that the peak of IPO activity and the peak and tipping point of the market are in the same year. Both offer proceeds and market capitalization show a similar pattern. Furthermore, IPOs in the beginning of the sample period are smaller on average than in the more recent years. Especially 2011 stands out in terms of offer proceeds and market capitalization, which
15
19 is caused by the largest IPO ever in the UK16; Glencore plc, with offer proceeds of £6,193 million and a market capitalization of £36,344.25 million. We suppress the potential effects of this outlier by taking the logarithm of offer proceeds and market capitalization. Furthermore, all returns are equally weighted, so this IPO does not excessively influence our results.
TABLE I
Distribution of IPOs per year
Table 1 reports the year-number distribution of 229 IPOs in the sample (1995-2013), together with the mean - and cumulative offer proceeds and mean - and cumulative market capitalization per year. Offer proceeds contain the total offer proceeds of the IPO and market capitalization refers to the total market capitalization at the time of the offer.
Year Frequency Offer proceeds (GBPm) Market capitalization (GBPm)
Nr. Obs.¹ % of total Mean Cum.² Mean Cum.²
1995 3 1.31% 45.54 136.62 127.19 381.58 1996 3 1.31% 77.17 231.50 213.87 641.62 1997 7 3.06% 59.50 416.48 171.64 1,201.47 1998 26 11.35% 119.73 3,113.10 272.68 7,089.74 1999 18 7.86% 181.31 3,263.52 528.20 9,507.59 2000 46 20.09% 87.67 4,032.73 403.28 18,550.87 2001 5 2.18% 231.33 1,156.66 294.38 1,471.88 2002 12 5.24% 314.62 3,775.42 635.19 7,622.25 2003 4 1.75% 424.62 1,698.47 837.05 3,348.21 2004 16 6.99% 143.25 2,291.99 233.14 3,730.29 2005 18 7.86% 247.05 4,446.94 768.27 13,828.80 2006 21 9.17% 222.92 4,681.30 689.23 14,473.82 2007 21 9.17% 226.81 4,763.09 572.36 12,019.56 2008 3 1.31% 372.31 1,116.93 1,494.54 4,483.63 2009 1 0.44% 61.97 61.97 192.05 192.05 2010 10 4.37% 291.78 2,917.85 1,213.08 12,130.76 2011 4 1.75% 1,747.88 6,991.51 10,320.06 41,280.23 2012 3 1.31% 98.29 294.87 295.86 887.59 2013 8 3.49% 397.45 3,179.60 1,472.00 11,776.01 Total 229 48,570.54 164,617.94
1; Number of observations, 2; Cumulative
16
Glencore IPO Moves to Next Step. (2011). Retrieved December 27, 2016 from The Wall Street Journal: http://www.wsj.com/articles/SB10001424052748703509104576331363602040414
FIGURE I
Distribution of IPOs per year
0 1000 2000 3000 4000 0 15 30 45 60 n u m b er o f IPO s Year
IPOs per year FTSE ALL SHARE
20
4.3. Performance and IPO firm characteristics
Table II reports the summary statistics of the sample. Panel A presents issuing firm characteristics and panel B depicts the IPO - and underwriter characteristics. The mean excess volatility first 20-days (hereafter ‘uncertainty 20-days’) is 6.99% and for excess rolling volatility first 255-days (hereafter ‘uncertainty 255-days’) the mean is 7.30%. A test for mean equality shows that excess volatility does not change significantly during the first year, where you would expect a decline as more information becomes available. On the other hand, as can be seen in figure I, most firms went public just before the market significantly dropped in value, which might partially explain the insignificant difference in volatility in the first year after floatation. The scatter plot in figure IV of appendix D furthermore shows that the trend is relatively similar, but the dispersion of uncertainty 20-days is considerably higher than for uncertainty 255-days. This could be explained by the larger estimation window for uncertainty 255-days. Also, the volatility varies greatly through time, with the years 2000 and 2001 showing the highest peak and greatest dispersion. The mean (median) market share of underwriter reputation is 10.49% (7.10%), which is relatively high compared to the findings of US IPOs (Carter, Dark, & Singh, 1998; Beatty & Welch, 1996). The Main Market is the stock market for larger, more established companies. Such companies are in most cases underwritten by one or several of the larger investment banks, resulting in that the top ten banks are involved in almost 60% of the sample’s IPOs. Furthermore, IPO firms listed between 1995 and 2002 have higher mean excess volatility, both 20-days and 255-days, than the sample of IPOs between 2003 and 2013, as shown in figure III. It appears that IPOs issued during the first IPO wave of our sample were more uncertain than IPOs issued during the second wave. Also, underwriters’ reputation is higher for the 2003-2013 sub-sample. More IPOs were underwritten by high reputable investment banks in the second wave, which is likely the result of consolidations in the financial services industry. These two observations might suggest that there is some positive interaction between underwriters’ reputation and uncertainty.
