__________________________________________________________________________________
AN ANALYSIS OF THE EFFECTS OF VENTURE CAPITAL
BACKING AND UNDERWRITER’S REPUTATION ON
INFORMATION ASYMMETRY AT THE IPO LOCKUP
EXPIRATION DATE
BY
BALINT GULYAS
(106 212 29)
________________________________________________________
Bachelor Thesis
Faculty of Economics and Business
Specialization Finance and Organization
Supervisor: Guilherme de Oliveira (MSc)
Date: June 2016
Abstract
Lockup agreements prohibit insiders from selling their stocks immediately after the IPO. This clause is inserted to the underwriting contract to ensure the market that insiders act in the best interest of the company even after it went public. I study 1023 firms that went public between 2003 and 2014. I analyze how Venture Capital backing and underwriter reputation affects the cumulative abnormal returns at the unlock date. I find a statistically significant 3-day cumulative return of -0.7%. These returns are driven by asymmetric information between insiders and the market and the following revised expectations. Even though Venture support is useful in decreasing pre-IPO informational problems, I find that is negatively associated with returns at the unlock date. Underwriter reputation on the other hand does not seem to have an effect on these returns when controlling for firm specific variables.
Statement of originality
I, Balint Gulyas, herby declare to take full responsibility for the contents of this paper. I declare that the text presented in this document is original and that no sources other than those mentioned in the bibliography have been used in writing it. The University and the Faculty of Economics and Business is solely responsible for the supervision of the work.
Acknowledgements
I am grateful for many people, without whom this project would not have been possible. First of all, I am truly thankful for my supervisor Guilherme de Oliveira, who helped me overcome many difficulties and gave me lots of valuable comments on how to improve my final version. I thank Mark Dijkstra, who also provided valuable comments, and was always ready to answer my questions. I would like to thank my family for the continuous support throughout the last 3 years. They stood by me and supported me both financially and morally, without them I would have never been able to study abroad. I am extremely grateful for my Dad, my Mom and all my brothers and sisters. Also thanks to all my friends especially to Philipp.
Table of Contents
1.
INTRODUCTION... 1
A
.
R
ESEARCH QUESTION: ... 2
2.
MOTIVATION OF RESEARCH ... 2
3.
LITERATURE REVIEW ... 3
A
.
N
EGATIVE CUMULATIVE ABNORMAL RETURN: ... 3
B
.
A
SYMMETRIC INFORMATION,
LOCKUP PERIODS ANDV
ENTUREC
APITALS: ... 4
C
.
U
NDERWRITER REPUTATION AND ASYMMETRIC INFORMATION: ... 5
4.
HYPOTHESES ... 6
5.
DATA ... 7
6.
MODEL ... 8
7.
RESULTS ... 9
A.
E
VENT STUDY: ... 9
B.
M
ULTIPLE REGRESSION: ... 10
C
.
G
ROUPED MULTIPLE REGRESSIONS: ... 11
D
.
D
ISCUSSION: ... 12
E.
R
OBUSTNESS CHECK: ... 13
8.
CONCLUSION ... 14
A.
L
IMITATIONS: ... 14
BIBLIOGRAPHY: ... 16
APPENDIX ... 18
1. Introduction
Almost all initial public offerings (hereafter; IPO(s)) include a so called lockup periods. During this periods insiders, who are generally assumed to have superior information about a company’s business and future performance, are prohibited from liquidating their holdings. This provision is initiated by the underwriter of the IPO and it serves as a guarantee that the company and its management is focusing on long term value creation rather than short term profit. Newly public companies have little publicly available past financial information, thus their management have to convince the market that their business is sustainable and they have faith in it. The terms of the lockup and the expiration date, along with the percentage of “locked” shares are all public knowledge as they are disclosed in the IPO prospectus.
Despite the fact that most major financial news sites have a section dedicated to expiring lockups, most firms experience a statistically significant negative cumulative abnormal return (hereafter: CAR) at the unlock date. My results are consistent with this anomaly, I observed an average CAR of - 0.71%, which is significant at the 1% level, across all US Common Stock Offerings between 2003 and 2014. There is no consensus about the reasons behind the observed CARs, however the most widely cited explanation argues that the asymmetric information between insiders and outsiders is one of main drivers (Field and Hanka, 2001; see also Yung and Zender, 2010).
Asymmetric information arises when one party is better informed than another, thus the informed party can profit from the information difference (Aboody and Lev, 2000). In the context of lockups, insiders - pre-IPO shareholders - have more information on the business model, future potential or possible pitfalls, thus they are better informed than average investors. The existence of informational problems is recognized by most investors as a potential source of uncertainty, which can greatly influence firm value. Signaling is an important tool to lower uncertainty, thus investors try to observe and interpret every action and decision of the better informed party. In this paper I analyze the Cumulative Abnormal Returns around the lockup expiration date and I study the effect of positive pre IPO signals on them.
Venture Capital (hereafter: VC) backing and underwriter reputation are positive signals in the context of pre-IPO asymmetric information (Gompers and Lerner, 2001). However, my results reveal that firms backed by VCs experience significantly larger negative returns, thus are potentially more exposed to post-IPO informational issues than firms not backed by VCs. The large difference among CARs suggest that the market reacts to lockup expirations differently.
Moreover, the average CAR of the firms underwritten by renowned investment banks is negative and differs from firms taken public by less well known institutions. However, underwriter reputation does not have a significant effect on CARs when I controlled for firm and industry specific characteristics. Thus, following my terminology, underwriter reputation does not seem to have an effect on post-IPO informational problems at the lockup expiration date.
The rest of this paper is structured as follows; in the next subsection I specify my research question. In section 2 I motivate my research and then briefly discuss existing relevant academic literature in section 3. After specifying my hypothesis, I move on to my data analysis part, where I specify the research method in section 6. The results of my statistical analysis is summarized in section 7, which is followed by a brief discussion and my conclusion.
a. Research question:
It has been shown that underwriter reputation and the backing of a VC firm both can reduce pre-IPO informational issues; my paper aims to identify whether these signals are still valid post-pre-IPO. My research question reads as follows: Is Venture Capital backing and underwriter reputation decreasing post IPO information asymmetry, measured by abnormal returns at the lockup expiration date? I examine this relation focusing on all US Initial Common Stock Offerings between 2003 and 2014. 2. Motivation of research
The Efficient Market Hypothesis (hereafter: EMH) states that market participants use all available information to determine the value of a share (Nam, Park & Arthurs, 2014). There arethree forms of this theory based on what constitutes “all available information”. The semi-strong form of EMH, which is generally regarded as the most realistic form, assumes that expectations are formed using all publicly available information. Proponents argue that insider information is by nature private, therefore expectations and prices do not reflect them. Unlock dates are public knowledge, as such they should not imply the reformulation of existing expectations only if informational issues are involved. Assuming that the semi-strong form of the EMH holds, lockup expiration dates provide a good opportunity to study post IPO information asymmetry. Without informational issues trades are uninformative and simply the lockup date should not imply any decrease in value. Thus previous literature commonly attributes the observed negative returns to asymmetric information and the following market reaction to insider sales activity. However whereas some firms do not experience any share price drop, others suffer multi-day negative returns of even 31%1, thus there must be some signals available for analysis.