21 period shows negative returns for both the sample firms as for the market index, whereas in the latter period all returns (except BHAR) are positive. Furthermore, initial underpricing is more than twice as high during the first sample period compared to the second period. The first period includes the dot-com (IPO) bubble, which is characterized by high levels of underpricing.
TABLE II
Issuing firm-, IPO- and underwriter characteristics
Table II presents the summary statistics of 229 IPOs in the sample (1995-2013). Panel A presents the issuing firms’ characteristics. All variables are measured at the time of issuing. Panel B shows the IPO and underwriter characteristics. Furthermore, the means of all variables are reported of the sub-samples 1995-2002 (N:120) and 2003-2013 (N:109). Excess volatility first 20 days is measured as the return volatility of the first 20 days, excluding the first day, in excess of the market volatility for that same period. Excess rolling volatility first 255 days is the volatility over 21 days, rolled over 255 trading days, in excess of the rolling market volatility for that same period and commencing on day six. Underwriter reputation is measured as the percentage of total market value brought to the market. VC (venture capital) backed and hot issue are both binary variables.
Variable 1995-2002 2003-2013 Nr.
Obs.¹ Mean Median Std. Dev² Min.³ Max.⁴
Prop. of
sample Mean Mean
Panel A: Issuing firm characteristics
Age 229 9.34 3.23 21.58 0.02 192.90 8.40 10.37
Offer proceeds (GBPm) 229 212.10 78.40 484.04 2.10 6,193.48 134.38 297.66
Market capitalization (GBPm) 229 718.86 224.47 2,542.30 15.68 36,344.25 387.23 1083.95 Panel B: IPO- and underwriter
characteristics
Uncertainty 20-days (%) 229 6.99 4.73 8.42 -5.74 55.67 8.38 5.47
Uncertainty 255-days (%) 229 7.30 5.80 6.80 -2.30 64.52 7.67 6.89
Underwriter reputation (% market
share) 229 10.49 7.10 9.54 0.00 27.24 8.39 12.81
VC backed (%) 229 35.81
Hot issue (%) 229 64.19
1; Number of observations, 2; Standard deviation, 3; Minimum, 4; Maximum, 5; Proportion of the sample
22 TABLE III
Initial underpricing and three-year buy-and-hold abnormal return
Table III lists the summary statistics of the performance characteristics of the 229 IPO firms (1995-2013). The table presents the firm's buy-and-hold return, the benchmark buy-buy-and-hold return and the BHAR (buy-buy-and-hold abnormal return), all measured over the first 36 months after floatation (3-years). Furthermore, the means are reported of the sub-samples 1995-2002 (N: 120) and 2003-2013 (N: 109). Statistics reported are the mean, median, standard deviation, minimum, maximum and percentage of negative returns.
Variable 1995-2002 2003-2013
Nr. Obs.¹ Mean Median
Std.
Dev² Min.³ Max.⁴ Negative Mean Mean
Buy-and-hold return (%) 229 -1.86 -31.41 95.08 -99.43 404.16 62.88 -13.36 10.80
Buy-and-hold return FTSE AS (%) 229 6.34 3.63 30.24 -44.52 85.66 48.03 -3.87 17.59
BHAR (%) 229 -8.20 -35.24 83.34 -162.97 355.15 65.07 -9.48 -6.79
Initial underpricing (%) 229 9.03 5.06 18.48 -20.43 189.65 22.71 12.66 5.03
1; Number of observations, 2; Standard deviation, 3; Minimum, 4; Maximum
Furthermore, in figure III the average BHR (black line), BHAR (gray line) and the benchmark returns (dashed line) per month are depicted. As can be seen from the graph, during the first year, the sample of IPO firms outperforms the market, with a starting downtrend after 3 months. The subsequent two years, both the buy-and-hold return and the buy-and-hold abnormal return remain negative with a short but steep drop in return at 12 months. These declines could be the result of expiring lock-up agreements. Interesting is the development of BHAR in the graph when the first-day return is added. The buy-and-hold return after 36 months is approximately zero, which gives rise to the idea that underwriters deliberately underprice such that the long-run underperformance after 3-years does not become negative to protect their own reputation.
FIGURE II Underpricing per IPO firm
-50 0 50 100 150 200 U n d er p ri ci n g ( % ) IPO date Underpricing
23 FIGURE III
Buy-and-hold return and buy-and-hold abnormal return over 36 months
V. Results
5.1. Characteristic differences: underwriters’ reputation vs. uncertainty
To better understand the relation between the different variables, tables IV and V show the issuing firm-, IPO- and return characteristics of the high- and low underwriters’ reputation sub-samples and uncertainty sub-sample. Uncertainty 20-days and uncertainty 255-days are split into two sub-samples (high and low) based on the average volatility. The columns ‘statistic (p-value)’ report the t-statistic (for the mean difference)17
and the Mann-Whitney statistic (for median difference) with the corresponding p-value.
In both sub-samples in table IV, the difference in underpricing between the high- and low sub-sample is significant at a 5% confidence level, for both the mean and the median. The mean (median) first-day return in the high reputation sub-sample on the median cut-off point is 5.66% (4.25%) point lower compared to the low reputation group. The results confirm our expectations that IPO firms underwritten by a prestigious underwriter experience less underpricing than when underwritten by a less reputable underwriter. Also, the median difference of BHARs between the different two median sub-samples is significant at a 5% confidence level, but this does not hold for the other observations. However, the BHARs of
17
First an equal variance test is performed. If the variance between the high- and low group is significantly different at a 5% significance level, the t-test is replaced by the Satterthwaite-Welch t-test.
24 the high underwriters’ reputations groups are consistently more positive (less negative) than for the low underwriters’ reputation sub-sample. Nevertheless, from figure V in appendix F it appears that the positive effect of underwriters’ reputation is only present in year +3, whereas no difference can be identified for years +1 and +2. Furthermore, offer proceeds and market capitalization are significantly larger (at a 1% significance level) for both high-reputation groups. This result can have two explanations. First, on average, prestigious investment banks underwrite larger firms and these firms raise more money with an IPO. Second, the high reputable underwriters sub-samples display lower levels of underpricing. Assuming that less underpricing results from an offer price closer to its intrinsic value, less underpricing leads to higher market value and higher offer proceeds. Finally, the median excess uncertainty 20-days and uncertainty 255-20-days are significantly larger for the high reputation group based on the average cut-off point, which seems to contradict our expectations.
25 uncertainty 20-days and low uncertainty 255-days group. Younger firms are generally harder to value and their future profitability is likely to be more uncertain. Finally, the number of firms that went public in a hot-issue period is significantly higher in the high uncertainty 20-days group, than when uncertainty is low. However, for the uncertainty 255-20-days groups the effect is the opposite.