Lockup expirations represent a potential release of insider information and therefore they directly affect investors’ holdings. The unlock dates are listed on several financial news sites, for example in the Wall Street Journal and are included in most of the financial databases. The reasons for listing these dates are to warn and prepare shareholders of a potential release of insider knowledge. The behavior of the informed party is an important signal for investors. The semi-strong form of the EMH suggest that once insider trading activity is publicly available it will be priced in. However certain insider transactions, such as VC distributions, are not required to be reported to the SEC, thus are not immediately available at the unlock date.
Focusing on VC backing and high quality underwriters can also be interesting knowing that there are currently more than 160 private companies whose valuation exceed $1 Billion - so called “unicorns” - who are heavily linked to VCs (data from CBInsights). These large valuations imply larger amount of shares locked up and potentially more private information regarding the sustainability of their businesses. In fact, in the fast paced (tech) world many companies secured considerable investments only for their “traction”, without even producing a single cent of profit, which further increases uncertainty.
Previous studies often explain this apparent violation of the semi-strong form of the EMH by the existence of asymmetric information between insiders and investors (Field and Hanka, 2001, see also: Bradley, Jordan & Roten 2001; Brau, Carter, Christophe & Key 2004). These studies however used pre 2000’s data, when the IPO market was flourishing. During the dot-com bubble, which was fueled
1 Odimo Inc (Nasdaq: ODMO) lost 31% during the days surrounding the lockup expiration date. For more information, see Appendix 1.
by the booming IPO market, pre-IPO investors were eager to cash out as soon as possible (Ritter and Welch, 2002). Prior to the burst investment banks and the market in general were overly optimistic and investors largely ignored the worrying signals. After a collapse investors became more cautious and start searching for insider signals in every transaction (Michealy and Shaw, 1994).
Thus it is not obvious that same relations still hold, as markets tend to “correct” expectations after publications highlighted certain issues. Due to the strict regulations associated with public companies, most of the insiders’ private information is revealed over time. They have strong incentives of signaling that they have faith in the firms. However, when they sell large amounts the market will regard it as negative signal and as the demand for the shares decreases the price will fall. Thus besides the pre-IPO signals insiders can also signal sustainability if they keep their shares even after the lockup. This means that lockup periods, and especially the unlock dates, are relevant in the context of asymmetric information.
3. Literature review
a. Negative cumulative abnormal return:
Prior the 2000’s lockup periods did not receive much attention, however in 2001 both Field and Hanka and Bradley et al. examined the market reaction to IPO lockup expirations. They were studying different event windows (3 and 5-day window respectively), however they both found similar, statistically significant negative CARs at the expiration date. Field and Hanka (2001) identified an overall CAR around the unlock date of -1.5% and 40% rise in trading volume. Both of these effects were approximately 3 times higher among VC backed companies. Abnormal trading volume was a dependent variable of their studies, and they found that stock performance is positively associated with the number of trades. The relation was even stronger among VC backed firms, suggesting that VCs aggressively sell their shares at the earliest possible time, if price has risen. This result is consistent with Ferris et. al (1988) who concluded that investors in general are more likely to sell their holdings when stock price has risen. Here, it is important to note, that VCs are exempt from insider sales regulations, they are not required to report their intention to distribute their shares, thus their trading activity is not immediately observable.
Field and Hanka (2001) investigated several hypotheses that try to explain the observed price movement, but concluded that there is no unambiguous explanation. The price pressure theory specifies that on unlock date, due to the increased sell order by insiders, stock prices temporarily decrease, however evidence suggest that the price drop is not temporary thus this explanation was rejected (Field and Hanka, 2001). On the other hand, the downward sloping demand curves theory explain permanent price drops by the permanent increase of the number of shares available for the public, however Field and Hanka (2001) claim that the observed CARs are not exclusively a result of increased trading volume. They found significantly negative CARs in cases where the trading volume was under 1% of the pre-unlock average, thus concluding that the most probable explanation falls under asymmetric information and market expectations. Bradley et al. (2001) found a significant, -1.61% 5-day CAR around the unlock date using a sample of 2529 companies. They also examined the above mentioned theories and arrived to the same conclusion as Field and Hanka, citing information differences as a possible explanation.
Thus while the reasons of the observed CARs are not yet completely understood, there seem to be a consensus that the price movement is closely connected to information asymmetry problems (Field and Hanka, 2001, Bradley et. al. 2001, Brau et. al., 2005).
b. Asymmetric information, lockup periods and Venture Capitals:
As mentioned above, most of the previous studies stress the importance of asymmetric information as one of the main drivers behind the negative returns. Lockup agreements increase the price investors are willing to offer for multiple reasons. Firstly, the market expects that negative information held back at the IPO is likely to come out before the lockup expires (Ibbotson and Ritter, 1995). Moreover, until insiders keep their shares, their incentives are similar to outsider shareholders’ incentive, and this aligned interest lowers the risk of moral hazard problems. As Bradley et al. (2001) pointed out transactions by insiders have a significant impact on information asymmetry as the market tries to determine the sign of their private information based on their trading activity. Thus an increased trading volume is associated with a market belief that insiders are cashing out, thereby signaling poorer outlook. Adverse selection arises because insiders are better informed about a company’s actual and future business performance and therefore they can make better decisions. This issue is highly relevant among newly public firms, since there is only limited amount of past financial data available. The main source of adverse selection is that insiders generally are more aware of potential threats, weaknesses and competitors, thus they can evaluate all these before transactions (Chiang and Venkatesh, 1988). Therefore, lockup periods can be successful in decreasing adverse selection.