5.2. Returns and the interactive effects of underwriters’ reputation and uncertainty
In this section, we delve a bit deeper into the role of underwriters’ reputation and effect of uncertainty on underpricing and long-run performance. This analysis attempts to shed additional light on whether underpricing and long-run performance are likely to be related to underwriters’ reputation. We examine the issue by constructing four different portfolios based on uncertainty 255-days and underwriter reputation (median). In table VI results for the mean (median) underpricing and year +1, year +2 and year +3 buy-and-hold abnormal returns can be found for each portfolio. Furthermore, a test for differences will shed light on the return differences between the constructed portfolios.
TABLE IV
Mean and median values across underwriters’ reputation
Table IV lists the issuing firm, IPO and return characteristics for two reputational sub-samples. The first sub-sample is based on the median cut-off point and the second sub-sample based on the average cut-off point. All of the issuing firm characteristics are measured at IPO date. Panel A lists the issuing firms age, offer proceeds (money raised with the IPO) and market capitalization. Panel B lists the IPO's uncertainty 20-days (in excess of the market volatility over the same period), uncertainty 255-days ( measured as the return volatility of 21 days rolled over 255 days (starting at trading day 6), in excess of the market volatility over the same period), whether the IPO was backed by a venture capitalist and whether the firm went public during a hot issue period with high IPO volume. Panel C lists the 36-months BHAR and the initial underpricing. The absolute difference between the low and high reputation sub-sample is reported in the 'difference' columns. The subsequent column reports the t-statistic (for the mean difference) and the Mann-Whitney statistic (for the median difference) and the corresponding p-value is shown between parenthesis.
Variable Underwriter reputation median Underwriter reputation average
Total Low High Difference Statistic (p-value) Low High Difference Statistic (p-value)
Number of observations 105 124 118 111
Underwriter reputation (% market share) mean 1.17 18.39 1.88 19.66
median 0.78 19.00 1.02 19.08
Panel A. Issuing firm characteristics
Age mean 9.34 8.90 9.71 0.81 0.281 (0.779) 9.22 9.47 0.25 0.088 (0.930)
median 3.23 3.27 3.10 -0.17 0.488 (0.625) 3.62 2.34 -1.28 1.078 (0.281)
Offer proceeds (GBPm) mean 212.10 61.71 339.45 277.74*** 4.866 (0.000) 69.21 363.99 294.78*** 4.677 (0.000)
median 78.40 32.85 168.06 135.21*** 9.248 (0.000) 34.50 185.75 151.25*** 9.682 (0.000)
Market capitalization (GBPm) mean 718.86 190.40 1166.34 975.94*** 3.182 (0.002) 196.31 1274.35 1078.04*** 3.171 (0.002)
median 224.47 97.90 493.69 395.79*** 9.750 (0.000) 101.69 531.51 429.82*** 10.335 (0.000)
Panel B. IPO characteristics
Uncertainty 20-days (%) mean 6.99 7.06 6.93 -0.13 -0.119 (0.909) 6.79 6.90 0.11 0.373 (0.709)
median 4.73 3.73 5.51 1.78* 1.851 (0.064) 4.01 5.28 1.27** 2.043 (0.041)
Uncertainty 255-days (%) mean 7.30 7.09 7.47 0.38 0.410 (0.682) 6.68 7.95 1.27 1.417 (0.158)
median 5.80 5.00 6.14 1.14 1.466 (0.143) 4.82 6.30 1.48*** 2.735 (0.006)
VC backed (%) mean 35.81 33.33 37.90 4.57 0.718 (0.473) 33.05 38.74 5.69 0.895 (0.372)
Hot issue (%) mean 64.19 67.62 61.29 -6.33 -0.993 (0.322) 66.95 61.26 -5.69 -0.895 (0.372)
Panel C. Return characteristics
Initial underpricing (%) mean 9.03 12.09 6.43 -5.66** -2.326 (0.021) 12.29 5.56 -6.73*** -2.830 (0.005)
median 5.06 7.01 2.76 -4.25*** 2.833 (0.005) 7.57 2.50 -5.07*** 3.430 (0.001)
36-Month BHAR (%) mean -8.20 -15.30 -2.19 13.11 1.184 (0.238) -9.74 -6.57 3.18 0.289 (0.774)
median -35.24 -48.28 -18.38 29.90** 2.179 (0.029) -44.79 -25.57 19.22 1.240 (0.215)
* 10% significance level
** 5% significance level
27 TABLE V
Mean and median values across uncertainty
Table V lists the issuing firm, IPO and return characteristics for two uncertainty sub-samples. The first sub-sample is based on the first 20 days volatility, in excess of the market volatility for the same period (excluding the first trading day) with the cut-off point based on the average volatility. The second sub-sample is based on the first 255 days volatility, in excess of the market volatility for the same period (excluding the first 6 days) with the cut-off point based on the average volatility. All of the issuing firm characteristics are measured at IPO date. Panel A lists the issuing firms age, offer proceeds (money raised with the IPO) and market capitalization. Panel B lists the underwriters' reputation (measured as the percentage of total market value brought to the market, whether the IPO was backed by a venture capitalist and whether the firm went public during a hot issue period with high IPO volume. Panel C lists the 36-months BHAR and the initial underpricing. The absolute difference between the low and high reputation sub-sample is reported in the 'difference' columns. The subsequent column reports the t-statistic (for the mean difference) and the Mann-Whitney statistic (for the median difference) and the corresponding p-value is shown between parenthesis.
Variable Uncertainty 20-days Uncertainty 255-days
Total Low High Difference Statistic (p-value) Low High Difference Statistic (p-value)
Number of observations 142 87 142 87
Panel A. Issuing firm characteristics
Age mean 9.34 11.96 5.07 -6.89*** -2.852 (0.005) 11.57 5.70 -5.87** -2.474 (0.014)
median 3.23 3.82 2.25 -1.57* 1.665 (0.096) 3.26 2.96 -0.30 0.416 (0.677)
Offer proceeds (GBPm) mean 212.10 229.82 183.18 -46.64 -0.705 (0.481) 233.10 177.81 -55.29 -0.837 (0.404)
median 78.40 78.82 75.00 -3.82 0.384 (0.701) 77.41 78.74 1.33 0.036 (0.971)
Market capitalization (GBPm) mean 718.86 798.97 588.09 -210.88 -0.607 (0.544) 781.77 616.17 -165.60 -0.477 (0.634)
median 224.47 197.34 259.19 61.85 1.033 (0.302) 183.73 259.19 75.46 1.622 (0.105)
Panel B. IPO & underwriter characteristics
Underwriter reputation (% market share) mean 10.49 10.46 10.56 0.10 0.077 (0.939) 10.00 11.11 1.11 1.003 (0.317)
median 7.10 7.10 11.34 4.24 0.063 (0.950) 7.10 4.87 -2.23 0.959 (0.337)
VC backed (%) mean 35.81 38.03 32.18 -5.85 -0.893 (0.373) 38.73 31.03 -7.70 -1.192 (0.235)
Hot issue (%) mean 64.19 59.15 72.41 13.26** 2.041 (0.042) 65.49 62.07 -3.42 -0.522 (0.602)
Panel C. Return characteristics
Initial underpricing (%) mean 9.03 6.14 13.73 7.59*** 3.066 (0.002) 7.75 11.11 3.36 1.144 (0.255)
median 5.06 5.10 4.87 -0.23 1.307 (0.191) 5.14 4.87 -0.27 0.066 (0.948)
36-Month BHAR (%) mean -8.20 -3.36 -16.10 -12.74 -1.121 (0.263) 4.98 -29.71 -34.69*** -3.108 (0.002)
median -35.24 -22.79 -51.62 -28.83* 1.651 (0.099) -7.92 -55.51 -47.59*** 3.832 (0.000)
* 10% significance level
** 5% significance level
The mean and median difference tests between the LR/LU - and HR/HU portfolio for +1- and +2-year periods shows that the LR/LU group IPOs significantly outperform the HR/HU group. The outperformance of the LR/LU portfolio compared to the HR/HU portfolio suggests that uncertainty exerts a greater influence on long-run performance than underwriter reputation. Contrarily, the influence of underwriters’ reputation on underpricing is greater than the effect of uncertainty, given the significant median difference between the LR/LU and HR/HU portfolios and the significant mean (median) differences between HR/HU vs. LR/HU and HR/LU vs. LR/LU. A visualization of the 36 months buy-and-hold abnormal returns is presented in figure VII of appendix F. Against our expectations, these findings suggest that the relationship between underwriters’ reputation and investors’ returns decreases over time after controlling for uncertainty. Even though the majority of studies assumes there is a positive relation between underwriters’ reputation and long-run performance, as mentioned in the literature section, there are arguments supporting the opposite. It could be that these alternative explanations better fit our results. Part of the role of underwriters during and after