On the other hand, Brau et al. (2004) claim that lockup agreements cannot entirely diminish these issues, since the period is rather short. The apparent trend towards a 180-day lockup period only allows for maximum of two earnings announcements, and management can withhold negative information if they are considering liquidating their holdings at the unlock date (Brau et. al, 2004). Withheld information is rather negative, for the simple reason that management would want to publish all positive private information before they cash out to increase the value of their holdings. Thus around the expiry date agency costs increase, which in turn lowers the demand for stocks. However, since specific insider trading activity is not necessarily available at the unlock date, market participants have to form (rational) expectations about the severity of asymmetric information from other signals (Bradley et. al. 2001).
Lashinsky (1999) argued that unlock dates are rather “meaningless”, he claims that stocks with a normal volatility should not be affected by the expiration of these agreements. The observed negative abnormal returns, across multiple samples and timeframes, reject his claim, and give room for empirical research.
There is a highly technical discussion about CARs as a potential measure of asymmetric information by Dierkens (1991), who focuses on seasoned equity issues. She shows that larger asymmetric information is associated with a significantly larger price drop at equity issues, and conclude that information asymmetry and the subsequent market reaction is the main source of price fluctuations following equity issuance. Therefore, in my research I will interpret a negative coefficient as a factor that increases information asymmetry, thus a negative signal, and vice versa.
In the next paragraphs I briefly explain what VCs do and their role in decreasing pre-IPO asymmetric information. VCs are professionally managed, equity focused investment pools who support privately held companies both strategically and financially (Gompers and Lerner 2001). They provide funding to early stage firms with perceived, long-term growth potential and are critical for commercializing innovations when young companies start developing their products (Block and Sandner, 2009, Gompers and Lerner, 2001). Firms with high level of intangible assets, such as Technology or Healthcare companies, whose performance and value are difficult to assess, are
especially heavily relying on VC financing. There is a lack of collateral in these capital-intensive industries, due to the intangibles and the specialized intellectual property values, thus debt financing is costly or even unavailable2 (Aboody and Lev, 2000).
As VCs specialize in early-stage investments they have several tools to decrease agency problems, thereby allowing these companies to scale their product or service and potentially create a sustainable public company. As Gompers (1995) states, VCs particularly concentrate their investments in areas where agency costs are high, thus their control and monitoring function have the greatest value. Before any capital injection they scrutinize companies thoroughly and after the investment they closely monitor them. The most important tools of monitoring include spreading out capital injection over time, diversifying risks by partnering up with other VCs, capture seats on board of directors and aligning incentives by appropriate compensation agreements (Gompers and Lerner, 2001). However, Nam et al. (2014) found that VCs have a greater tendency to engage in earnings management, due to their short term focus.
A distinction must be made here; reputable VCs have more incentive of preserving their reputation, thus they serve an external audit function, thereby signaling lower level of asymmetric information to the market. The important signaling role of VCs, is also documented by Megginson and Weiss (1991) who found significant evidence that IPO underpricing is less severe among VC backed firms, arguing that it is due to the certification role of VCs.
On the other hand, Lee and Wahal (2004) arrived to significantly different conclusions by studying all US IPOs between 1980 and 2000 and correcting for selection bias, arising from the unequal distribution of industries due to the above mentioned reasons. They found that VC backed IPOs experience 5-10.3% higher first day returns, meaning that VC backed IPOs were more severely underpriced than non VC backed IPOs. These results opposed previous studies, however as their research method was more sophisticated, it seems to me more credible.
They explain their findings by Gompers’ “grandstanding theory”, which claims that younger VCs aim to build their reputation thus they have an incentive to take “their” companies public as soon as possible. This urgency is only applicable to young VCs, whose reputation is not yet well established and who are more willing to bear the costs associated with higher underpricing (Lee and Wahal, 2004). Gompers (1996) also found greater underpricing among companies taken public by younger VCs. In general, greater underpricing is associated with more uncertainty, thus the effect of VC backing on asymmetric information is ambiguous. Moreover, underpricing is more severe among firms taken public by less well established investment banks; this guides us to the next section.
c. Underwriter reputation and asymmetric information:
The role of underwriter certification in decreasing asymmetric information and adverse selection has been investigated by several previous researchers. On the one hand, investment banks are required by law to ensure that the offer price is fair (Securities Act of 1933), but they are also required to analyze and certify the information about their client. First of all, high quality investment banks are usually having higher standards of certification, which implies that they can produce superior due diligence about the firms they underwrite (Francis, Hasan, Lothian and Sun, 2010). As they stated, since investors use the investment bank’s reputation as a proxy for correct valuation, firms can signal
2 This unequal distribution is supported by my sample, see Appendix 2 for a distribution of VC backing across industries.
positive outlook and complete information by choosing a highly renowned underwriter. Thus reputable underwriters signal higher quality IPOs and less uncertainty.
This certification role is double sided, investment banks have an incentive of keeping their reputation, thus high quality investment banks perform thorough screening before taking a firm public (Beatty and Ritter, 1986). In fact, according to Carter et al. (1998) this screening capacity is responsible for the superior performance of issues backed by reputable investment banks. In the context VC distributions, Gompers and Lerner (1998) reported considerably less negative (higher) returns among firms underwritten by high quality underwriters. They also cited the profound pre-IPO screening as a potential explanation.
Lower underpricing is interpreted as better service; thus investment banks can obtain more clients by pricing the shares more fairly. Francis et al. (2010) found that firms with symmetric information (who has nothing to hide) select high quality underwriters in order to communicate positive signals to the market. Another research carried out by Michaely and Shaw (1994) concluded that reputable underwriters are successful in reducing information asymmetry not only between insiders and the market, but between general and institutional investors, which also leads to lower underpricing. Institutional investors carry out their own analyses, however by thorough screening, investment banks can eliminate this competitive advantage of the institutional investors.
Investment banks set lockup periods to further decrease uncertainties and adverse selection surrounding a new issue. Therefore, lockup periods are value enhancing. Underwriters are aware that a sudden selloff by insiders would signal negative outlook and could damage a freshly public firm (Brav and Gompers, 2001). Thus they include provisions to convince the market that they do not have to be prepared for such possibility. Even though lockup lengths were argued to signal the severity of informational issues, this signal largely lost its relevance, due to the apparent standardization.
4. Hypotheses
According to the semi-strong form of the EMH, prices fully reflect all publicly available information, and as mentioned above the unlock date, which is public knowledge, should not result in any share price movement. Thus investors form rational expectations based on all this information, and market prices should reflect these expectations (Bradley et al. 2001). No information asymmetry would mean that the CAR - on average - at the event window equals zero.