TABLE VI
BHAR and underpricing differences of IPO firms: Underwriter reputation and uncertainty
Table VI presents the buy-and-hold abnormal returns for 12, 24 and 36 months and first-day returns of four IPO stock portfolios. Underpricing is measured as the first-day raw return. Buy-and-hold abnormal returns are measured per month, where in the first month the first trading day is excluded. Underwriter reputation is measured as the percentage of total market value brought to the market by an underwriter. Firms underwritten by investment banks with an above median market share take the value of 1 and 0 otherwise. Uncertainty is measured as the return volatility of 21 days rolled over 255 days (starting at trading day 6), in excess of the market volatility over the same period. Firms with an above average volatility take the value of 1 and 0 otherwise. The IPO portfolios consist of (1) firms underwritten by high-reputable investment banks with low uncertainty (HR / LU); (2) firms underwritten by high-reputable investment banks with high uncertainty (HR / HU); (3) firms underwritten by low-reputable investment banks with low uncertainty (LR / LU) and (4) firms underwritten by low-reputable investment banks with high uncertainty (LR / HU). The tests for differences between the groups are performed using the t-statistic (for the mean difference) and the Mann-Whitney statistic (for the median difference). The corresponding p-value is shown between parenthesis.
Underpricing year 1 year 2 year 3
IPO stock portfolios N Mean Median Mean Median Mean Median Mean Median
29 the IPO is to support and stabilize the share price. Furthermore, most IPOs have a lockup period where for a predetermined amount of time large shareholders are restricted from selling shares. Both aspects are particularly important for the first (half) year. Consequently, as find by Logue, Rogalski, Seward & Foster-Johnson (2002), the effect of underwriters’ reputation might diminish after a year. Going back to figure V in appendix F, the BHAR line starts to decline after three months, and the drop becomes even larger at 12 months. These observations could be an indication of price stabilization efforts during the first few months and negative returns when lockup agreements expire. However, based on figure III, no difference can be found between high and low underwriters’ reputation for years +1 and +2. Besides, the results of table VI do not present evidence of a decreasing relation between underwriters’ reputation and long-run performance after a year. The difference in relation between underwriters’ reputation and long-run performance in our sample does not gradually decrease as documented by Logue, Rogalski, Seward & Foster-Johnson (2002), but appears to diminish after the first day.
5.3. Regression analysis
5.3.1. Cross-sectional regression analysis
Table VII (table VIII) exhibits the cross-sectional regression results of initial underpricing (BHAR) on underwriters’ reputation, uncertainty 20-days (uncertainty 255-days) and the control variables. In columns (1) through (3), the independent variables are separately regressed on underpricing (BHAR). Columns (4) to (6) show regressions of the independent variables on underpricing (BHAR), including the control variables. In column (7) and (8) all variables are regressed, including the control variables and in column (9) and (10), the interaction term is added.