However, my overall sample reveals that the CAR on average is negative and significant, which suggest informational issues. In this paper I analyze the effect of pre-IPO positive signals on post-IPO asymmetric information. I am interested to see how VC backing and underwriter reputation affects the CARs at the unlock date. I prepare a double sided statistical test to test the coefficients of these variables. Double sided test is preferred as there is no consensus on the direction of these effects3. Given that the CARs are a result of asymmetric information, and these signals are still valid post-IPO then their coefficient should be positive. On the other hand, if they worsen asymmetric information their coefficient should be negative, whereas if they do not affect informational issues at the unlock date the betas should be zero. My hypotheses expressed mathematically are the following:
H0A ; βVC = 0 H1A ; βVC ≠ 0
3 A two‐sided test is more conservative than a one‐sided test because it requires a more extreme test statistic to reject the null hypothesis. Thus it is preferred in all cases unless there is a scientific reason to assume the direction of the difference.
and
H0B ; βBIG4= 0 H1B ; βBIG4 ≠ 0 5. Data
The initial sample of companies were obtained from Thomson ONE database. I looked up all US IPOs with an IPO date between 01/01/2003 and 31/12/2014, then I filtered all common stock offers. After eliminating all the irrelevant or erroneous entries my final sample consisted of 1023 companies4. After downloading my sample, I researched and computed all the variables, starting with the lockup periods. I followed Yung and Zender’s (2010) logic when there were multiple lockup expiration dates I took the management lockup as this is the one most closely connected to information asymmetry problems. In 63 cases, when management lockup was not listed, I took the “company lockup” date. I manually went through my sample and determined the lockup expiration date used for calculations. After setting the above mentioned dates I used a function (=days) to find the length of the lockup period.
There appears to be a significant standardization of lockup lengths at 180 days. It has been mentioned by older studies as well and my data further confirms this. The average lockup length is 189.8, but the median length is 180 in all subsamples (see table 3 in the appendix). The descriptive statistics of table 1 show significantly lower standard deviation of length among firms underwritten by the big four, or backed by VCs. Thus the standardization is even more apparent there.
I used the Venture Capital flag in Thomson ONE to identify the companies that were backed by VCs, in my final sample there are 382 VC backed firms, which amounts to approximately 37.3% of total. The distribution of VC backing across industries support the hypothesis that VCs concentrate their investment in certain sectors, especially to Healthcare and High Technology. These industries suffer more from informational problems due to the potentially high level uncertainties hence their heavy reliance on VCs. Please see appendix 2 for the distribution of VC backed IPOs across industries.
Similarly to previous studies, I measured underwriter reputation by market share. Field and Hanka (2001) regarded Morgan Stanley, Goldman Sachs and Merrill Lynch as the big three and classified them as reputable investment banks. However, the investment banking landscape changed a lot since their research. Thus I prepared a test to identify which investment banks had significant market share as lead bookrunner in my sample. As expected Citigroup gained significant market share by being lead in 116 IPOs. Therefore, I modified the BIG3 to BIG4 and assigned the dummies accordingly. 49.5% (506 firms) of my sample was underwritten by one of the big four investment banks. See appendix 4 for the distribution of underwriters over years.
For industries I use the “TF macro descriptions” obtained from Thomson ONE, I assigned dummies to fix the industry effects. Industry dummies are present and needed to control for industry specific characteristics. See table 2 in the appendix for a distribution of IPOs across industries.
Similarly, to previous research I control for firm size, I calculated the total market capitalization after the IPO. Shares outstanding and prices were obtained from CRSP database but due to some mismatch of the identification numbers and erroneous entries I could only obtain data for 888 firms. In order to exclude some of the “IPO underpricing effect” I multiplied shares outstanding by the first
4 I identified 1895 IPOs during the period of interest, then similarly to previous studies I eliminated all entries without an indicated lockup provision and all penny stock offerings, with offer price below $5. After allowing only common and ordinary share codes and eliminating all incorrect entries and missing values my final sample size shrank to 1023 firms.
day closing price instead of the IPO offer price. IPO underpricing is a well known (and heavily researched5) phenomenon, which states that new issues are usually priced below their fair value. The reasons for underpricing include liquidity and uncertainty issues, however it is only a temporary effect as market mechanism drive the prices towards their intrinsic value. IPO underpricing would slightly bias market capitalization values, so it is desirable to use first day closing price to counteract this issue. Average market capitalization equals $669 million, but values range from $20 million to $24 billion. Due to the magnitude of these numbers I take the natural logarithm of the values when controlling for firm size. This is in line with the research practice and it limits the range to 16.8-23.9.
I also calculated the stock price performance from the IPO until 10 days before the unlock date. I use this value to control for additional firm specific effects. There were significant differences across groups. As expected VC backed companies in general were doing better, their average performance was almost twice as big as the average calculated over the whole sample (10.6% versus 5.4% respectively). Moreover, firms underwritten by a leading investment bank also outperformed the overall sample mean, these companies averaged an 8.1% growth. In fact, firms who received VC support and were underwritten by large investment banks achieved an average return of 16.4% throughout this period. Please see table 1 in the appendix for the detailed summary statistics.
6. Model
I test my hypotheses by setting up regressions, for more robust results I will add new variables gradually to my model and see how the coefficients behave. My proposed multivariate model is the following:
𝐶𝐴𝑅
𝑖= 𝛽
0+ 𝛽
1𝑉𝐶
𝑖+ 𝛽
2𝐵𝐼𝐺4
𝑖+ 𝛽
3𝑙𝑛(𝑆𝑖𝑧𝑒)
𝑖+ 𝛽
4𝐿𝑒𝑛𝑔𝑡ℎ
𝑖+ 𝛽
5𝑃𝑒𝑟𝑓
𝑖+ 𝛽
6−16𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑖𝑒𝑠
𝑖+ 𝜀
𝑖The dependent variable is the multi-day Cumulative Abnormal Return (CAR), measured in a 3-day window, from t = -1 until t = 1, where t = 0 is the lockup expiration date. Abnormal returns are estimated using the Capital Asset Pricing Model (CAPM). βCAPM was estimated for each company during a 70-day estimation window until 10 days before lockup expiration.
𝐴𝑅
𝑖= 𝑟
𝑖− 𝐸(𝑟
𝑖)
where𝐸(𝑟
𝑖) = 𝛼
𝑖+ 𝑟
𝑓+ 𝛽
𝐶𝐴𝑃𝑀𝑖(𝑟
𝑚𝑘𝑡− 𝑟
𝑓)
The CARs are the sum of the abnormal returns during the event window [t=-1; t=1] and are presented in the next section. After calculating the dependent variables, my main explanatory variables are venture capital backing (VC) and underwriter reputation (BIG4). Both of these variables are dummies, and the sign of their coefficient will tell me how they are related to CARs and thus to informational problems at the unlock date.
The remaining variables, which are selected based on previous studies, will be control variables and are explained below. Ln(Size) is the total market capitalization of a firm following the IPO, calculated as first day closing price times shares outstanding. As I mentioned above, first day closing prices were used instead of offer price to exclude some of the underpricing effect.
Length refers to the lockup period in days, using the management lockup period for reasons
mentioned above. Perf refers to the stock price performance from IPO until 10 days before the lockup
expiration dates. This variable is important for controlling firm specific noise. I am controlling for industry characteristics by using dummy variables for each but one industry. The base industry was Consumer Staples.
Regarding the methodology, I prepare a standard event study followed by a multiple regression model. After careful investigation I decided to study a 3-day window, starting 1-day prior the lockup expiration day and finishes 1 day after. This shorter event window is preferred to limit the effects and noise of other events thus focus only on the lockup expirations.
7. Results a. Event study:
First of all, I had to run a standard event study to find the CARs of each firm, where the event date (t=0) is the lockup expiration date. The abnormal returns were calculated using the CAPM with pre-event implied beta and CRSP6 value weighted returns as market return. When preparing an event study a longer estimation window is preferred, as it yields more accurate estimates. However, there is a trade-off due to the varying lockup lengths. As there are some shorter lockup periods it is necessary to choose a shorter window to only include pre-event data. Bradley et al. (2001) points out that it is undesirable to use the whole pre-unlock period as an estimation window due to the post IPO underwriter stabilization effect. This refers to the practice that underwriters propose to purchase shares on the secondary market at the offer price, to stabilize prices, which could potentially influence expected returns. Therefore, when setting up the event study I used an estimation window of 70 days and a minimum of 10 trading days gap similarly to a Bradley et. al (2001). My event study was not sensitive to the choice of estimation method, see robustness check for more details.
To assess the significance of the average cumulative abnormal return (CAAR) I used the standardized residual test as described by Patell (1976). The average AR on the unlock date is -0.16% which is significant at the 1% level. The average AR one day following the unlock date (t=1) is -0.46% which is similarly significant at the 1% level. The average CAR of the 3-day window across the whole sample equals -0.71% with a Patell z-value of -4.2. This implies that the negative CAAR is also significant at the 1% level. Most of the previous research found similar abnormal returns, thus there is overwhelming evidence that firms experience negative abnormal returns at the unlock date (Field and Hanka, 2001., Bradley et. al 2001., Yung and Zender 2010.)
Some studies try to explain these results with the price pressure effect or the bid-ask bounce, mostly resulting from increased trading volume. These theories are discussed in detail by Field and Hanka (2001); their main prediction is that prices should rise to original level after the drop experienced at the unlock date. They proposed a test to assess the validity of these theories and concluded that none of them explain the observed CARs. I prepared the same test in my sample by setting up an event study for a 27-day period, starting 6 days before the event data and finishes 21 days after. As we can see from Chart 3 in Appendix 5, CARs do not really normalize after the expiration date, but stay below 0. In fact, there seems to be a long-lasting price drop which is almost exclusively materialized around the unlock date. These results are consistent with the results of Field and Hanka (2001), thus I conclude that the CARs are not explained by the above mentioned theories.
6
CRSP (Center for Research in Security Prices) is a provider of historical stock market information. It is available through Wharton Research Data Services. The CRSP value weighted return represents the value-weighted market portfolio and it is the most commonly used market portfolio for research purposes.
My results imply that lockup expirations are, in general, followed by negative abnormal returns. An average decline of less than 1% does not really seem to be an issue, however it has been shown that certain characteristics can influence this price drop significantly. Thus in next sub-section I will present my multiple regression model (explained in section 6) to see how VC backing and underwriter reputation affects information asymmetry at the unlock date.
b. Multiple regression:
Along with research practice I gradually added new control variables to my models to see how the coefficients change, and obtain more robust results. I started my regressions with Model 1, where I only regressed the VC dummy on the CARs and added the above explained variables one-by-one. I refer to these models as “ungrouped”, because here I ran the test on my whole sample without any classifications. My initial sample consisted of 1023 companies, however market capitalization and stock price performance, which I calculated using CRSP data, were not available for 135 firms due to mismatch between Thomson ONE and CRSP identifiers. Thus in model 5, 6 and 7 the sample size is smaller, only 888 firms. Regression output can be found in the appendix, please see table 4.
The most straightforward conclusion from these regressions is that the coefficient of VC backing is negative and significant at the 1% level in every model. This coefficient tells us that given two almost identical firms, the one backed by a venture capitalist is on average experiencing a 2-2.6% lower Cumulative Abnormal Return during a 3-day period surrounding the unlock date. In model 7, which is the most detailed regression VC backing had a coefficient of -0.025 with p-value equals 0.005.
I also found a somewhat significant coefficient for the variable BIG4 in models 2, 3 and 5. However, despite being negative and significant at the 10% level in those models, it loses its significance when controlling for industry and/or stock performance. In model 7 the coefficient of BIG4 equals -0.006 with a p-value of 0.18. This result is insignificant, thus observing the whole dataset, it seems that choosing a highly renowned underwriter does not affect the CARs if I control for firm specific variables.
Another interesting pattern revealed by the ungrouped multiple regressions is that stock price performance is negatively associated with CARs at the unlock date. As model 6 and 7 predicts, companies with better performing stocks are expected to experience a 2.2-2.4% drop in cumulative returns. These results are significant at the 5% level and could partly be explained by increased sell orders lowering the price.