5.3.1.1. Underpricing
Results in columns (1) of table VII show that underwriter reputation (median) is negatively (-0.056) related to underpricing at a 5% confidence level. Yet, in column (4) and column (7), the coefficient loses its significance but nonetheless exerts a negative influence on initial underpricing.
Underwriter reputation (average) is also negatively (-0.067 for column (2)) related to
underpricing and is able to retain its significance, as shown in columns (2), (5) and (8). Hence, after controlling for a variety of other explanations for underpricing, underwriters’
30 negative relation between underwriter reputation and underpricing are consistent with the findings of Carter, Dark, & Singh (1998). The results suggest that underpricing is a compensation for the risk uninformed investors take. The reputation of underwriters lowers this risk of uninformed investors by certifying the quality of the issuing firm. The findings can furthermore suggest that high-reputable underwriters are more skilled in extracting value-relevant information during the book-building process, improving the pricing of the issue and reducing underpricing. (Pollock, Chen, Jackson & Hambrick, 2010). Our findings confirm the expected negative relation between underwriters reputation and underpricing of hypothesis I. In column (3) of table VII, uncertainty 20-days enters the regression with a positive (0.542) and significant coefficient at a 5% confidence level. The significantly positive effect of
uncertainty 20-days is persistent through regression (6), (7) and (8). However, once the
interaction term is added in column (9) and (10), uncertainty loses its significance, while maintaining a similar positive coefficient (0.573 for column (10)). The large coefficient highlights the (economic) importance in explaining underpricing and confirms hypothesis III where is expected that uncertainty has a positive relation with initial underpricing. The results indicate that more uncertainty about the intrinsic value of the firm results in greater underpricing and can imply that investors are compensated for increased risks, again in line with Carter, Dark, & Singh (1998). In combination with our findings on the relation between underpricing and underwriters’ reputation, the results strengthen our expectation for the existence of a certification effect. However, it should be noted that other factors could have amplified the positive relation between uncertainty and underpricing as well. Lui, Lu, Sherman & Zhang (2016) for instance finds that media coverage prior to the IPO is particularly positively related to underpricing when ex-ante uncertainty is high.
31 situations, despite the high reputation of an underwriter, uncertainty may not decrease because particularly uninformed investors are ambiguous how to interpret the signal given by reputable underwriters.
5.3.1.2. Long-run performance
Table VIII reports the cross-sectional regression results of BHAR on underwriters’ reputation, uncertainty 255-days, and the control variables. Results in column (1) show that the coefficient of underwriter reputation (median) is positive (0.131), but insignificant.
Underwriter reputation (median) gains significance at a 5% (10%) confidence level in
column (4), (column (7)) and (8), when control variables and the interaction term are added. A similar result is found for underwriter reputation (average) in column (2). Underwriter
reputation (average) does however not gain significance at any conventional level in
regressions (5), (8) and (10). Despite the predicted positive coefficients of underwriter reputation in all regressions, evidence for a significant positive effect on long-run abnormal returns is weak. Again it appears that the expected positive signaling effect in the premarket does not hold for longer holding periods. Interestingly, however, is the significantly negative coefficient of hot issue dummy. In line with the behavioral timing hypothesis of Loughran and Ritter (2000), there is a strong negative relation between long-run underperformance and firms that start floating during a hot issue period. This negative relation is also observable in figure I. A high number of firms went public in the years of the dot-com bubble or just prior to the financial crisis. It seems that issuers bring their firm to the market when stocks are overvalued. Eventually, firms’ returns will decline back to their intrinsic value, resulting in negative long-run returns. Whether a firm over- or underperforms seems to be more related to market timing than to underwriters’ reputation. In any case, the regression results do not confirm hypothesis II that predicts a positive relation between underwriter reputation and long-run abnormal returns.
Uncertainty 255-days is regressed on BHAR in column (3) with a negative (-2.670) and