When lockups expire, pre-IPO shareholders can finally sell their shares, and this increased selling volume could drive down prices. The increased sell orders can be a result of portfolio rebalancing by pre-IPO investors or simply the fact that they want to cash out, given that the price of their shares increased since the IPO. I did not measure insider trading activity in my paper, however Francis et. al (1988) found a positive relation between insider trading activity and stock price performance. They concluded that that investors are more likely to cash out (sell) when prices have risen. I cannot test this relation in my dataset, however my results are consistent with this prediction, thus when the share price of a company has risen since the IPO, one can expect lower (more negative) cumulative abnormal returns around the unlock date.
c. Grouped multiple regressions:
The CARs differ significantly across subsamples, so I grouped my sample based on my explanatory variables and repeated the full multiple regression (model 7). I refer to these models as “grouped” regressions. First I divided my sample in two subsamples based on whether a firm received VC support or not. As I am using the complete model, my samples are limited to the 888 companies for the remainder of my paper.
The average CAR of the ungrouped sample (888 firms) is 0.702%, which is significant (t = -2.78) at the 5% level. After grouping the data, the CAAR of the VC backed group equals -2.1%, which is significant at the 1% level (t = -3.73). On the other hand, the non VC backed subsample has an average cumulative abnormal return equals 0.15% which is not significantly different from zero at any level (t = 0.69). Please see table 5 in the appendix for CAARs and table 6 for the regression outputs.
The model on the VC backed subsample explain the variance of the abnormal returns considerably better than any previous model, as measured by its doubled R-squared value (R2 = 0.11). The observed negative coefficient of stock price performance is considerably larger in absolute value. It has almost doubled comparing to the ungrouped model with a coefficient equals -0.04 (with a p-value of 0.001). This means that the CAR of a venture backed company is expected to be lower as the stock performance increases, whereas in a non VC backed company there is not such a relation. It is possible that some of the problems related to asymmetric information is indeed solved by VCs but not all of them.
On the other hand, this difference might be explained by VC distribution policies. Venture capitals usually distribute shares among limited partners during the first year from the IPO. And the lockup expiration date represents the first possible distribution (Gompers and Lerner, 1998). Once partners have the shares they are more eager to sell if price has risen, which put a downward pressure on the price (Francis et. al 1988). Moreover, it can signal negative outlook for the future and the market can reformulate expectation which further lowers the price, resulting in more negative CARs.
The coefficient of underwriter reputation remains insignificant in both groups, suggesting that it does not have an effect on abnormal returns even in the grouped samples. However, when calculating the average CAR of firms underwritten by reputed investment banks there seems to be a large difference. In the next section I discuss the results of a grouped regression based on underwriter reputation. See table 7 in the appendix for reference.
The average cumulative abnormal return is -0.38% and not significant (t = -0.88) among companies not underwritten by the BIG4. On the other hand, companies underwritten by Morgan Stanley, Goldman Sachs, Merrill Lynch or Citigroup experience an average CAR of -1.02%, which is significant at the 1% level (t = -3.81). These results are inconsistent with the expectation that choosing a high quality underwriter helps lowering the magnitude of CARs (and asymmetric information) at the unlock date. The summary of the regression outputs of these two groups can be found in table 8 in the appendix.
The most striking result of these grouped regressions is that among companies not underwritten by highly renowned investment banks VC backing does not have a significant effect on the CAR. Its coefficient equals -0.018, with a p-value of 0.21. Despite being (negative as in any other cases), it is not significantly different from zero at any level. The other variables are not really different than in any other model, stock price performance is also only significant in the BIG4 subsample.
In the non BIG4 subsample 43% of the companies are supported by venture capitalists. Thus one would expect to see some negative effect of VC backing on the cumulative abnormal returns given all the other results. In the next section I will briefly discuss some existing theories that could help understanding how VC backing could be negatively associated with abnormal returns at the unlock date. Unfortunately, I do not have available data to test the relevance of these theories, however it worth mentioning them as they might partly explain the observed outcome and are closely related to my paper.
d. Discussion:
Given the above highlighted results it seems that VC backing does have an effect on CARs. The coefficients of the dummy is negative and significant in all cases except in the subsample of firms not underwritten by leading investment banks. Thus it can be concluded that VCs are negatively associated with abnormal returns around the unlock date. Therefore, my first hypothesis is rejected, meaning VC backing seem to have a negative association with post IPO information asymmetry as measured by the CARs around the unlock date.
There are some possible explanations why VC support seems to be a negative signal. When limited partners start selling their shares the market might expect that they are satisfied with the returns they made and they assume it is the best time to cash out. When lockups expire and VCs exit, their certification role “disappears” and the positive signals of quality and/or monitoring cease to exist. Outside investors and VCs had their interest closely aligned, both of them aimed to maximize share price, however when VCs exit, the interests of outsiders are not safeguarded anymore by the monitoring role of VCs. Thus the higher value that the market attached to a firm due to VC participation simply vanishes when VCs are free to trade their shares. This can explain why VC backing has a negative relation with abnormal returns at the unlock date in general. On the other hand, according to Bradley et. al (2001), it is unlikely that the market repeatedly fails to anticipate transactions undertaken by insiders. They argue that since lockup expirations are public knowledge the market should anticipate insider sales and prices should reflect it already before the unlock dates. They claim that the observed results are not exclusively due to insider trading activity, since VCs for example are not required to disclose their transaction activity, thus the market cannot react immediately.
Nonetheless VC backing and lockup expirations can have an impact on expectations. VCs are not required to notify the SEC days before the transaction as they simply distribute the shares among limited partners (Gompers and Lerner, 1988; see also: Bradley et. al 2001). Then it is up to the partners whether they want to sell their shares or not. However, the market reaction is very similar to the response to public announcements of secondary stock sales. When VCs distribute their shares the market views it similarly to the sale of shares by insiders. Hence the negative association.
However, this negative relation does not mean causality. Companies do not experience price drops because they have been backed by VCs. As I explained in section 3, VCs concentrate their investments in certain sectors, thus it is very difficult to assess the effect of VC backing due to the lack of comparability. But even if this support does not cause negative abnormal returns, it is apparent that venture backed companies are more exposed to negative CARs at the unlock date.
On the other hand, as the last regression shows, VC backed companies underwritten by not highly reputed underwriters are not associated with such negative returns. A possible explanation could be draw upon Gompers’ “grandstanding hypothesis”.
The grandstanding hypothesis states that young VCs have an incentive of taking firms public earlier, to establish reputation and access more funds (Gompers, 1996). He claims that young VCs are very sensitive to capital and the best way to signal their quality is by successful exits. In fact, they are willing to sacrifice some of their returns to gain reputation. Rushing with an IPO is costly for the VC due to more severe underpricing, which in fact stems from the reduced certification role of the VC. By taking a firm public earlier, VCs have less control and monitoring over the company; among others their tenure at the board is shorter and thus some of the certification role disappears. Moreover, young VCs use less well known investment banks, all these factors lead to higher information asymmetry comparing to older, more established VCs.
Thus it is possible that younger VCs do not reduce informational problems as much as older VCs, therefore at the unlock date there is smaller adjustments in value. Large VCs and BIG4 are most likely associated to each other, due to their repeated interactions. In fact, it is possible that large and well established VCs even have an agreement with leading investment banks. If it is indeed the case, this theory could explain why I observed negative coefficient of VC backing only in the BIG4 subsample. Unfortunately, I do not have data on the VC of each IPO, thus I cannot test this theory in this paper, however it could provide interesting opportunity for further research.
Underwriter quality on the other hand does not seem to affect abnormal returns post IPO. Thus my second hypothesis is not rejected; the dummy variable of underwriter quality does not influence information asymmetry at the unlock date. The market most likely incorporated all positive signs associated with the certification of a leading investment bank at the time of the IPO. When the lockup agreement expires the underwriter is not involved anymore in the transactions, therefore the market expectations are unchanged irrespective of the identity of the underwriter. In fact, in the next subsection, I briefly discuss why underwriter reputation might have lost its significance when I controlled for firm specific noise.
e. Robustness check:
My event study was not sensitive to the choice of estimation method. I repeated the calculations using the Fama-French three factor model and the Market-Adjusted Model as well. The former method is comparable to the CAPM, however it is less relevant for newly public companies. The results were very close to that of the CAPM. The latter estimation method yields less correct estimation as the betas of the firms are not estimated with a regression, but assumed to be 1. Thus it calculates excess returns as actual return less the CRSP value-weighted market return. I performed these tests solely to assess the sensitivity of my results. The average and standard deviation of the abnormal returns were similar, thus I proceeded with the CAPM model as it has been shown to be more successful in predicting returns.
Moreover, my ungrouped regression coefficients did not change significantly when my sample size shrank due to the mismatch between identifiers. In fact, when I added the control variables the coefficients of VC backing did not change significantly.
On the other hand, it is possible that reputed underwriters are slightly negatively associated with asymmetric information, but this association is captured by industry or performance control variables. Most importantly stock price performance could capture some of the relation. Firms underwritten by the big four have an average of 8.15% return from IPO until 10 days before the lockup expiration, whereas other firms only reach an average of 2.65%. The standard deviation of the performance is smaller as well, thus high quality investment banks are associated with better performing stocks. The
coefficient of performance is negative and significant in all cases. Thus it might be possible that some of the underwriter’s effect is captured by the stock performance variable. Nonetheless the low and barely significant coefficient in model 1, 2, 3 and 4 still suggest that even if there is an effect it is very little.
8. Conclusion
Venture Capital backing and underwriter reputation has been shown to be effective in decreasing pre-IPO information asymmetry. In this paper I aimed to study the effects of these variables on post-IPO informational issues. I wanted to study how a newly public company’s share price moves when the so called lockup agreement expires and insiders are allowed to sell their shares. Assuming symmetric information, there is not supposed to be any significant abnormal return around these days. However, I (and previous studies) observed a statistically significant, negative three-day cumulative abnormal return around the unlock date, which is commonly attributed to informational problems. Thus the lockup expiration date provides a good opportunity to study post-IPO asymmetric information. I looked at an initial sample of 1023 US companies (decreased to 888 in more complex models) that had an IPO between 2003 and 2014. Despite the overall significance of CARs, there are large differences among subgroups, meaning not every company experience similar price drops. To answer my research question I ran several regressions.
According to my results Venture Capital support is negatively associated with abnormal returns. Thus, instead of a positive signal it is a negative one in the context of post-IPO asymmetric information. Even though VC backing successfully lowers adverse selection prior to the initial public offering, it is associated with larger (more negative) abnormal returns around the lockup expiration date. Possible explanations include VC distribution policies and the loss of monitoring, which in turn increases agency costs. The market attached higher value of VC backed shares due to the monitoring role of VCs, however when they exit their investments this monitoring role disappears. This can result in revised expectations, which in turn lowers the value of the shares.
Underwriter quality, on the other hand does not seem to affect CARs in the whole sample. Even though firms with high quality underwriters experience 2.7 times lower abnormal returns, the coefficient of underwriter quality loses its significance when controlling for firm specific variables. Thus it does not decrease, nor increase post-IPO asymmetric information problems at the lockup expiration date.
a. Limitations:
On the other hand, it must be noted that some of the negative returns might in fact due to increased trading volume and thus increased supply of shares. The demand curve hypothesis states that if supply increases prices will go down. Field and Hanka (2001) also examined abnormal trading volume at the unlock date and they found a significant increase in average abnormal trading volume. Despite this average effect, they found significantly negative returns when trading volume was way below the pre-event average. Thus they concluded that it does not entirely explain the price drop. In case of VCs they found even higher abnormal trading volume. They prepared a detailed analysis of abnormal trading volume as a determinant of abnormal returns, however they concluded that it alone cannot entirely explain the negative returns.
I was not able to incorporate abnormal trading volume to my analysis, even though it might increase the power of my model. Thus it represents a possible improvement for further research.
Moreover, identifying the VCs who backed each company and categorizing them by their age or reputation could also reveal some interesting results. One might find supporting evidence that grandstanding is indeed responsible for the insignificant coefficient of VC backing among companies not underwritten by well-known investment banks. Further research is needed to investigate this hypothesis.
Bibliography:
Aboody, D., & Lev, B. (2000). Information Asymmetry, R&D, and Insider Gains. The Journal of Finance,
55(6), 2747-2766.
Beatty, R. P., & Ritter, J. R. (1986). Investment banking, reputation, and the underpricing of initial public offerings. Journal of Financial Economics, 15(1-2), 213-232.
Block, J., & Sandner, P. (2009). What is the effect of the financial crisis on venture capital financing? Empirical evidence from US Internet start-ups. Venture Capital, 11(4), 295-309.
Bradley, D. J., Jordan, B. D., Yi, H., & Roten, I. C. (2001). Venture Capital And Ipo Lockup Expiration: An Empirical Analysis. Journal of Financial Research, 24(4), 465-493.
Brau, J. C., Carter, D. A., Christophe, S. E., & Key, K. G. (2004). Market reaction to the expiration of IPO lockup provisions. Managerial Finance, 30(1), 75-91.
Brav, A., & Gompers, P. A. (1999). Insider Trading Subsequent to Initial Public Offerings: Evidence from Expirations of Lock-Up Provisions. SSRN Electronic Journal SSRN Journal.
Carter, R. B., Dark, F. H., & Singh, A. K. (1998). Underwriter Reputation, Initial Returns, and the Long-Run Performance of IPO Stocks. The Journal of Finance, 53(1), 285-311.
Chiang, R., & Venkatesh, P. C. (1988). Insider holdings and perceptions of information asymmetry: A note. The journal of finance, 43(4), 1041-1048.
Dierkens, N. (1991). Information Asymmetry and Equity Issues. The Journal of Financial and Quantitative
Analysis, 26(2), 181.
Field, L. C., & Hanka, G. (2001). The Expiration of IPO Share Lockups. The Journal of Finance, 56(2), 471-500.
Francis, B. B., Hasan, I., Lothian, J. R., & Sun, X. (2010). The Signaling Hypothesis Revisited: Evidence from Foreign IPOs. Journal of Financial and Quantitative Analysis, 45(01)
Ferris, S., Haugen, R., & Makhija, A. (1988). Predicting contemporary volume with historic volume at differential price levels: Evidence supporting the disposition effect, Journal of Finance 43(1) Gompers, P. A. (1995). Optimal Investment, Monitoring, and the Staging of Venture Capital. The Journal
of Finance, 50(5), 1461-1489.
Gompers, P. A. (1996). Grandstanding in the venture capital industry. Journal of Financial Economics,
42(1), 133-156.
Gompers, P. A., & Lerner, J. (1998). Venture capital distributions: Short-run and long-run reactions,
Journal of Finance 53(1)
Gompers, P. A., & Lerner, J. (2001). The Venture Capital Revolution. Journal of Economic Perspectives,
Ibbotson, R. G., & Ritter, J. R. (1995). Initial public offerings. Handbooks in operations research and
management science, 9, 993-1016.
Lee, P. M., & Wahal, S. (2004). Grandstanding, certification and the underpricing of venture capital backed IPOs. Journal of Financial Economics, 73(2), 375-407.
Ljungqvist, A., & Wilhelm, W. (2002), IPO pricing in the dot-com bubble. Journal of Finance, 58(2), 723-752
Megginson, W. L., & Weiss, K. A. (1991). Venture Capitalist Certification in Initial Public Offerings. The
Journal of Finance, 46(3), 879.
Michaely, R., & Shaw, W. H. (1994). The Pricing of Initial Public Offerings: Tests of Adverse-Selection and Signaling Theories. Review of Financial Studies, 7(2), 279-319.
Mishkin, F. S., (1990), Asymmetric Information and Financial Crises: A Historical Perspective,
Financial Markets and Financial Crises by R. Glenn Hubbard, University of Chicago Press: ISBN: 0-226-35588-8, obtained via: http://www.nber.org/books/glen91-1
Nam, D., Park, H. D., & Arthurs, J. D. (2014). Looking Attractive until You Sell: Earnings Management, Lockup Expiration, and Venture Capitalists. Journal of Management Studies, 51(8), 1286-1310. Patell, J. M. (1976). Corporate Forecasts of Earnings Per Share and Stock Price Behavior: Empirical Test.
Journal of Accounting Research, 14(2), 246.
Ritter, J. R., & Welch, I. (2002). A Review of IPO Activity, Pricing, and Allocations. The Journal of Finance,
57(4), 1795-1828.
Yung, C., & Zender, J. F. (2010). Moral hazard, asymmetric information and IPO lockups. Journal of
Appendix
1. Odimo Inc (footnote 1):
Odimo Inc. (Nasdaq: ODMO) is a VC backed online retailer of luxury goods such as watches and jewelry. They lost about 40% of their value between the IPO and the unlock date, which was followed by a 31% price drop around the unlock date. Their share closed at $4.63 one day before the unlock date but it closed at $3.2 one day after, which is a 30.88% loss. There was not any extraordinary announcement, thus this loss is most likely due to market reaction (adjustment) to the expiring lockup provision, fueled by uncertainty, which in fact stems from asymmetric information.
Date Adjusted closing price ($)
Aug 12 4.63
Aug 15 (lockup expiration date) 3.10
Aug 16 3.20
2. Summary Statistics:
3. Distribution of VC backing across industries (footnote 2):
Chart 1:
VC backing across industries
My sample confirms that VCs invest in industries with high uncertainty as their “certification role” is the most valuable there. ~84% of all VC support were in the High Technology and Healthcare industry:
4. Lockup length:
Table 3:
5. Underwriter reputation:
Chart 2:
IPOs underwritten by the BIG 4 per year (Morgan Stanley, Merrill Lynch, Goldman Sachs or Citigroup) Interesting to see the effect of the Financial Crisis not only on the amount of IPOs but also on smaller investment banks, they lost most of their market share, and only recovered in 2014.
Morgan Stanley had the largest market share by deal volume, they were lead bookrunners in 151 IPOs. Merrill Lynch followed them by 135, whereas Citigroup accounted for 116 IPOs. Goldman Sachs was lead in 104 cases, thus overall the big four accounted for 506 IPOs. This is 49.46% of my final sample.
6. Chart 3 plots the CARs over a 27-day period, starting 6 days prior the unlock date until 21 days after:
Chart 3:
Mean and 95% confidence interval of Cumulative Abnormal Returns over a 27-day window
Field and Hanka (2001) discussed a few possible theories that could possibly explain the observed returns. The price pressure and the bid-ask effect are commonly cited explanations in the context of seasoned equity offerings or stock splits, they both predict a temporary price drop. They explain this drop by the increased sales executed at the bid. Field and Hanka (2001) prepared a test and rejected this explanation; they found a parallel fall at the unlock date in the bid and ask price as well. In fact Chart 3 shows that this price drop seems to be permanent rather than temporary in my sample, thus my results are consistent with their.
7. Ungrouped regression outputs (model 1, 2, 3, 4, 5, 6, 7):
Table 4:
Standard Errors are presented in parentheses underneath the coefficients in all of my regression outputs.
8. CAAR in the VC backed vs non VC backed groups:
Table 5:
9. Summary of regression outputs of VC backed vs non VC backed subsamples:
10. CAAR in the non BIG4 vs BIG4 groups
Table 7:
11. Summary of regression outputs of non BIG4 vs BIG4 subsamples:
