Initial Public Offerings on the London Stock Exchange

Home , Initial public offering, Public offering

A thesis submitted to The University of Manchester for the degree of Doctor of Philosophy in the Faculty of Humanities

2014

Dimitris Kostas

Manchester Business School

Table of Contents Abstract 4 Declaration 5 Copyright statement 6 Dedication 7 Acknowledgements 8

Chapter 1 Introduction 1 Motivation 9 2 Institutional framework 13 3 Thesis overview and contributions 14 4 Thesis structure 17 References 18

Chapter 2 The Use of Warrants in Non-Underwritten Initial Public Offerings Abstract 20 1 Introduction 21 2 Literature review & hypotheses 23 3 The Alternative Investment Market 29 3.1 The role of the broker 30 3.2 Broker’s compensation structure 31 4 Data 32 4.1 Pricing of warrants 33 5 Methodology 35 6 Empirical results 44 6.1 Descriptive statistics - univariate analysis 44 6.2 Warrant characteristics 45 6.3 Test of the cost minimisation hypothesis 46 6.3.1 Probit regression model 46 6.3.2 Second stage regression estimates 47 6.3.3 Robustness tests 50 7 Conclusion 51 References 53

Chapter 3 Warrants in Underwritten IPOs Abstract 68 1 Introduction 69 2 Literature review 71 Non-cash and broker/underwriter compensation regulations between the UK 3 72 and the US markets 4 Hypotheses 74 5 Data 76 6 Methodology 77 7 Results 84 7.1 Descriptive statistics - univariate analysis 84

7.2 Warrant characteristics 87 7.3 Test of the cost minimisation hypothesis 88 7.4 Robustness tests 92 8 Conclusion 94 References 96

Chapter 4 The IPO when-issued market Abstract 110 1 Introduction 111 2 Literature review & hypotheses 113 3 When-issued dealing 117 Differences of the when-issued dealing among the UK and other developed 3.1 119 markets 4 Data 122 5 Methodology 123 5.1 When-issued market price accuracy 123 Determinants of the when-issued market and how the conditional trading 5.2 124 affects the offer price? 5.3 Volume in the first day of trading 131 6 Results 132 6.1 Descriptive statistics 132 6.2 Univariate analysis 133 6.3 Informational accuracy of the when-issued market 136 6.3.1 The when-issued market facilitates price formation 138 The when-issued market allows investors the earliest opportunity to 6.3.2 139 agree an entry or exit price 6.4 Probit regressions 140 6.5 Does the decision to have a when-issued market affect the offer price? 142 6.6 Volume in the first day of trading 144 6.7 Robustness tests 145 6.7.1 Heckman two stage model 145 6.7.2 Volume OLS regression 146 7 Conclusion 146 References 148

Chapter 5 Conclusion 1 Summary of results 165 2 Future research 167 References 170

This thesis contains 60088 words including title page, tables, and footnotes.

Abstract The University of Manchester Dimitris Kostas Doctor of Philosophy (PhD) Initial Public Offerings on the London Stock Exchange April 2014

This thesis examines the non-cash compensation paid to the underwriters/brokers during the flotation process and the IPO when-issued dealing market in one of the most successful and international stock exchanges around the world, the London Stock Exchange (LSE). The thesis consists of three essays that try to answer the following questions: Do IPO firms minimise their costs of going public by issuing warrants to their financial advisers? Does the when-issued dealing affect the setting of the offer price? The first essay examines the issue of warrants to brokers as part of their compensation package in non-underwritten offerings on the Alternative Investment Market of the LSE. The main finding is that IPO firms are able to make efficient decisions and choose the contract that minimises their costs. For companies that issue warrants to their brokers the total costs of going public are 22.74% (as a percentage of gross proceeds), but would have been 25.61% had they not issued them. This 2.87% reduction in costs is equivalent to 70.34% of the commission paid to the brokers by the IPO firms. The main source of this decrease in the costs is the lower underpricing the companies incur by granting warrants to their brokers. The second essay examines the use of non-cash compensation in underwritten IPOs. The findings suggest that firms that are cash constrained are more likely to issue warrants to their underwriters. In addition, underwriters appear to have the ability to time the issue of warrants because they include them as part of their compensation package when the market is doing well. Interestingly, warrant issuers are still able to minimise their costs of going public even under a very light regulatory setting underlying the use of non-cash compensation. The third essay examines the when-issued dealing in the Main Market of the LSE for an extensive period of time, 1996 to 2012. The main finding is that, in an institutional setting in which the when-issued dealing commences only after the allocation of shares and the offer price are announced, investors pay ‘rents’ to the underwriters in order to acquire IPO shares that will trade within the when-issued dealing. These ‘rents’ take the form of a higher offer price. In other words the when-issued dealing affects the setting of the offer price. For companies that have a when issued dealing the offer price is £3.4 but would have been 54% lower (£1.55) had these firms not had a when issued dealing.

Declaration

I, Dimitris Kostas, declare that no portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trade marks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=487), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.manchester.ac.uk/library/aboutus/regulations) and in The University’s policy on Presentation of Theses

Dedication

This thesis is dedicated to my family

Acknowledgements

I am very grateful to my supervisors, Prof. Arif Khurshed and Mr. Brahim Saadouni, who gave me the opportunity and freedom to conduct my research in the area of my interest. Their continuous encouragement and guidance motivated me throughout the PhD programme to always try to improve my work.

I would like to thank my PhD examiners Dr. Susanne Espenlaub and Prof. Ranko Jelic for their valuable comments and feedback during my viva. I would also like to thank other committee members, university staff and colleagues, such as Dr. Konstantinos Stathopoulos, Prof. Stuart Hyde, Dr. Maria Marchica, Dr. Ning Gao, Prof. Norman Strong, Dr. George Christodoulakis, Prof. Ser-Huang Poon, Prof. Richard Stapleton, Dr. Edward Lee, Dr. Marie Dutordoir, Dr. Roberto Mura, Dr. Abdulkadir Mohamed and Dr. Shuxing Yin.

In addition, I would like to thank all my friends in Manchester, and especially Zhe Wen and Huixin Zhang, who provided me with unforgettable memories of joy and laughter.

Furthermore, I would like to thank the UK regulator, the Financial Conduct Authority, who invited me to present my research findings. Finally, I would like to thank the participants of the 15th Annual Conference on Money and Finance, 2013 "Merton H. Miller" Doctoral Seminar and World Finance and Banking Symposium.

Chapter 1

Introduction

1 Motivation

Recently there has been mounting shareholders’ concern related to the fees paid to intermediaries during the flotation process. The Institutional Investor Committee (IIC)1, reports that many issuers were reluctant to challenge the total underwriting fees paid to their advisers because they were not familiar with the equity capital raising process (Institutional Investor Committee, 2011). One of the proposals of the IIC was that the boards of the companies should not only require a full breakdown of the fees paid to their advisers, but also ensure that they understand the purpose of each part of the underwriting fees that is paid for. In addition, PricewaterhouseCoopers’s (PwC) survey (PricewaterhouseCoopers, 2012) documented that 48% of the chief financial officers of the companies that have gone public said that the costs of an initial public offering (IPO) were higher that what they initially expected. According to PwC one of the main factors that affects the cost of an IPO is the underwriting fees.

Further to the ongoing debate on underwriting services and fees, Dunbar (1995) reports that IPO firms are able to minimise their total costs of going public through the issue of non-cash compensation to financial intermediaries. In the first two essays of this thesis I examine whether the IPO firms make efficient decisions and choose the contract that is more favourable for them. More specifically, do IPO firms minimise their costs of going public through the issue of warrants to their underwriters/advisers? What are the reasons that may lead underwriters to also include non-cash compensation in their compensation package and not get paid all their fees only in cash?

Furthermore, many developed markets around the world have a grey/when- issued/conditional dealing/market for IPOs. The when-issued market is the trading that takes place on a security a few days prior to its official admission on the stock exchange. All the transactions occurring during the when-issued market are conditional on the fact that the security will be admitted to trading on the stock exchange. If the security is not listed then all the when-issued dealing trades will be void. The when-issued market has received a lot

1 The IIC is a group of trade associations that represents institutional investors and consists of the Association of British Insurers, the Investment Management Association and the National Association of Pension Funds.

9 of attention in The Financial Times (2012) as to whether it is beneficial or risky for investors because the liquidity within this market may be invariably low. In addition, previous academic literature has reported that underwriters pay informational ‘rents’ to investors, which take the form of the allocation of underpriced IPOs shares. This is due to the fact that the latter provide underwriters with truthful information about the value of the IPO, through their orders during the subscription period (Benveniste and Spindt, 1989). However, Aussenegg et al. (2006) document that when there exists an IPO when-issued market, which takes place simultaneously with the subscription period, financial intermediaries do not pay any ‘rents’ to investors. The reason is that their private information is revealed to the underwriters through their trades in the when-issued market. But, apart from the underwriters, is there any case in which investors also have to pay ‘rents’ in order to buy the IPO shares? Is liquidity low or high within the when-issued market? Is the when-issued market beneficial for investors and how does a well-regulated when-issued market differs from an over the counter grey market in terms of the companies that it attracts. In chapter 4 I try to find answers to all the aforementioned questions.

My thesis is focused on the London Stock Exchange (LSE) because it is considered one of the most successful stock exchanges around the world as more than half of the IPOs taking place on it trade above their offer price a year after their flotation. When compared to New York and Hong Kong Stock Exchanges, LSE has the largest percentage of IPOs trading above their issue price since the financial crisis (London Stock Exchange, 2012a). In addition, LSE is one of the most international markets around the world because over a third of the IPOs conducted on it come from overseas. Moreover, LSE’s IPO institutional framework is very distinct from that of other developed markets.

More specifically, my first and second thesis chapters examine the issue of warrants to financial advisers on the Alternative Investment Market (AIM) of the LSE. The warrants issued to financial intermediaries are different from those issued by unit IPOs (IPOs that issue both shares and warrants). One of the differences is that financial advisers’ warrants are issued only to them, whereas unit IPO warrants are issued to all investors (Chemmanur and Fulghieri, 1997, Schultz, 1993). In addition, the warrants issued to financial advisers constitute part of their compensation package for advising a company to conduct an IPO, whereas this is not the case with unit IPO warrants. Furthermore, according to Schultz (1993, p. 206, Table 2) almost all unit IPOs issue warrants to their underwriters, while only 39.7% of the share only IPOs issue warrants to their investment banks as part of their compensation

10 package. Due to the fact that unit IPOs have very different characteristics when compared to share only IPOs (Schultz, 1993), as they are smaller, younger and riskier, and based on the previous literature (Barry et al., 1991, Dunbar, 1995) I exclude unit offerings from my data.

The main reason that I study the use of compensation warrants on AIM is that approximately one third of the total IPOs make use of the non-cash compensation. The popularity of the use of compensation warrants on AIM may be explained by the fact that it has a light regulatory environment. AIM’s regulatory set-up has been specifically designed to meet the needs of smaller/growing companies and give them access to investors and equity capital at an early stage of their development where a listing on the Main Market is not yet appropriate (London Stock Exchange, 2007). For instance, no trading record and prescribed level of shares to be in the hands of the public are required for companies that are listed on AIM. Barry et al. (1991) and Dunbar (1995) report that it is usually smaller and younger companies that issue warrants to their underwriters. This may also explain the very limited use of warrants on the Main Market (less than 4%) and this is why I exclude it from my sample. The importance of the AIM market is evident from the fact that it is Europe’s largest market in terms of number of companies (London Stock Exchange, 2007). From 1995 to 2009, nearly half of all IPOs, conducted in major European stock markets took place on AIM (Vismara et al, 2012). In addition, AIM’s institutional setting is unique and very different from that of the US, as it carries almost no regulations underlying the use of non-cash compensation.

My first thesis chapter is focused only on non-underwritten offerings whereas the second one on underwritten IPOs listed on AIM. The main difference between underwritten and non-underwritten offerings is that the financial advisers have to buy any unsold IPO shares if the issue is underwritten. If it is not, then the financial advisers only have to do their best endeavours to procure investors for the IPO shares. Apart from the differences in the underwriting obligation and the fact that underwritten and non-underwritten IPOs have very different characteristics in terms of issue size, market capitalisation, etc., I cannot conduct an analysis on both of them simultaneously because the vast majority of non-underwritten offerings do not disclose any accounting data. This restricts me from testing new hypothesis that have never been examined in the previous literature. For instance, if I want to examine whether IPO firms issue warrants to their underwriters because they are cash constrained,

11 then I can only test this in the underwritten offerings sample, as most of the non-underwritten IPOs do not report the cash available the year prior to their listing in their prospectuses. But most importantly, I cannot run a probit model in which the dependent variable takes the value of one if the IPO firms issue warrants and zero otherwise and also include as an independent variable another dummy which takes the value of one if the IPO is underwritten and zero otherwise. The reason is that the decision to underwrite an offering may be a non- random one and financial intermediaries may self-select the contract (underwrite or not) that is more favourable for them. Hamilton and Nickerson (2003) report that firms rarely make decisions randomly. So, endogeneity may be an issue and this is why it is important to study each group of IPOs individually (Golubov et al., 2012, p. 307).

On AIM there exist two advisers, the Nominated adviser (Nomad), which is the ‘regulator’ of the company as it ensures that the firm satisfies all the criteria set by the AIM rules, and the broker, which is the party standing between the investors and the firm (London Stock Exchange, 2007). When the LSE launched the AIM market to attract smaller and younger companies, it realised that a lot of these smaller firms do not have an experienced management team in running a public company. This is why the Exchange decided to assign the responsibility for ongoing regulation to a new financial adviser, the Nomad. Its responsibility is to determine if the company is suitable for admission on AIM and to provide ongoing advice on regulatory matters. If the firm does not have a Nomad then it is effectively unregulated and, based on the AIM rules, its shares should be suspended and eventually its admission on AIM will be cancelled (London Stock Exchange, 2007).

In my thesis I only focus on the warrants issued to the brokers as they are essentially the underwriters in the case of underwritten IPOs and the advisers responsible for the fundraising in the case of non-underwritten offerings. More specifically, the broker is responsible for organising the roadshows, procuring investors for the company’s shares, building the book, collecting the funds from investors and allocating the shares. The broker can act as a market maker for the company’s shares and play a pivotal role for the IPO firms post admission (London Stock Exchange, 2007).

The following example illustrates the different components of the issue costs that an IPO firm pays for the underwriting services/advice a broker provides at the time of listing. SRS Technology Group was listed on the Alternative Investment Market on 20 August 2001. The

12 company offered a total of 2,800,000 new shares at an offer price of £1.25 per share. The offering was underwritten by Numis Securities Limited (broker). The company paid an underwriting commission of £105,000 (3% of gross proceeds). In addition, the broker was granted with warrants which can be converted into 446,079 shares at an exercise price of £1.25. The warrants can be exercised anytime within the first 2 years starting from day one of listing. Given a risk free rate of 5%, time to maturity of 2 years, first day closing price of £1.275, total number of shares after listing of 14,869,300, and a standard deviation of the FTSE AIM All Share Index of 0.74%, the total value of the warrants is £ 64,215.2 As a result, the broker enhances its fee by 61.2% (64,215/105,000) by including warrants in its compensation package. Moreover, the total broker compensation (commission plus value of warrants) is approximately 4.83% of the gross proceeds.

In my third thesis chapter I focus only on the Main Market of the LSE because the percentage of IPOs that have a conditional trading on AIM is very small. In addition, the number of companies that conduct an IPO through a when-issued dealing on the Main Market has increased substantially as this percentage was only 6.56% in 1996 and soared to more than 50% from 2002 onwards. Moreover, the UK’s when-issued dealing has a very different institutional setting from that of other major developed markets around the world, such as that of Hong Kong, Germany or France. Also, to the best of my knowledge, there is no other paper that has focused on the when-issued dealing of the Main Market over an extensive period of time (1996-2012). Interestingly, the volume in the first day of the when issued market is very high, 19.74%, as a percentage of shares outstanding, and then drops substantially. Some of the questions that I will try to answer in the fourth chapter are the following: Does the existence of the when-issued dealing affect the setting of the offer price and the volume in the first day of trading? Is the when issued dealing beneficial for IPO firms and investors?

2 Institutional framework

The institutional setting underlying the use of non-cash compensation is very different between the UK and the US markets. In the US the use of warrants is constrained by a number of regulatory requirements related to the exercise price, lock-in period and minimum value of warrants, whereas LSE carries none of these regulatory constraints. More

2 The warrants are valued using the Cox constant elasticity variance model as described in Barry et al. (1991).

13 specifically, in the US stock exchanges warrants have a lock-in period of 180 days, must be exercised within five years and must have a minimum value of 0.2% of gross proceeds for 1% amount of warrants (Financial Industry Regulatory Authority or FINRA3 Rule 5110). However, LSE has none of these requirements. Dunbar (1995) finds that in the US market underwriters do not force firms to use costlier compensation contracts. This is why he suggests that FINRA should relax some of its requirements underlying the non-cash compensation as they are unnecessarily restrictive. But, in an institutional setting in which there are almost no regulations related to the non-cash compensation, do IPO firms still minimise their costs of going public?

As far as the IPO when-issued market is concerned the UK’s institutional setting is very different from that of other developed markets. More specifically, the UK’s when-issued market is regulated by the LSE and commences only when the allocation of shares is completed and the issue price is announced. However, in other markets, such those of Hong Kong, Germany, France, Spain, Switzerland, etc. the grey market is conducted over the counter and takes place simultaneously with the subscription period as the issue price is announced the last day of the when-issued market (Cornelli et al., 2006). Companies listed on the LSE should satisfy certain requirements in order to have a when-issued dealing, such as there must be sufficient liquidity and demand for the security during the conditional trading period. But, this is not the case in other developed over the counter/grey markets. The existing empirical literature has reported that in non-regulated when-issued markets it is usually smaller companies that have a conditional trading (Dong, 2009). But, does this still hold in the case of a well regulated market, such as that of the UK? Are the characteristics of companies that have a conditional trading, which is regulated by the stock exchange, different than those in which it is not? Furthermore, does the when issue market affects the setting of the offer price and the liquidity of the securities?

3 Thesis overview and contributions

In this thesis there are three different essays on the IPOs taking place on the LSE. In my first empirical chapter I examine the issue of warrants to brokers in non-underwritten IPOs on AIM. The contribution of this chapter is twofold. The first one is that it is the only paper that studies the use of non-cash compensation in a group of IPOs that has never been examined before, the non-underwritten offerings. The second one is that it sheds light and contributes

3 FINRA is the largest independent regulator for all securities firms in the US.

14 to the ongoing debate related to the dissatisfaction of investors and companies with equity underwriting services provided by the financial intermediaries.

More specifically, my main finding is that IPO firms are able to make efficient decisions as they choose the contract that minimises their costs of going public. The total costs (underpricing and broker compensation) the IPOs incur (expressed as a percentage of gross proceeds) in order to obtain a listing on AIM are 22.74%, but would have been 25.61% had these firms not issued warrants to their advisers. This 2.87% reduction in costs is equivalent to 70.34% of the commission the IPO firms pay to their brokers. This decrease in the costs is mainly caused by the lower underpricing the IPO companies incur due to the issue of warrants. The actual underpricing is 15.95%, but would have been 22.12% had the firms not issued warrants. The rationale for this is that companies can credibly signal to the market that they do not sell overpriced securities by issuing non-cash compensation to their brokers. Warrants align the brokers’ compensation with the aftermarket performance of the stock price. If the price of the shares increases in the aftermarket then the value of warrants will also increase, leading to a higher broker compensation. Due to the certification through warrants, IPO investors will require a lower discounting of the IPO price.

In the second essay I examine the use of non-cash compensation in underwritten offerings on the AIM market. This essay makes three important contributions to the existing literature. It is the first one that studies the issue of compensation warrants in a market outside the US. This is very important because the AIM market has a very different institutional setting when compared to that of the US.

The second contribution is that I examine and find supportive evidence of hypotheses that have never been tested in the existing literature. More specifically, I find that companies which are cash constrained have a higher probability of issuing warrants. In addition, brokers include warrants in their compensation package when the market is doing well. This is consistent with the argument that brokers appear to be able to time the issuance of warrants. If the market return is positive then probably the company’s stock price will increase, leading to an increase in the value of warrants and consequently the compensation of the brokers. Moreover, I find that underwritten offerings that issue warrants are usually advised by reputable brokers. This is in stark contrast to Dunbar (1995) who reports exactly the opposite. Furthermore, the value of warrants is equivalent to 75% (2.4%/3.2%) of the commission the companies pay to brokers. As a result brokers enhance their compensation fees by 75%.

Finally, my results have important policy implications. I find that the total costs the IPO firms incur during the flotation are 28.9% and would have been almost double, 43.96% had they not issued warrants. Dunbar (1995) also finds similar results for US firm commitment offerings. He reports that the US regulations underlying the use of non-cash compensation are unnecessarily restrictive and should be relaxed. My findings further show that even in an environment in which there are almost no regulations related to the use of compensation warrants, IPO firms are still able to choose the contract that minimises their total costs of going public. As a result, the non-cash compensation regulations on AIM should not be restricted as the IPO companies achieve a 15.06% (43.96%-28.9%) reduction in costs via the use of compensation warrants.

In the third essay I study the conditional trading on the Main Market of the LSE which has a very different regulatory setting from that of other developed markets.

The contribution of the third essay is threefold. First, I shed light on a highly under- researched when-issued market of one of the major stock exchanges in the world, the LSE. Second, I study what are the determinants of a regulated when-issued market. To the best of my knowledge, my thesis chapter is the first paper that examines the determinants of the conditional trading. My findings suggest that companies that have a when-issued market have very different characteristics when compared to those that do not. Firms that are larger, less risky, with higher future growth opportunities and underwritten by more reputable underwriters are more likely to have a when-issued market. However, this is not the case in the most active European grey market, which is the German one (Cornelli et al., 2006), in which the characteristics of the companies that have a grey market are quite similar to those that do not. In fact, the companies that have a conditional trading in Germany are slightly smaller than those that do not (Dorn, 2009). One potential explanation for that may be the fact that the when-issued market in the UK is regulated by the LSE and has to follow certain rules, whereas that in Germany takes place over the counter. In addition, the when-issued market affects the volume in the first day of trading. For companies that have a when-issued dealing the volume of trading will approximately be 14% higher than those that do not.

The third contribution of the aforementioned essay is that my results have important policy implications for market regulators as the decision to have a when-issued market affects the setting of the offer price. More specifically, for companies that have a conditional trading the actual offer price is £3.4 but would have been 54% lower (£1.55) had these firms not had

16 a when-issued market. So, investors actually pay a ‘rent’, that takes the form of a higher offer price in order to acquire the company’s shares and would have paid a lower price if this company did not have a when-issued market. The reason is that the conditional trading has informational value for investors and offers them certain advantages, such as price formation before the commencement of the unconditional/aftermarket trading and earliest entry/exit price (London Stock Exchange, 2012b). Furthermore, for companies that do not have a when-issued market the actual offer price is £1.7 but would have been £3.14 had they had a conditional trading. So, under the LSE’s when-issued market institutional setting the Securities and Exchange Commission’s (SEC’s) argument, according to which the grey market for IPOs is not allowed in US stock exchanges because it will lead to a lower offer price, does not hold.

4 Thesis structure

The chapters in this thesis are self-contained as each one of them has its own literature review, answers unique questions and uses different datasets. The tables, figures and equations are independent and are numbered from the beginning of each chapter.

This thesis is structured as follows. Chapter 1 includes the introduction, Chapters 2 and 3 examine the use of non-cash compensation in non-underwritten and underwritten IPOs, Chapter 4 studies the when-issued market and whether the decision to have a conditional trading affects the setting of the issue price and Chapter 5 concludes.

From Chapter 2 to Chapter 4, I use the third person (we) instead of the first (I), because these chapters are working papers co-authored with my supervisors.

References

Aussenegg, W., Pichler, P., Stomper, A., 2006. IPO pricing with bookbuilding and a when- issued market. Journal of Financial and Quantitative Analysis 41, 829.

Barry, C. B., Muscarella, C. J., Vetsuypens, M. R., 1991. Underwriter warrants, underwriter compensation, and the costs of going public. Journal of Financial Economics, 29, 113-135.

Benveniste, L. M., Spindt P. A., 1989. How investment bankers determine the offer price and allocation of new issues. Journal of Financial Economics 24, 343-361.

Chemmanur, T. J., Fulghieri, P., 1997. Why include warrants is new equity issues? A theory of unit IPOs. Journal of Financial and Quantitative Analysis, 32, 1-24.

Cornelli, F., Goldreich D., Ljungqvist A., 2006. Investor sentiment and pre-ipo markets. The Journal of Finance 61, 1187-1216.

Dorn, D., 2009. Does sentiment drive the retail demand for ipos? Journal of Financial and Quantitative Analysis 44, 85.

Dunbar, C. G., 1995. The use of warrants as underwriter compensation in initial public offerings. Journal of Financial Economics 38, 59-78.

Golubov, A., Petmezas, D., Travlos, N. G., 2012. When It Pays to Pay Your Investment Banker: New Evidence on the Role of Financial Advisors in M&As. The Journal of Finance 1, 271-311.

Hamilton, B. H., Nickelson, J. A., 2003. Correcting for Endogeneity in Strategic Management Research. Strategic Organisation 1, 51-78.

Institutional Investor Committee, 2011. Best Practice Guidance for Issuers when Raising Equity Capital.

London Stock Exchange, 2012a. Perspectives on the global markets.

London Stock Exchange, 2012b. When Issued Dealing. http://www.londonstockexchange.com/traders-and-brokers/rules-regulations/when-issued- dealing-guidance.pdf

London Stock Exchange, 2007. Joining AIM: A Professional Handbook. http://www.faegre.com/files/22472_Joining%20AIM%202.pdf

PricewaterhouseCoopers, 2012. Considering an IPO? The costs of going and being public may surprise you. http://www.pwc.com/en_US/us/transaction- services/publications/assets/pwc-cost-of-ipo.pdf

Schultz, P. 1993. Unit initial public offerings: a form of staged financing. Journal of Financial Economics 34, 199-229.

The Financial Times, 2012. IPOs: outlook brightens for grey market.

Vismara, S., Paleari, S., Ritter, J. R., 2012. Europe's Second Markets for Small Companies. European Financial Management 18, 352-388.

Chapter 2

The Use of Warrants in Non-Underwritten Initial Public Offerings

Abstract

We study the issue of warrants to the brokers as part of their compensation package in non- underwritten offerings. We find that IPO firms that use compensation warrants minimise their costs of going public as the actual costs are 22.74%, but would have been 25.61% had they not issued these warrants. This 2.87% reduction in costs is equivalent to approximately 70.34% of the commission paid to the brokers by the IPO firms. In addition, warrant issuers are smaller, riskier and are advised by less reputable brokers.

1 Introduction

Recently there has been a lot of debate in the market related to the fees paid to financial intermediaries during an Initial Public Offering (IPO). According to the Institutional Investor Committee (IIC)4, a large number of firms that raise capital do not question the compensation paid to the financial advisers simply because they are not familiar with the equity capital raising process (Institutional Investor Committee, 2011). Furthermore, PricewaterhouseCoopers (PwC) reports that almost half of the Chief Financial Officers (CFOs) said that the costs of conducting an IPO exceeded their initial expectations (PricewaterhouseCoopers, 2012). According to PwC’s study underwriting fees were one of the main factors that affected the costs of an IPO. In this paper we examine whether IPO firms minimise their costs of going public through the use of non-cash compensation (i.e. warrants). More specifically, we try to answer the following question: Do IPO firms make efficient decisions and choose the contract that minimises their flotation costs?

Our study is focused only on non-underwritten offerings on the Alternative Investment Market (AIM) of the London Stock Exchange (LSE) for four important reasons. First, to the best of our knowledge, this is the only paper that examines the use of non-cash compensation in non-underwritten IPOs as the existing empirical literature has been focused on the issue of warrants in firm commitment offerings in the United States (US). But, firm-commitment offerings and non-underwritten IPOs are totally different. In firm-commitment offerings the underwriters have the obligation to buy the unsold IPO shares (underwritten IPOs), whereas in non-underwritten IPOs the financial advisers will only do their best endeavours to procure investors for these shares. So, the incentive to use non-cash compensation may be different between firm commitment and non-underwritten IPOs. Second, the non-underwritten offerings on AIM constitute approximately 70% of the total IPOs. Third, the vast majority of non-underwritten offerings on AIM are placings, which means that the potential investors of these offerings are institutional (informed) investors. However, in the US market best effort offerings are usually purchased by individual (retail) investors (Ritter, 1987). Finally, 43.42% of the non-underwritten offerings issue warrants to their financial advisers. This is in stark contrast with the US market in which almost all best-efforts5 offerings issue warrants

4 The IIC is a group of trade associations that represent institutional investors and consists of the Association of British Insurers, the Investment Management Association and the National Association of Pension Funds. 5 Although the best efforts offerings in the US may have differences with the non-underwritten IPOs in the UK, they share one important characteristic; in both IPO groups the financial intermediaries will only do their best efforts to sell the IPO shares, but are under no obligation to buy the unsold shares (non-underwritten).

21 to their investment banks (Dunbar, 1995). So, the AIM market enables us to conduct a unique analysis on the use of warrants in non-underwritten offerings which cannot be done in the US market.

We only include AIM IPOs in our sample, and not those listed on the Main Market of the LSE because almost all Main Market IPOs are underwritten by their sponsors. In addition, AIM is considered Europe’s largest market by number of companies and from 2004 onwards there were more listings on AIM than on Nasdaq (Gerakos et al., 2011). AIM is also a very popular destination for international companies as from 2005 onwards more than 20% of the AIM IPOs were incorporated outside the United Kingdom (UK). According to Vismara et al. (2012), nearly half of all IPOs listed on major European stock markets took place on AIM.

We examine the issue of warrants only to the brokers (and not the Nomads) because the brokers are responsible for procuring subscribers/buyers for the company’s new/selling shares, organise the roadshows, conduct the bookbuilding, organise the settlement of the issue and determine the final offer price, in collaboration with the company’s management, based on the level of investors’ interest and their evaluation of the market (Burton et al., 2006, London Stock Exchange, 2007c). So, the brokers play a very important role in the flotation process as they are responsible for the fundraising.

Our study contributes to the existing empirical literature in two different ways. The first one is that it sheds light in the use of non-cash compensation by a group of IPOs that has never been examined before, the non-underwritten offerings. The second one is that it provides insights and contributes to the ongoing debate on the dissatisfaction of companies and institutional investors with equity underwriting services in the UK, especially the increase in the underwriting fees (Office of Fair Trading, 2011).6 While the Office of Fair Trading (OFT) report is focused on the financial intermediaries’ compensation in rights issues, we provide direct evidence of the financial advisers’ fees on initial public offerings.

Our sample consists of IPOs that took place during the period June 1995, the initiation of the AIM market, until December 2008. Our main finding is that companies that issue warrants are able to make efficient decisions and choose the contract that minimises their

6 An investigation by the Office of Fair Trading (UK) into the underwriting fees of rights issues in the UK found that there had been a significant increase in the fees since the onset of the current financial crisis. The average fee increased to 3% in 2009 from 2% to 2.5% over 2003-2007. IPO underwriting fees were excluded from the investigation because they were considered a significantly different type of transaction involving different types of underwriting risk.

22 costs. More specifically, the total costs of going public for the IPOs that issue warrants to their brokers are 22.74%, but would have been 25.61% had they not issued them. Although, a 2.87% (25.61% - 22.74%) reduction in costs, through the use of warrants, appears to be a small percentage, it is approximately equivalent to 70.34% of the commission (2.87%/4.08%) paid to the broker(s) by the company. So, the IPO firms are able to reduce their flotation costs significantly by issuing warrants to their financial advisers. Dunbar (1995) also finds that IPO firms in US minimise their costs of going public through the use of non-cash compensation and the reduction in costs is equal to 3.8%. But Dunbar’s (1995) study is focused only on US firm commitment IPOs.

This 2.87% decrease in the total costs is mainly caused by the fact that the IPO firms incur a lower underpricing; the actual underpricing is 15.95%, but would have been 22.12% had the firms not issued warrants. This is due to the fact that through the use of non-cash compensation, companies can credibly signal that the IPO is not overpriced and investors will require a lower discounting of the IPO price. Furthermore, the probability of issuing warrants is higher for smaller and riskier offerings which are advised by less reputable brokers.

The remainder of this paper is organised as follows. Part 2 summarises the literature review and explains our hypotheses, part 3 provides background information for the AIM market, the role of the broker and its compensation structure, part 4 describes the data and pricing of warrants, part 5 includes the methodology, part 6 presents the empirical results and part 7 concludes.

2 Literature review & hypotheses

The existing empirical literature is mainly focused on the use of warrant as compensation to the underwriters in firm commitment offerings in the US market. There are three main hypotheses that have been suggested in the literature in order to explain the issuance of compensation warrants to underwriters: the certification hypothesis (Booth and Smith, 1986), the monopsony power hypothesis (Dunbar, 1995, US Securities and Exchange Commission, 1963a, 1963b, as cited in Ng and Smith, 1996) and the circumvention hypothesis (Barry et al., 1991). According to the certification hypothesis, underwriters include warrants as part of their compensation in order to certify that the issue is not overpriced. The monopsony power hypothesis argues that underwriters have a monopsony power over issuing firms and ask for warrant-based compensation. The alternative of the

23 aforementioned hypothesis is the cost minimisation, according to which IPO firms choose the contract that minimises their costs of going public. According to the circumvention hypothesis, underwriters use warrants as a way to avoid the maximum compensation guidelines set by the National Association of Securities Dealers (NASD). The circumvention hypothesis cannot be tested within the context of the AIM market as there are no regulatory requirements that set maximum or minimum limits on the brokers’ compensation.

Ritter (1987) compares best efforts and firm commitment offerings and finds that only 35.4% of the IPOs in the US are best efforts offerings. He reports that the cost of going public (investment banking fees and underpricing) is much lower for firm commitment offerings when compared to those of best efforts offerings. This is due to the fact that firm commitment offerings have substantial economies of scales, higher sales and book value, than best efforts offerings. Furthermore, best efforts offerings are mainly bought from individual investors whereas firm commitment offerings from institutional. Ritter (1987) also reports that firms that have higher aftermarket volatility, usually use best efforts offering contracts to get listed. In addition, he finds that firm commitment offerings are taken for listing by prestigious underwriters. Out of 364 best efforts offering IPO contracts, only one was advised by a prestigious investment bank during the period 1977 to 1982.

There is also an extensive literature on unit IPOs that issue warrants to the investors. Schultz (1993) reports that unit IPOs, in comparison with share only IPOs, are multi-stage financing arrangements that reduce the agency cost of free cash flow (Jensen, 1986). More specifically, unit IPOs bond the managers to optimal investment decisions. This is due to the fact that managers receive only the portion of cash that is required to fund the company’s projects, thus mitigating the free cash flow problem. If managers invest in successful/value-creating projects then the company’s stock price will increase enough (above the warrants exercise price) to allow the exercise of warrants which will lead to additional funds for the company (second stage financing). If not, and the project has a negative Net Present Value, then the stock price will stay below the warrant exercise price and the second round of financing will not occur (no exercise of warrants).

Schultz (1993) also reports that unit IPOs are usually issued by companies that are younger, smaller and riskier as they face a high level of uncertainty related to their growth opportunities. In addition, companies with lower levels of managerial ownership post-IPO will have greater agency costs and will be more likely to issue unit IPOs. Furthermore, given

24 the younger age and smaller size of the firms that issue warrants, unit IPOs will experience greater underpricing when compared to share only IPOs. Moreover, unit IPOs will receive additional financing (exercise of warrants) only if their investment projects are profitable and the stock price will increase. As a result unit IPOs with negative NPV projects will be less likely to survive as they will not receive these additional funds.

Chemmanur and Fulghieri (1997) extend the model of Leland and Pyle (1977) and propose a signalling model which requires the interaction of asymmetric information and manager’s risk aversion. Insiders have private information about the riskiness and the expected value (quality) of the firm. According to their model high quality firms (high expected future cash flows) distinguish themselves from low quality firms by mainly using three different signals; the underpricing, issue of warrants and percentage of equity retained by insiders. Risk averse insiders of high quality firms will choose the signal that will maximise their utility. All three aforementioned signals are costly for the company. However, warrants may be preferred because insiders will share the firm value with warrant holders only if warrants are exercised.

The signalling hypothesis provides several testable empirical predictions. After controlling for retained ownership, the proportion of the firm value sold in the form of warrants will increase with the riskiness of the firm. In addition, the percentage of underpricing and the fraction of equity retained by insiders have a positive and negative relationship with the riskiness of the firm. Furthermore, unit IPOs are likely to take place when the information asymmetry and the risk of failure are high.

Although some of the predictions of Chemmanur and Fulghieri (1997) are similar to those of Schultz (1993), such as unit IPOs are riskier than shared only IPOs, the former predicts that, after controlling for retained ownership, the proportion of the firm value sold in the form of warrants will increase with the level of the IPO firm riskiness, whereas Schultz (1993) makes no prediction.

Mazouz et al. (2008) use Hong Kong IPO data to examine whether firms issue warrants to investors (unit IPOs) in order to reduce the agency cost of free cashflow (Schultz, 1993) or to signal their higher quality (Chemmanur and Fulghieri, 1997). Overall their findings support the signalling hypothesis as the profitability and asset utilisation rates are higher for IPOs that issue warrants when compared to those that do not. Furthermore, Mazouz et al. (2008) use a self-selection model and find that the underpricing is lower for IPOs that issue warrants than what it would have been if they had not issued them.

How and Howe (2001) and Lee et al. (2003) compare whether the agency cost or the signalling hypotheses can explain the issue of warrants to investors during an IPO on the Australian Stock Exchange, which has a different institutional setting from that of the US. Their findings support the signalling hypothesis, because firms use warrants as a signalling mechanism in an environment which is characterised by informational asymmetry. In addition, both Lee et al. (2003) and How and Howe (2001) report that there is no significant difference in the underpricing between unit and share only IPOs.

Garner and Marshall (2005) examine Chemmanur and Fulghieri’s (1997) predictions related to the signalling model and find that the level of riskiness is not uniform across all unit IPOs. More specifically, they report that companies with longer warrant exercise periods and a lower ratio of shares to warrants are riskier. This is due to the fact that the market associates shorter exercise periods and a lower proportion of warrants with lower levels of risk. Furthermore, Garner and Marshall (2005) provide evidence that there is a trade-off among the signalling mechanisms (proportion of firm value sold as warrants, fraction of equity retained by insiders and underpricing) reported in Chemmanur and Fulghieri (1997).

Byoun and Moore (2003) test the agency cost/sequential financing and signalling hypotheses in seasoned equity offerings (SEOs). They find that warrants can signal a firm’s future prospects. Furthermore, SEOs that issue warrants (units) experience an average abnormal return of -1.96%, but would have been -2.71% had they not issued them. In contrast, share only SEOs experience an abnormal return of -2.13%, but would have been -1.27% had they issued warrants. In addition, unit offerings experience less severe underpricing when compared to that of shares only offerings.

Ng and Smith (1996) report that the warrants issued to the underwriters as part of their compensation package are used as a signal to certify that the season equity offering will not be overpriced. Warrants link the underwriters compensation with the performance of the company’s share price as they function as a bonding mechanism that enables underwriters to certify the offer price. In the existing literature there have also been reported other signalling mechanisms which are used by IPO firms to signal their quality, such as earnings forecasts and lock-up agreements. More specifically, some papers find that the voluntary disclosure of the earnings forecasts in the IPO prospectuses can have a favourable impact on the degree of underpricing and the post-IPO return performance (Jog and Mcconomy, 2003). In addition, earnings forecasts may be an indication of the managers’ ability to anticipate

26 future changes and make the necessary adjustments to the production (Trueman, 1986). Liu and Ziebart (1999) and Skinner (1994) report that the market underreacts to good news earnings forecasts and overreacts to bad news forecasts. Moreover, Firth (1998) examined Singaporean IPOs and reported a statistically significant and positive relationship between the size of earnings forecasts and the IPO value, implying that the forecast size is an important signal of the firm value. However, on AIM the vast majority of companies do not repot earnings forecasts.

As far as the lock-up agreements is concerned, Goergen et al. (2006) and Brav and Gompers (2003) report that IPO firms with severe agency problems tend to extend the periods of the lock-up agreements. Huang et al. (2009) find that IPOs use underpricing and the length of the lock-up periods as substitutes for signalling the firm’s quality. Arthurs et al. (2009) find that IPOs that have an ongoing concern/issue are able to decrease the level of underpricing at the IPO day by accepting a longer lock-up period. In addition, Espenlaub at al. (2001) find that UK IPOs do not have standardised lock-up periods when compared to those of the US. Instead IPO firms in the UK have relative lock-up periods which are linked to the company’s calendar events, such as the publication of preliminary annual results. Furthermore, they find that high-tech firms are more likely to choose absolute expiry dates because they are characterised by greater informational asymmetries. On AIM the vast majority of IPOs issue warrants to their brokers without lock-in periods. This implies that the brokers can exercise them from the very first day of listing on the stock exchange.

In this study we examine three different hypotheses. Chemmanur and Fulghieri (1997), Schultz (1993), Mazouz et al. (2008), How and Howe (2001), Lee et al. (2003) and Garner and Marshall (2005) report that it is usually smaller companies that conduct unit IPOs and sell a combination of shares and warrants to investors. Furthermore, Barry et al. (1991), Dunbar (1995) and Ng and Smith (1996) document that smaller companies issue compensation warrants to their underwriters. The rationale of the two aforementioned strands of literature is that warrants are a way of certifying that the offering is not overpriced as the value of warrants is directly linked to the aftermarket performance of the stock price. This certification through warrants should be more valuable for smaller companies which have greater informational asymmetries. Based on this argument our first hypothesis is the following:

Hypothesis 1: Smaller firms are more likely to issue warrants to their brokers.

In addition, Logue and Lindvall (1974) report that if insiders are selling more shares, they may have a better bargaining power in lobbying for lower floatation costs. Dunbar (1995) confirms their argument and finds that the percentage of existing shares sold in the IPO can send a positive signal to the market. The reason is that if insiders are selling part of their holdings then this implies two facts; first, insiders are willing to give up some of their shares in order to expand the number of investors that hold the company’s stocks, and second the firm is not dependent on a small number of existing shareholders.

However, the selling of existing shares may also send a negative signal to the market as investors may think that the IPO is an opportunity for insiders to sell overpriced securities. If this is the case companies will issue warrants to their brokers in order to certify that the shares are not overpriced, as the value of warrants is linked with the aftermarket performance of the stock price. The two aforementioned arguments formulate our next hypotheses:

Hypothesis 2a: The probability of issuing warrants will be smaller, the higher the percentage of the total gross proceeds raised from the selling of existing shares.

Hypothesis 2b: The probability of issuing warrants will be higher, the higher the percentage of the total gross proceeds raised from the selling of existing shares.

Moreover, the US Securities and Exchange Commission reported in 1963 (US Securities and Exchange Commission 1963a and 1963b cited in Ng and Smith 1996) that the underwriting services market is segmented between two groups, large issuers that are underwritten by prestigious underwriters, and small that are restricted to be underwritten by less prestigious investment banks, which demand warrants to be part of their compensation. One of the SEC findings was that firms that issue warrants to the underwriters pay higher cash spreads and experience higher underpricing. This is consistent to the monopsony power hypothesis according to which the total costs of going public are greater when warrants are used than when they are not. This is due to the fact that investment banks have a monopsony power over issuing firms. Dunbar (1995) reports that underwriters are likely to ask for security based compensation (i.e. warrants) in riskier offerings in which information asymmetries between the issuer and the underwriter are expected to be much more prevalent. If the underwriter controls the pricing of the IPO then it can increase its compensation by reducing the issue price. The lower the issue price the higher the underpricing and the value of warrants respectively. So, if investment banks have a monopsony power then the issuer cannot choose the contract that minimises its costs.

But, Dunbar (1995) and Ng and Smith (1996) find evidence of the alternative to the monopsony power hypothesis, which is the cost minimisation. According to the cost minimisation hypothesis the cost of going public when warrants are used is lower than it would have been if warrants had not been issued. The main source of this reduction in the cost comes from the lower underpricing of the IPO. If insiders can credibly signal to the market that they are not selling overpriced securities, then investors will require a lower underpricing. This credible signal can be achieved through the issue of warrants to the underwriters as they link their compensation with the aftermarket performance of the stock price. As far as the unit IPO literature is concerned, Mazouz et al. (2008) document that the underpricing is lower for IPOs that also issue warrants to their investors than what it would have been had they not issued them. Mazouz’s et al. (2008), Dunbar’s (1995) and Ng’s and Smith’s (1996) findings formulate out last hypothesis:

Hypothesis 3: Companies minimise their costs of going public through the issue of warrants.

3 The Alternative Investment Market

The AIM, which was launched in June 1995, is the LSE’s junior/second-tier market and is designed mainly for smaller, growing companies that want to raise capital at a very early stage of their development. AIM is a prescribed,7 or in other words an exchange-regulated, market as it is not supervised by the Financial Services Authority8 (FSA) of the UK, but by the LSE (Arcot et al., 2007).

The most common method of issue on AIM is a placing in which the company shares are sold to qualified investors (informed investors) by its broker via an admission document (rather than a prospectus). The admission document is not pre-vetted by the exchange or the UK Listing Authority (UKLA) and is not approved by the AIM. In the case of larger companies, if the goal is to raise a large amount of capital, a public offer is usually used. In the public offer, the securities are issued to the public through a prospectus that must comply with the European Prospectus Directive requirements. In this case, the prospectus is pre- vetted by the UKLA.

7 http://www.londonstockexchange.com/companies-and-advisors/aim/faq/faq.htm 8 In 2012 the UK regulator, the Financial Services Authority, was split into two institutions, the Financial Conduct Authority which is responsible for regulating the financial services industry, and the Prudential Regulation Authority, which is responsible for the prudential supervision of the financial advisers.

One plausible explanation for the success of the exchange-regulated markets in attracting a high volume of IPOs is their light regulatory regime. AIM has such a regulatory system, tailored to the needs of growing and smaller companies, and its rules are less onerous than those of other markets that target larger and more established companies (Arcot et al., 2007). Table 1 reports the main differences in the listing criteria between the AIM and Main markets of the LSE. Some of these differences are the no trading record before admission, no requirement for a prescribed level of shares that should be in the hands of the public and no minimum market value for companies listed on AIM (London Stock Exchange, 2007c). However, AIM companies should have a Nominated Advisor (Nomad) and a broker at all times (ongoing advisers), whereas firms listed in other markets do not.

The success of the AIM market is evident from the fact that, during the period from 1995 to 2009, 44% of all IPOs conducted in the German, French, Italian and UK markets, and 77% of all IPOs listed on the exchange-regulated markets9 of the aforementioned countries, took place on AIM (Vismara et al., 2012). In addition, between 1995 and 2008, 1,743 companies have been listed on AIM, raising a total of £29.1 billion (market capitalisation of £65 billion). According to PricewaterhouseCoopers (2012), London is the most international of all the capital markets around the world because more than 21% of all companies listed maintain operations outside the UK. More specifically, from 1995 to 2008, 317 of the companies listed on AIM were incorporated outside the UK and from 2005 onwards more than 20% of the IPOs taking place on AIM are international companies. The importance of the AIM market is evident from two facts; it is Europe’s largest market by number of companies (London Stock Exchange, 2007c) and from 2004 onwards there are more listings on AIM than on Nasdaq (Gerakos et al., 2011).

3.1 The role of the broker

The broker plays an integral part not only during the flotation process but also after admission has taken place. According to Rule 35 of the AIM rules for companies (London Stock Exchange, 2010) a firm listed on AIM should have a broker at all times. If there is a resignation or a dismissal of the broker, then the company should notify the market immediately (Rule 17). The purpose of Rule 35 is to ensure that there will be an orderly market for the securities of a company.

9 The exchange-regulated markets include the Entry Standard of Deutsche Börse in Germany, Marche Libre and Alternext in France, Mercato Alternativo dei Capitali and AIM Italia in Italy and AIM in the UK.

The roles of the broker and the Nominated Adviser (Nomad) are sometimes confused, especially when the same adviser is performing both roles. However, even when the Nomad and the broker are the same institution, the two roles are completely different. The Nomad is responsible for determining whether a company is suitable to be admitted on the AIM market and provides ongoing advice regarding the company’s obligations and compliance with the AIM rules. On the other hand, the broker is responsible for arranging the fundraising, maintaining a liquid aftermarket and generally ensuring that there is sufficient interest in the company’s securities.

More specifically, the broker is responsible for organising the roadshows, during which the company’s management meets with potential investors, procuring investors for the company’s shares, building the book, collecting the funds from investors and allocating the shares. The broker can even advise the company against flotation if it believes that the market sentiment is not favourable for the IPO to take place (Burton et al., 2006, London Stock Exchange, 2007c). It also acts as a market maker for the company’s shares (London Stock Exchange, 2007b). As far as the secondary market is concerned, the broker advises the company on how to maintain good investor relations, provides information about relevant market and trading issues (i.e. changes in market conditions), and assists the firm with other corporate finance services and fundraising activities (i.e. rights issues, mergers) (London Stock Exchange, 2007b).

When one adviser is simultaneously playing both roles (integrated house) then it should be an approved Nomad (approved by the LSE) and there should be a clear distinction between its responsibilities as a Nomad and those as a broker (Chinese wall). The broker plays a critical role in the IPO process on AIM since it is the party standing between the investors and the company, whereas the Nomad (regulator) is the party standing between the company and the AIM market.

The existing literature on AIM has paid no attention to the role of the broker, as the focus has always been on the role of the Nomad (Gerakos et al., 2011, Mallin and Ow-Yong, 2011, Mallin and Ow-Yong, 2010, Mendoza, 2008). To the best of our knowledge, this is the first empirical study that describes and focuses on the role of the broker during the flotation process.

3.2 Broker’s compensation structure

The compensation the broker receives for advising a company seeking a listing on AIM consists of two main parts (London Stock Exchange, 2007b, p. 34). The first is the commission, which is calculated as a percentage of the amount of money raised in the flotation, and is paid to the broker in return for procuring subscribers for the new shares and buyers for the selling/existing/secondary shares offered during the IPO. The second component is the retainer fee, which is paid annually to retain the adviser on an ongoing basis, post-listing.

In some cases warrants can be considered a third component of the brokers’ compensation because some companies issue warrants to their brokers in return for the services they provide. In these cases, the compensation of the brokers is directly tied up with the aftermarket performance of the stock price. If the offering is overpriced and the stock price drops in the aftermarket, then the value of the warrants decreases and the total compensation of the advisers also decreases. If the offering is underpriced and the stock price increases in the aftermarket, the value of the warrants increases and the total broker compensation (including warrants) increases.

On AIM there is no regulation which determines a maximum or a minimum limit on the brokers’ compensation. The compensation the broker is (commission, annual retainer fee and issuance of warrants) paid depends on the initial agreement that has made with the issuer, which is described in the IPO admission document.

4 Data

The data includes non-financial IPOs that took place on the AIM market of the London Stock Exchange during the period June 1995 to December 2008. After excluding all financial, unit and underwritten IPOs and those that were listed in other stock exchanges prior to obtaining a listing on AIM the initial sample consists of 679 non-underwritten offerings.

8 IPOs are dropped because the admission documents are not available in Thomson One Banker, Thomson Research or PI Navigator databases. In 16 IPOs the stock price series is not available in Datastream, Thomson One Banker or Bloomberg and in 5 IPOs it is not clear in the admission document whether the company issues warrants or not. In addition, 2 IPOs that issue warrants are excluded from the final sample since the exercise price of the warrants changes over time making it difficult to value them with conventional valuation formulas

(Barry et al., 1991, Dunbar, 1995). We also omit 36 offerings that issued warrants to other advisers (i.e. Nomad).

When all the aforementioned IPOs are excluded the remaining non-underwritten offerings sample consists of 612 IPOs. 103 of these issues took place from June 1995 to December 1998 and are used to construct the dynamic brokers’ reputation (reputation based on gross proceeds and number of IPOs during the 3.5 years prior to the flotation date). The final sample consists of 509 (612 - 103) non-underwritten offerings. 221 issued warrants to their brokers and 288 did not. So, 43.42% of all non-underwritten IPOs issued warrants to their brokers.

The stock price series is collected from Datastream, Thomson One Banker and Bloomberg. The data regarding the compensation paid to the brokers, such as the commission and warrant characteristics (shares underlying the warrants, exercise price and time to expiration) are collected from the admission documents. All the rest of the data, such as the issue price, date of incorporation, issue size, shares outstanding, etc., are collected from the admission documents as well.

4.1 Pricing of warrants

In order to find out the value of warrants, and consequently the cost that the IPO firms incur when they issue them to their brokers, we follow Barry’s et al. (1991) and Dunbar’s (1995) approaches and make use of the Cox constant elasticity variance model (CEV). The first day closing price, and not the issue price, is used in the CEV formula as the price of the underlying stock (Dunbar, 1995). Barry et al. (1991) report that whichever price is used the inferences are qualitatively the same. The first day closing price can be justified by the fact that brokers may have inside information for the company value when the offer price is decided (Barry et al., 1991). This is due to the fact that the brokers are responsible for conducting the bookbuilding, in which they contact institutional investors, obtain bids for the company’s shares and consequently determine the issue price (Burton et al., 2006, Jenkinson and Jones, 2004, London Stock Exchange, 2007b).

Due to the fact that the company’s shares are not publicly traded before its admission on the AIM market, it is not possible to compute the volatility of the security in order to use it in CEV formula. Barry et al. (1991) use a time period of 126 days prior to the offer in order to calculate the average standard deviation for all the NASDAQ stocks that exist in CRSP

33 database. This is equivalent to forming a CRSP equity index, as they use all available stocks in the CRSP database. In this study we compute the average standard deviation of the FTSE AIM all share index for 126 days before the offering. Also, according to the AIM admission timetable (London Stock Exchange, 2009) a company appoints its advisers (broker, Nomad) and agrees the timetable for listing 12 to 24 weeks (approximately 120 working days) prior to admission. This implies that the advisers, in cooperation with the company, will decide when the listing will take place 12-24 weeks before the actual IPO date. This is an additional reason that we use a time period of 126 days in order to capture the volatility that exists in the market near the time of listing. Barry et al. (1991) also use sector volatilities, and their results are qualitatively the same. We cannot use AIM sector indices to calculate the volatility since some of them were introduced after the examined period of this study (i.e. automobiles and parts and food and beverage were introduced in the end of the year 2000). As far as the risk free rate is concerned, the Bank of England base rate in the month of the offering is used (Dunbar, 1995).

Moreover, Barry’s et al. (1991) and our measure of volatility have one disadvantage. Companies that are already listed for many years in the stock exchange (i.e. 5 years) will probably be less volatile when compared with those that are listed a few months (i.e. 3 months) before the IPO of interest. This is why we also use an alternative measure of volatility, which is based on 20 daily company specific returns in the aftermarket. Although this is an ex post measure that is not known at the time of listing, it capture the specific risk of the company of interest.

However, the non-underwritten IPO market on AIM, which mainly includes the smallest capitalisation IPOs, is characterised by thin trading/illiquidity. When the shares do not trade at every consecutive interval on the stock exchange then we have thin trading. This is evident from the fact that when we calculate a company’s daily returns for a period of time, i.e. one month, then most of them give the value of zero. Thin trading introduces autocorrelation in the time series of returns (Antoniou et al., 1997). In order to correct for that we use Miller’s et al. (1994) approach. According to Miller et al. (1994), by using an AR(1) (autoregressive) model we can correct for thin trading. More specifically we estimate the following equation:

푅푡 = 푎1 + 푎2푅푡−1 + 푒푡 (1)

푅푡 = return at time t

푅푡−1 = return at time t-1

Then by using the residuals 푒푡 from the regression we obtain the adjusted returns as follows:

푎푑푗 푅푡 = 푒푡⁄(1 − 푎2)

푎푑푗 푅푡 = return at time t, corrected for thin trading

The aforementioned approach assumes that the non-trading adjustment ( 푎2 ) remains constant throughout the examined period. Although this may be true for highly liquid markets, most probably the required adjustment will vary through time in the case of illiquid markets, such as AIM. This is why equation (1) is estimated recursively (Antoniou et al., 1997). Due to the fact that we have to run a regression, just 20 aftermarket observations cannot give us robust results. As a result, we use a period of 126 days in the aftermarket in order to calculate recursive daily returns and control for thin trading.

5 Methodology

The issue of warrants may be related to the underpricing and the total broker compensation costs incurred by the companies during the IPO. The decision to issue warrants may be a non-random one because companies may choose the contract that is more favourable for them (endogenous choice) (Li and Prabhala, 2007). This is true as companies rarely make decisions randomly (Hamilton and Nickerson, 2003). As a result, if the IPOs that issue warrants are not a random subset of the whole population then OLS regressions do not yield consistent estimates. The two stage Heckman (1979) selection model can control for this selection bias.

We make use of the endogenous switching model, which is an extension of the baseline Heckman self-selection model. In the first stage this model (selection model) consists of a binary choice equation which captures the decision to issue warrants and is the following:

∗ 퐼푖 = 푍푖훾 + 휀푖 (2)

Vector 푍푖 includes all observable independent variables that may influence the decision to issue warrants. Some of these variables may also affect the underpricing and total broker compensation. Vector 훾 includes all parameters that must be estimated and 휀푖 is the error

35 term. We use a probit model to estimate the selection model in which the dependent variable ∗ 퐼푖 is equal to one if the firm issues warrants and zero otherwise. ∗ ∗ 퐼푖 = 1 iff 퐼푖 > 0, and 퐼푖 = 0 iff 퐼푖 ≤ 0 (3)

By using the estimated value of 푍푖훾̂ we compute the inverse Mills ratio (IMR), which is different for the IPOs that issue warrants and for those that do not. The IMR is then included in the second stage of the Heckman model, in which we have two regression equations for the variable of interest conditional on the choice made in the first stage. The second stage equations are the following:

푦1푖 = 푋푖훽1 + 푢1푖 (4)

푦2푖 = 푋푖휃2 + 푢2푖 (5)

Equation (4) is the underpricing equation for the IPOs that issue warrants to the brokers, and equation (5) is the underpricing equation for those that do not. We only observe either 푦1푖 or

푦2푖, for each IPO, based on the outcome of 퐼푖:

푦푖 = 푦1푖 iff 퐼푖 = 1, and 푦푖 = 푦2푖 iff 퐼푖 = 0 (6)

푦푖 includes the underpricing and 푋푖 includes the independent variables that affect the underpricing when the IPO firms issue warrants or when they do not. We use the same methodology in order to examine the relation between the use of warrants and total broker compensation, by replacing the two underpricing equations with two total broker compensation equations. In this case, 푥푖 will consist of the variables that affect the total broker compensation when warrants are or are not used.

The independent variables included in vectors 푍푖 and 푋푖 can also be identical (Golubov et al., 2012, p. 304). It is not necessary to apply any exclusion restrictions in the second stage regressions because they are not critical in the Heckman selection model, as this model is identified by the nonlinearity of the IMRs. So, the second stage models are still valid even without any exclusion restrictions (Golubov et al., 2012, p. 304).

This is called a ‘what-if’ type of analysis because we only observe the underpricing and total broker compensation for companies that issued warrants, but we cannot observe what would the underpricing and total broker compensation be had the same company chosen not to issue warrants (counterfactual). The selection bias arises from the fact that the errors in equations

(4) and (5) may be correlated with the error in equation (2). So, the covariance between 휀푖 from equation (2) and 푢1푖 and 푢2푖 from equations (4) and (5) may not be zero. If we try to estimate equations (4) and (5) by OLS then we make the assumption that all variables that affect both the decision to issue warrants and the underpricing are observable and are included in the regressions. But, this is rarely true as there may be variables that are not observed by the researcher, which will consequently lead to potential endogeneity problems.

By using a self-selection regression model we take into account the aforementioned problems as we allow the error in equation (2) to be correlated with the errors in equations (4) and (5), so that unobserved or missing variables in the binary outcome equation (2) can also have an effect on the underpricing. Parameters β1 and 휃2 cannot be estimated directly by using OLS because this will generate inconsistent estimates since the expectation of 푦1푖 does not have a zero mean (u1 and ε may be correlated).

For this reason in the second stage we estimate equations (4) and (5) by OLS, including one additional regressor to adjust for the potential non-zero expectation of the errors. This regressor is the IMR, which allows equations (4) and (5) to be estimated consistently using OLS (Lee, 1978, Heckman, 1979). The IMR is defined as:

휑(푍푖훾) 퐼푀푅1 = − for IPOs that issue warrants and Φ(푍푖훾)

휑(푍푖훾) 퐼푀푅2 = for IPOs that do not issue warrants 1− Φ(푍푖훾)

휑 is the standard normal density function and Φ is the standard normal cumulative distribution function. If at least one of the IMRs in equations (4) and (5) is statistically significant then this implies that there is self-selection and we have to use the two stage Heckman model. If none of the IMRs is statistically significant then OLS estimates are not affected by selection bias (Golubov et al., 2012, p. 291).

We do not make use of a two simultaneous equations system, one for underpricing and one for warrants, in which the underpricing and warrants equations are functions of each other, for three important reasons:

First, in a simultaneous equation system both dependent variables affect each other. More specifically, the equation in which the underpricing is the dependent variable will include the issue of warrants as one of the independent variables and the equation in which warrants is the dependent variable will include the underpricing as one of the independent variables. However, underpricing cannot affect the issue of warrants because the decision to issue warrants is made before the company is listed on AIM and the first day return (underpricing) is observed. The company already discloses the issue of warrants in the IPO prospectus, then the company is listed on the stock exchange and the underpricing is observed. As a result, underpricing cannot be included as an independent variable in the warrants equation.

Second, in a simultaneous equation system the warrants equation makes the assumption that the slope coefficients of the independent variables are equal for the cases where companies issue and do not issue warrants. But, as we mentioned above the decision to issue warrants may be a non-random one because companies may choose the contract that is more favourable for them (Li and Prabhala, 2007). According to Hamilton and Nickerson (2003) companies rarely make decisions randomly. This implies that the slope coefficients may differ between warrant and no warrant issuers. If we believe that the issue of warrants has a slope effect (i.e. the beta coefficients differ according to the issue or no issue of warrants) then a sample selection (i.e. Heckman two stage model) model should be used. This is also the main difference between sample selection and endogeneity. Sample selection bias refers to the issue where the dependent variable is observed only for a restricted, nonrandom sample. For instance, we observe the underpricing and the cash compensation of the warrant IPO firms only if the companies issue warrants. However, we cannot observe the underpricing and cash compensation for the same firms had they not issued warrants. Endogeneity refers to the case where an independent variable that is included in the model is probably a choice/dummy variable which is correlated with the error term and is observed for all observations in the data (Heckman, 1979, Maddala, 1983, Main and Reilly, 1993, Lowry and Shu, 2002).

Third, the main goal of this study is to examine what the counterfactual would have been for firms that issue and do not issue warrants. For instance, for an IPO that does not issue warrants, what would the underpricing be if the firm chose to issue warrants. This analysis can be performed by using an extension of the Heckman self-selection model, the endogenous switching model. This is one additional reason that we cannot use a simultaneous equation system.

In this study, the analytical form of the first stage probit10 selection model is the following:

퐼푊 = 훾0 + 훾1푃푢푏푙𝑖푐_퐹푙표푎푡 + 훾2푀퐶

+ 훾3푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠

+ 훾4퐵푟표푘푒푟_푅푒푝. +훾5퐷푢푚푚푦_푁_퐵푟 + 훾6퐴𝑔푒 (7) + 훾7푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 훾8푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥) + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠 + 휀

where 퐼푊 is a dummy variable that takes the value of one if the firm issues compensation warrants to its broker(s) and zero otherwise, Public_Float is the ratio of the total number of shares sold in the IPO divided by the outstanding shares, MC is the natural logarithm of the market capitalisation of the company, Secondary_Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds), Broker_Rep. (GP and N of IPOs) is the reputation of the broker based on the gross proceeds and number of IPOs raised and advised by each broker over the previous 3.5 years, Dummy_N_Br is a dummy variable that takes the value of one if the broker and the Nomad are the same firm and zero otherwise, Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age), Vol_(FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date multiplied by 100, Ret_(FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date, Yearly_dummies are the yearly dummies for the examined period and 휀 is the stochastic error term.

10 Tucker (2010) reports that the calculation of the IMR requires the numerator and denominator functions to be normally distributed. This is why we should make use of a probit, and not a logit model.

The second stage OLS regressions for the underpricing are the following:

푈푛푑푒푟1 = 훽0 + 훽1푃푢푏푙𝑖푐_퐹푙표푎푡 + 훽2푀퐶

+ 훽3푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠

+ 훽4퐵푟표푘푒푟_푅푒푝. +훽5퐴𝑔푒

+ 훽6푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥) (8)

+ 훽7푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 훽8퐼푀푅1 + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠

+ 푢1

푈푛푑푒푟2 = 휃0 + 휃1푃푢푏푙𝑖푐_퐹푙표푎푡 + 휃2푀퐶

+ 휃3푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠

+ 휃4퐵푟표푘푒푟_푅푒푝. +휃5퐴𝑔푒

+ 휃6푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥) (9)

+ 휃7푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 휃8퐼푀푅2 + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠

+ 푢2

where 푈푛푑푒푟1 and 푈푛푑푒푟2 are the first-day returns for the IPOs that issue warrants to their brokers and those that do not respectively and are calculated as: (closing price – issue price)/issue price. The independent variables in equations (8) and (9) have already been explained above. The only difference is that equation (8) includes only the observations of the companies that issue warrants to their brokers whereas equation (9) includes the observations for the firms that do not.

As far as the total broker compensation is concerned, the second stage OLS regressions are the following:

푇표푡푎푙 퐵푟표푘푒푟 퐶표푚푝푒푛푠푎푡𝑖표푛 1 훽0 + 훽1퐷푢푚푚푦_푁_퐵푟

= + 훽2 푃푢푏푙𝑖푐_퐹푙표푎푡 + 훽3푀퐶

+ 훽4푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠

+ 훽5퐵푟표푘푒푟_푅푒푝. +훽6퐴𝑔푒 (10) + 훽7푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 훽8푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 훽9퐼푀푅1 + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠

+ 푢1

푇표푡푎푙 퐵푟표푘푒푟 퐶표푚푝푒푛푠푎푡𝑖표푛 2 휃0 + 휃1퐷푢푚푚푦_푁_퐵푟

= + 휃2푃푢푏푙𝑖푐_퐹푙표푎푡 + 휃3푀퐶

+ 휃4푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠

+ 휃5퐵푟표푘푒푟_푅푒푝. +휃6퐴𝑔푒 (11) + 휃7푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 휃8푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 휃9퐼푀푅2 + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠

+ 푢2

where 푇표푡푎푙 퐵푟표푘푒푟 퐶표푚푝푒푛푠푎푡𝑖표푛1 and 푇표푡푎푙 퐵푟표푘푒푟 퐶표푚푝푒푛푠푎푡𝑖표푛2 are calculated as: (Commission + Warrant Value)/Gross Proceeds, for the IPOs that issue warrants to their brokers and those that do not. The independent variables in equations (10) and (11) have already been explained above. The only difference is that equation (10) includes only the observations of the companies that issue warrants to their brokers whereas equation (11) includes the observations for the firms that do not.

The endogenous switching regression model allows us to conduct a counterfactual analysis on the alternative choice which we cannot observe. The counterfactual is computed by multiplying the coefficient estimates obtained from the second stage regressions with the explanatory variables. For instance, in order to calculate what the underpricing would be for IPOs that issued warrants had they not issued them, we multiply the coefficient estimates from equation (9) with the independent variables from equation (8). In this way we obtain the underpricing if the alternative contract had been chosen (no issue of warrants). Then we

41 compare the estimated underpricing with the actual one and infer whether the issue of warrants has any effect on the level of underpricing.

The explanatory variables included in the probit model are selected from the existing empirical literature. Chemmanur and Fulghieri (1997), Schultz (1993), Mazouz et al. (2008), How and Howe (2001), Lee et al. (2003) and Garner and Marshall (2005) report that it is usually smaller companies that conduct unit IPOs and sell a combination of shares and warrants to investors. Moreover, Barry et al. (1991), Dunbar (1995), Ng and Smith (1996) and Jain and Kini (1999) report that smaller and riskier firms with greater ex ante uncertainty are more likely to issue warrants. According to Booth and Smith (1986) non-cash compensation will be more valuable for smaller firms in which we would expect informational asymmetries between insiders and outsiders to be higher. This is why we include the variables age, public float and market capitalisation of the company in our analysis.

In addition, brokers may take warrants as part of their compensation package when the market is doing well. Brokers may have the ability to time the issue of warrants and include them in their compensation when the economic environment is positive. This is why we use the return and volatility of the FTSE AIM index two months prior to the IPO date. The rationale is that if the FTSE AIM index is increasing, then it is likely that the company’s stock price will also increase and consequently the value of warrants will be higher.

Moreover, Logue and Lindvall (1974) report that if insiders are selling more shares in the IPO, they may have a better bargaining power in lobbying for lower floatation costs. Dunbar (1995) confirms their argument and finds that if existing shareholders sell their shares in the IPO then they can lobby for lower compensation fees. In order to take his finding into account a variable that captures the percentage of gross proceeds raised from the selling of existing shares is included in our analysis. Barry et al. (1991) and Dunbar (1995) also document that firms that issue warrants are underwritten by less reputable underwriters. Ng and Smith (1996) report that reputable underwriters, who have sufficient reputational capital to certify that the issue is not overpriced, will not include warrants in their compensation. On the other hand, less reputable underwriters, who lack reputational capital, will include warrants in order to certify the IPO. So, according to Ng and Smith (1996) warrants can substitute for reputational capital as they link the compensation of the underwriters with the

42 aftermarket stock price performance. This is why we also include the reputation of the broker as a control variable in our analysis.

We use four different variables to capture the reputation of the broker. Two of them are based on the number of IPOs and gross proceeds raised during the 3.5 years prior to the IPO date (dynamic reputation). So, according to the dynamic reputation measure, the first IPO, for which the broker reputation is available, took place in 1999 (reputation based on the previous 3.5 years, June 1995 – December 1998). The reasons for choosing a period of 3.5 years are the following; it takes some time for a broker to build its reputation and in periods of a depressed market reputable brokers may not advise any IPOs in order to avoid damaging their reputation (Goergen et al., 2006). The other two reputation measures are based on the total number of IPOs and gross proceeds advised and raised by the broker during the examined period (static reputation). Fang (2005) and Megginson and Weiss (1991) also compute their reputation variable in the same way (based on market share) in their papers for the US market. In order to maintain the same number of observations in our probit model (equation (2)) for all different measures of broker reputation (dynamic and static), we construct the static reputation from 1999 to 2008 and not from the end of June 1995.

Mallin and Ow-Yong (2011) include in their analysis a dummy variable that takes the value of one if the Nomad and the broker are the same firm and zero otherwise. They report that when the Nomad also acts as a broker the reputational capital that is at stake is higher than that if the Nomad and broker are different. Their reasoning is that when an adviser is conducting both roles (Nomad and broker), it will be more careful and sponsor companies that comply with the corporate governance recommendations of the AIM market because of the relatively higher reputational loss the adviser will incur if the company fails. So, if the IPO firm has two advisers then any reputational loss to the broker and Nomad will be shared between them. Due to this argument, if the broker also acts as a Nomad it may only sponsor good quality IPOs (less risky) that do not issue warrants, whereas if the Nomad and broker are different then they may sponsor more risky IPOs and may ask for warrants as a compensation for this extra risk. Another possible explanation that the probability of warrants may be higher when the broker and Nomad are two different firms is that when an adviser is acting both as a Nomad and broker then there may be a better coordination between the company’s management and the Nomad/broker, as the company has to deal with only one adviser and not two. So, the adviser may charge one fee for both roles. However, if the

Nomad and broker are different, then they may charge individual fees and the probability of also getting paid non-cash compensation (i.e. warrants) becomes higher.

In the underpricing second stage regression we omit the dummy variable that takes the value of one when the Nomad and the broker are the same firm and zero otherwise. The reason that we use the aforementioned variable only in the total broker compensation regression is because we want to examine whether the broker charges higher fees if it is a different adviser from the Nomad.

6 Empirical results

6.1 Descriptive statistics - univariate analysis

From Table 2 it is evident that 43.42% of all non-underwritten offerings issue compensation warrants to the brokers. This is in stark contrast with the US market in which almost all best efforts offerings issue warrants to their underwriters (Dunbar, 1995). As a result, the AIM market enables us to conduct a unique study on the use of warrants in non-underwritten offerings that has never been examined before.

Table 2 compares company and other characteristics between the warrant and no-warrant IPO groups. The no-warrant IPO firms sell their shares at higher issue prices when compared to those of the no-warrant IPOs (£0.94 vs. £0.66) and the differences in means and medians are statistically significant at conventional levels. Fang (2005), Klein and Leffler (1981), Shapiro (1983) and Allen (1984) report that higher prices can be an indicator of superior quality. This is due to the fact that when the quality of a firm that is listed for the first time in the stock exchange is ex ante unobserved, then a higher price implies a better quality. The reason is that the present value of future income exceeds the short term profit made from selling low quality securities at high prices. In accordance to the aforementioned argument, the inverse of the issue price, which is used as a proxy for the risk of the IPO, is significantly higher for the warrant IPO firms than that of the no-warrant firms. This implies that companies that issue warrants to their brokers are riskier when compared to those that do not.

In addition, companies that issue warrants raise less gross proceeds (£8.47 vs. £10.6 millions), less proceeds from the selling of existing/secondary shares (4.97% vs. 10.06%) and have a smaller market capitalisation (£28.7 vs. £35.7 millions) than those that do not.

Moreover, the median volatility of the companies’ returns for 20 and 126 days (recursive) in the aftermarket is significantly higher for the warrant IPO group.

Furthermore, the commission paid to the broker by the issuer (as a percentage of gross proceeds) is significantly lower for the no warrant IPO group when compared to the warrant one (3.26% vs. 4.08%). Table 2 also reports that the value of warrants is 2.72% of the gross proceeds. So, the warrant group of companies not only pay a higher commission, but also grant warrants to their brokers. Moreover, firms that issue warrants are advised by less reputable brokers when compared to those that do not. The aforementioned finding is consistent with that of Barry et al. (1991), Dunbar (1995) and Jain and Kini (1999). Table 3 provides the correlation matrix of the variables that will be used later in the two stage probit model. All in all, from Table 2 it is evident that companies that issue warrants to their brokers appear to be riskier than those that do not, as they have a higher aftermarket standard deviation of returns, sell their shares at lower prices, raise less gross proceeds from the IPO, are smaller and are underwritten by less reputable brokers. Barry et al. (1991), Dunbar (1995), Ng and Smith (1996) and Jain and Kini (1999) also report similar results for US firm commitment IPOs.

6.2 Warrant characteristics

Table 4 reports the characteristics of 242 warrants issued to brokers as part of their compensation package. The reason that the number of warrants is greater than the number of IPOs (242 warrants, 221 IPOs) is because in some cases the company has two brokers and issues warrants to both of them or may even issue warrants with different characteristics (i.e. time to expiration) to an individual broker. The average size and value of the warrants, expressed as a percentage of the number of shares issued and the gross proceeds raised in the IPO, are 7.3% and 2.5% respectively (Panel A). The value of warrants is based on the volatility of the FTSE AIM all share index for 126 days before the offering takes place. The average ratio of exercise price to offer price is almost equal to 1 (1.02 in Panel A). 214 warrants out of 242 (around 88.43%) have an exercise price equal to the offer price (panel D), whereas only 21 of them (8.7%) have an exercise price which is on average 36% higher than the offer price (panel F). Interestingly, there are also warrants (7 warrants or 2.9%), in Panel E, that are offered at an average discount of 30%. This means that the exercise price is 30% lower than the issue price. These seven warrants are already in the money even before trading commences.

The time to expiration is on average 4 years and there are warrants which have a time to maturity of up to 10 years. Another important finding is that only 26 (10.74%) warrants have an average lock in period of 1.07 years (Panel C). The remaining 216 (89%) warrants have no lock in period, so they can even be exercised from the first day of trading. Under the UK’s institutional setting which has almost no regulations underlying the non-cash compensation, warrants can have an expiration period of up to 10 years, no lock-in agreements and an exercise price lower than the offer price.

6.3 Test of the cost minimisation hypothesis

6.3.1 Probit regression model

Table 5 reports the results of two probit models in which the dependent variable is a dummy that takes the value of one if the company is issuing warrants to its broker and zero otherwise. The only variable that is different between Models 1 and 2 is the reputation of the broker. In Model 1 the reputation is based on the gross proceeds raised from the flotation whereas in Model 2 it is based on the number of IPOs advised by the broker during the 3.5 years prior to the flotation. From Table 5 it is evident that the probability of issuing warrants is higher for firms that have a smaller market capitalisation, raise less money from the selling of existing shares, are advised by less reputable brokers and the Nomad and broker are two different advisers. Dunbar (1995) also finds that riskier firms are more likely to issue warrants to their underwriters in his study for US firm-commitment offerings.

More specifically, from Table 5, there is a negative and statistically significant relationship between the market capitalisation of the company and the probability of issuing warrants. This implies that the larger the company the less likely it is to issue warrants to its brokers. In addition, the probability of issuing warrants is affected by the reputation of the broker. The coefficient has a negative sign, which suggests that the lower the reputation of the broker the higher the probability of the issuance of warrants. So, reputable brokers will get paid only cash compensation (commission) whereas those who lack reputational will get paid a combination of cash compensation and warrants.

Moreover, the money raised from the secondary/existing shares (as a percentage of the gross proceeds) has a significantly negative relationship with the dependent variable. This is consistent with the argument that when existing shareholders sell shares then they have a

46 higher bargaining power in lobbying for lower fees. The reason is that through the selling of secondary shares insiders can broaden the pool of investors that hold the company’s stocks.

Burton et al. (2006, p. 678, 688) also found that one of the main goals that smaller companies want to get listed on a stock exchange is to widen their shareholder base, since prior to listing there is a very limited number of dominant shareholders. This implies that when insiders sell shares this is considered a positive signal for the market since they are willing to give up some of their holdings and broaden the company’s investor base. Also, Brau et al. (2007) report that the selling of secondary shares in the IPO is not a negative signal for the market. Instead, it is possibly driven by liquidity reasons and not by opportunistically selling overpriced shares in the flotation. As a result, the existing shareholders, when they give up part of their holdings, during the flotation, they can achieve a better bargaining power in lobbying for paying lower compensation to their advisers. One possible reason may be the fact that the broker(s) will need to put less effort to attract investors for the IPO because the existing shareholders have already sent a positive signal to the market.

Another variable that is included in our analysis is a dummy that takes the value of one if the broker is the same adviser as the Nomad and zero otherwise. In both models this variable is statistically significant at 5% level and has a negative coefficient which means that if the broker and the Nomad are the same adviser then the probability of issuing warrants is smaller. This is consistent with two different arguments. The first one is that if the broker and Nomad are the same firm then there may be a better coordination between the IPO company and the adviser. As a result the broker may charge one fee for both roles and may not get paid warrants. The second one is that in the case of a riskier IPO the reputational capital which is at stake for an adviser who is conducting both roles (Nomad and broker) simultaneously is higher than what it would have been if the broker and the Nomad are two different firms (Mallin and Ow-Yong, 2011). So, if one adviser is conducting both roles then it will only advise less risky IPOs that do not issue warrants.

6.3.2 Second stage regression estimates

The results obtained from Model 1 in Table 5 are used to construct the inverse Mills ratio. Then we run OLS regressions (second stage regressions) in which the dependent variables underpricing and total broker compensation (commission + warrant value) are regressed on the IMRs and on the independent variables separately for the two IPO groups, those that issue warrants to their brokers and those that do not. The IMRs are included in the OLS

47 regressions in order to adjust for the selectivity bias because we can only observe the contracts used by the issuer but we cannot observe what would have happened if the issuers had used the alternative contract. The statistical significance of the IMRs in Table 6 implies that there exists selectivity bias and if we did not include them in the OLS regressions then the estimates would be biased and inconsistent.

For the IPO firms that issue warrants the underpricing has a negative and statistically significant relationship with the market capitalisation, broker reputation and age of the firm. This implies that companies that are larger, exist for a longer period and are advised by more reputable brokers will experience a lower underpricing. In addition, the volatility and return of the FTSE AIM all share index during the two month period prior to the IPO date have a positive relationship with the underpricing which means that in periods of high volatility and return in the market the IPO firms will be more underpriced. For the same group of companies the total broker compensation has a significantly negative relationship with the public float, which implies that the more shares sold in the IPO, the lower the total broker compensation will be.

For the IPO group that do not issue warrants the public float has a significantly negative relationship whereas the volatility has a significantly positive relationship with the underpricing respectively. Thus, the higher the volatility of the FTSE AIM all share index and the smaller the number of shares sold, as a percentage of the outstanding shares, the higher the underpricing will be. For the same group of companies the total broker compensation is positively affected by the return of the FTSE AIM all share index and negatively affected by the market capitalisation, the broker reputation and the dummy variable that takes the value of one if the broker and the Nomad are the same adviser. So, the no warrant IPO group will pay lower fees when the return of the FTSE AIM index is low, the market capitalisation of the company and the reputation of the broker are high and the Nomad and the broker are the same institution.

In order to examine what the underpricing and the total broker compensation would have been had the alternative contract been used we multiply the coefficient estimates from Table 6 (second stage OLS regressions) with the independent variables. Then we compare these values with the actual underpricing and total broker compensation (Table 7).

For companies that issue warrants to their brokers the actual underpricing is 15.95% and would have been 22.12% had warrants not been used. So, these companies would have

48 incurred a higher underpricing had they not issued warrants. For the same group of companies the total broker compensation is 6.79% and would have been 3.49% had warrants not been issued. This implies that IPO firms would pay a lower compensation to the brokers had they not issued warrants. Overall, the total costs of going public (underpricing and total broker compensation) are 22.74% and would have been 25.61% if the companies had not granted warrants. This means that the IPO firms that issue warrants are able to choose the contract that minimises their costs. This is consistent with Dunbar’s (1995) findings for US firm commitment offerings. Although a 2.87% reduction in the total costs (25.61% - 22.74%) may seem a small figure, it is equivalent to 70.34% of the commission paid to the brokers (2.87%/4.08%). The aforementioned decrease in the total costs is mainly caused by the fact that the IPO firms incur a lower underpricing (15.95% actual vs. 22.12% estimated) because, through the issue of warrants to the brokers, they certify that the IPO is not overpriced. As a result investors will require a smaller discounting for the IPO price.

Although warrants are beneficial for the issuers, it is still puzzling why do brokers not get paid all their fees in cash as there is no regulation that restricts their cash compensation. One reason may be the fact that they want to capture the upside potential of the stock price. Brokers play a crucial role for the survival of the IPO firm post-admission as they are responsible for assisting with further corporate finance and fundraising activities (i.e. rights issues, mergers), informing for relevant market issues (change in economic environment) and advising the company on how to maintain good investor relations. As a result, a firm may do well and grow significantly in the future under the guidance of its broker(s). If this happens, then the value of its stock price will rise, the value of warrants will increase and the compensation of the brokers will become higher. Another explanation may be that the companies that issue warrants may not have enough cash at their disposal to pay the brokers’ compensation. As a result, warrants may be used as a substitute of cash and brokers accept warrants instead of a higher commission. However, we cannot test this hypothesis in the non-underwritten IPOs because the vast majority of them do not disclose yearly accounting data (i.e. cash and cash equivalents) in their prospectuses. This is one of the limitations of this study.

For companies that do not issue warrants the actual underpricing is 19.73% and would have been 16.92% had warrants been used, but the difference in means is not statistically significant. The total broker compensation is 3.26%, but would have been 6.11% if warrants had been used. Overall, the total costs for the IPOs that do not issue warrants are 22.99%

49 and would have been 23.03% had they issued them. So, for this group of companies the issue of warrants is not beneficial as the total costs incurred would be similar regardless of the choice of the contract. This is why these firms choose not to grant warrants to their underwriters.

6.3.3 Robustness tests

In order to check the robustness of our results we conduct the following robustness tests:

We replicate the same analysis and instead of using the first day return (underpricing), we make use of the market adjusted first day return (MAIR), which is calculated as: first day return – FTSE AIM all share index return at the day the IPO took place. As shown from Table 2 the first day return and the MAIR are similar (18.09% vs. 18.03% respectively). Furthermore, we substitute the public float with the retained ownership variable in our model, which is the level of ownership retained by insiders of the IPO firm after the IPO. The reason we do that is because the unit IPO empirical studies include this variable in their analysis. More specifically, the agency costs hypothesis (Schultz, 1993) predicts that companies with lower managerial ownership have a higher probability of including warrants in their IPOs. Due to the fact that it is difficult to obtain data on the shareholdings of managers, previous papers make use of the retained ownership instead. From Table 2 it is evident that the retained ownership is not significantly different between the IPOs that issue compensation warrants and those that do not (70.03% vs. 68.16%). Mazouz et al. (2008) also find similar results in their unit IPO study. Then we conduct our two stage analysis again (including retained ownership and MAIR) and our results are qualitatively the same (Appendix. Tables 1 and 2), firms that issue warrants to their brokers achieve a 2.88% (as a percentage of gross proceeds) reduction in their total costs of going public (Appendix, Table 2).

We also value the warrants based on the volatility of 20 and 126 (recursive) days in the aftermarket and then run the two stage probit model. In addition, we replicate the first and second stage regressions but now we include one of the other three different variables of the broker reputation11 in equation (2). Furthermore, we construct the reputation of the broker for three and four years prior to the IPO date. Moreover, we calculate the volatility and return of the FTSE AIM index for one and three months period before the IPO date. We also include

11 These reputation variables are based on the number of IPOs advised by the broker during the 3.5 years prior to the flotation, on the gross proceeds raised and the number of IPOs advised by each broker during the period 1999 to 2008.

50 the dummy variable, which takes the value of one if the Nomad and the broker are the same institution and zero otherwise, in the second stage underpricing equation. After conducting all the aforementioned tests, our results are qualitatively the same as the cost minimisation hypothesis still holds on AIM.

Although we try different variables that capture the same information in our two stage analysis, such as the broker reputation based on the gross proceeds and number of IPOs (dynamic and static measures) for the 3, 3.5 and 4 years prior to the IPO date (statistically significant, Table 1), public float vs. retained ownership (statistically insignificant, Tables 1 in the main part of the analysis and Appendix respectively) and the volatility and return of the FTSE AIM index for one, two and three months period before the IPO date (statistically insignificant, Table 1), the R square in the probit model still remains low (approximately 10%). One potential explanation may be the data restriction and variables that we can include in our model as the non-underwritten IPOs are the smallest companies that exist on AIM and the vast majority of them do not report any accounting data in their prospectuses (cash available, total assets, revenues, etc.). As a result, we cannot use variables, such as the cash available before the IPO date, which may affect the issue of compensation warrants and consequently increase the R square of our probit model.

7 Conclusion

We examine the issue of non-cash compensation (i.e. warrants) to brokers as part of their compensation package in non-underwritten offerings on the AIM market of the London Stock Exchange. This is the first empirical analysis that examines the use of warrants in non- underwritten offerings. In the US we cannot conduct the same study as the vast majority of best efforts offerings issue warrants to their advisers. This study is also the first one that is focused on the significant role of the broker in the flotation process on AIM.

Our results shed light on the current debate of the dissatisfaction of the CFOs for the costs their companies incur for raising capital through an IPO. Our main finding is that issuers make efficient decisions and are able to choose the contract that minimises their costs of going public. More specifically, for firms that issue warrants, the total costs of going public are 22.74%, but would have been 25.61% had they not issued them. Although this 2.87% reduction in the total costs may seem a small figure, it is equivalent to 70.34% of the commission paid to the brokers. The main factor for the 2.87% savings in total costs is caused by the reduced underpricing the firms incur through the use of compensation warrants (15.95%

51 actual vs. 22.12% estimated underpricing). For firms that do not issue warrants the total costs of going public are similar regardless of the contract they choose to use. Moreover, the probability of issuing warrants is higher for smaller and riskier firms that are advised by less reputable brokers, raise less money from the selling of existing shares in the IPO and the Nomad and the broker are two different advisers.

References

Allen, F., 1984. Reputation and product quality. The Rand Journal of Economics 15, 311- 327.

Antoniou, A., Ergul, N., Holmes, P., 1997. Market Efficiency, Thin Trading and Non-linear Behaviour: Evidence from an Emerging Market. European Financial Management, 3, 175- 190.

Arcot, S., Black, J., Owen, G., 2007. From local to global: the rise of AIM as a stock market for growing companies. London Stock Exchange publication http://www.londonstockexchange.com/companies-and advisors/aim/publications/external/aim-global-local.pdf

Arthurs, J. D., Busenitz, L. W., Hoskisson, R. E., Johnson, R. A., 2009. Signaling and initial public offerings: The use and impact of the lockup period. Journal of Business Venturing 24, 360–372.

Barry, C. B., Muscarella, C. J., Vetsuypens, M. R., 1991. Underwriter warrants, underwriter compensation, and the costs of going public. Journal of Financial Economics, 29, 113-135.

Booth, J. R., Smith, R. L., 1986. Capital raising, underwriting and the certification hypothesis. Journal of Financial Economics, 15, 261-281.

Brau, J. C., Li, M., Shi, J., 2007. Do secondary shares in the IPO process have a negative effect on aftermarket performance? Journal of Banking & Finance, 31, 2612-2631.

Brav, A., Gompers, P. A., 2003. The role of lockups in initial public offerings. Review of Financial Studies 16, 1–29.

Burton, B., Helliar, C., Power, D., 2006. Practitioners’ Perspectives on the IPO Process and the Perils of Flotation. The European Journal of Finance, 12, 671-692.

Byoun, S., Moore, W. T., 2003. Stock vs. stock-warrant units: evidence from seasoned offerings. Journal of Corporate Finance, 9, 575-590.

Chemmanur, T. J., Fulghieri, P., 1997. Why include warrants is new equity issues? A theory of unit IPOs. Journal of Financial and Quantitative Analysis, 32, 1-24.

Dunbar, C. G., 1995. The use of warrants as underwriter compensation in initial public offerings. Journal of Financial Economics, 38, 59-78.

Espenlaub, S., Goergen, M., Khurshed, A., 2001. IPO Lock-in Agreements in the UK. Journal of Business Finance & Accounting 28, 1235-1278.

Fang, L. H., 2005. Investment bank reputation and the price and quality of underwriting services. The Journal of Finance 60, 2729-2761.

Firth, M., Liau-Tan, C., 1998. Auditor quality, signalling and the valuation of initial public offerings. Journal of Business Finance & Accounting 25, 145-165.

Garner, J. L., Marshall, B. B., 2005. Unit IPOs: What the Warrant Characteristics Reveal about the Issuing Firm. The Journal of Business 78, 1837-1858.

Gerakos, J. J., Lang, M. H., Maffet, M. G., 2011. Listing Choices and Self-Regulation: The Experience of the AIM. SSRN eLibrary.

Goergen, M., Khurshed, A., Mudambi, R., 2006. The Strategy of Going Public: How UK Firms Choose Their Listing Contracts. Journal of Business Finance & Accounting, 33, 79- 101.

Goergen, M., Renneboog, L., Khurshed, A., 2006. Explaining the diversity in shareholder lockup agreements. Journal of Financial Intermediation 15, 254–80.

Golubov, A., Petmezas, D., Travlos, N. G., 2012. When It Pays to Pay Your Investment Banker: New Evidence on the Role of Financial Advisors in M&As. The Journal of Finance 1, 271-311.

Hamilton, B. H., Nickelson, J. A., 2003. Correcting for Endogeneity in Strategic Management Research. Strategic Organisation 1, 51-78.

Heckman, J. J., 1979. Sample selection bias as a specification error. Econometrica 47, 153- 161.

How, J. C. Y., Howe, J. S., 2001. Warrants in Initial Public Offerings: Empirical Evidence. Journal of Business, 74, 433-457.

Huang, C. J., Chao, C. H., Liao, T. L., 2009. The joint decision to signal through IPO underpricing and lockup. Applied Economics Letters 17, 955–961.

Institutional Investor Committee, 2011. Best Practice Guidance for Issuers when Raising Equity Capital.

Jain, B. A., Kini, O., 1999. On investment banker monitoring in the new issues market. Journal of Banking & Finance 23, 49-84.

Jenkinson, T., Jones, H., 2004. Bids and Allocations in European IPO Bookbuilding. The Journal of Finance 59, 2309-2338.

Jensen, M., 1986. The agency cost of free cash flow, corporate finance and takeovers. American Economic Review 76, 323-329.

Jog, V., Mcconomy, B. J., 2003. Voluntary disclosure of management earnings forecasts in IPO prospectuses. Journal of Business Finance and Accounting 30, 125-167.

Klein, B., Leffler, K. B., 1981. The role of market forces in assuring contractual performance. Journal of Political Economy 89, 615-641.

Lee, L.-F., 1978. Unionism and Wage Rates: A Simultaneous Equations Model with Qualitative and Limited Dependent Variables. International Economic Review 19, 415-433.

Lee, M., Lee, P., Taylor, S., 2003. Unit initial public offerings: Staged equity or signalling mechanism? Accounting and Finance 43, 63-85.

Leland, H. E., Pyle D. H., 1977. Information asymmetries, financial structure and financial intermediaries. Journal of Finance 32, 317-287.

Li, K., Prabhala, N. R., 2007. Self-Selection Models in Corporate Finance. Handbook of Corporate Finance: Empirical Corporate Finance, 47-62.

Liu, C. S., Ziebart, D.A., 1999. Anomalous security price behavior following management earnings forecasts. Journal of Empirical Finance 6, 405–430.

Logue, D. E., Lindvall, J. R., 1974. The Behavior of Investment Bankers: An Econometric Investigation. The Journal of Finance, 29, 203-215.

London Stock Exchange, 2007a. AIM Rules for Nominated Advisers. http://www.londonstockexchange.com/companies-and-advisors/aim/publications/rules- regulations/aim-rules-for-nominated-advisers.pdf

London Stock Exchange, 2007b. From local to global. The rise of AIM as a stock market for growing companies. http://eprints.lse.ac.uk/23110/1/fromLocalToGlobal_Sept2007.pdf

London Stock Exchange, 2007c. Joining AIM: A Professional Handbook. http://www.faegre.com/files/22472_Joining%20AIM%202.pdf

London Stock Exchange, 2009. AIM admission timetable – an example. http://www.londonstockexchange.com/companies-and-advisors/aim/for- companies/joining/admission-timetable.pdf

London Stock Exchange, 2010. AIM Rules for Companies. http://www.londonstockexchange.com/companies-and-advisors/aim/advisers/rules/aim- rules-for-companies.pdf

Lowry, M., Shu, S., 2002. Litigation risk and IPO underpricing. Journal of Financial Economics 65, 309–335.

Maddala, G. S., 1983. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge University Press.

Main, B., Reilly, B., 1993. The employer size-wage gap: evidence for Britain. Economica 60, 125–142.

Mallin, C., Ow-Yong, K., 2011. Factors influencing corporate governance disclosures: evidence from Alternative Investment Market (AIM) companies in the UK. The European Journal of Finance 18, 515-533.

Mallin, C., Ow-Yong, K., 2010. The UK Alternative Investment Market – Ethical Dimensions. Journal of Business Ethics 95, 223-239.

Mazouz, K., Saadouni, B., Yin, S., 2008. Warrants in IPOs: Evidence from Hong Kong. Pacific-Basin Finance Journal 16, 539-554.

Megginson, W. L., Weiss, K. A., 1991. Venture Capitalist Certification in Initial Public Offerings. The Journal of Finance 46, 879-903.

Mendoza, J. M., 2008. Securities Regulation in Low-tier Listing Venues: The rise of the Alternative Investment Market. Fordham Journal of Corporate & Financial Law 13, 257 – 328.

Miller, M. H., Muthuswamy, J., Whaley, R. E., 1994. Mean Reversion of Standard & Poor's 500 Index Basis Changes: Arbitrage-Induced or Statistical Illusion? The Journal of Finance, 49, 479-513.

Ng, C. K., Smith, R. L., 1996. Determinants of Contract Choice: The Use of Warrants to Compensate Underwriters of Seasoned Equity Issues. The Journal of Finance 51, 363-380.

Office of Fair Trading, 2011. Equity underwriting and associated services. http://www.oft.gov.uk/shared_oft/market-studies/OFT1303.pdf

PricewaterhouseCoopers, 2012. Considering an IPO? The costs of going and being public may surprise you. http://www.pwc.com/en_US/us/transaction- services/publications/assets/pwc-cost-of-ipo.pdf

Ritter, J. R., 1987. The costs of going public. Journal of Financial Economics 19, 269-281.

Schultz, P. 1993. Unit initial public offerings: a form of staged financing. Journal of Financial Economics 34, 199-229.

Shapiro, C., 1983. Premiums for high quality products as returns to reputations. The Quarterly Journal of Economics 98, 659-679.

Skinner, D., 1994. Why firms voluntarily disclose bad news. Journal of Accounting Research 32, 38–60.

Trueman, B., 1986. Why do managers voluntarily release earnings forecasts? Journal of Accounting and Economics, 53-72.

Tucker, J., 2010. Selection bias and econometric remedies in accounting and finance research. Journal of Accounting Literature 29, 31–57.

Vismara, S., Paleari, S., Ritter, J. R., 2012. Europe's Second Markets for Small Companies. European Financial Management 18, 352-388.

Table 1 Differences in the admission criteria required for listing on AIM and MM

This table presents the main differences in the listing criteria between the Alternative Investment Market (AIM) and the Main Market (MM) of the London Stock Exchange.

Rules AIM MM

25% is the minimum level of shares Public float No regulation. that should be in public hands.

Trading record No regulation. 3 years trading record.

Shareholder approval in order for No prior approval is required for most Approval for transactions substantial acquisitions or disposals transactions. to take place.

Prospectus or Admission Admission document approved by Prospectus approval by UK Listing Document Nomad. Authority. Requirement of sponsors for certain Existence of a broker and a Nomad at transactions (i.e. purchase of own Adviser(s) all times. shares, transfer between listing categories, etc.). Minimum level of market value Market Value No regulation. (£700,000).

Table 2 Comparison between warrant and no-warrant IPO firms First day return/Underpricing is the 1st day return and is calculated as (closing price – issue price)/issue price. MAIR is the market adjusted first day return and is calculated as (first day return – FTSE AIM all share index return at the day the IPO took place). Inv. Issue Price is the inverse of the issue price. Age is the number of years from incorporation to flotation on AIM. Public Float is the shares sold in the IPO divided by outstanding shares. Retain. Owner. is the level of ownership retained by insiders of the IPO firm after the IPO. GP is the gross proceeds and is calculated as the offer price multiplied by the total number of shares sold in the IPO. Secondary Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). MC is the market capitalisation of the company. Volatility (20 days) is the standard deviation of the company’s returns for 20 days in the aftermarket. Recursive Vol (126 days) is the standard deviation of the company’s returns for 126 days in the aftermarket adjusted for thin trading. Vol (FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date. Ret (FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. Commission is the money paid to the brokers for procuring subscribers and buyers for the new and existing shares respectively. Total Broker Comp. is the summation of the commission and the value of warrants. The commission, value of warrants and total broker compensation are expressed as a percentage of gross proceeds. Dyn. Broker Rep is the dynamic broker reputation and is measured based on the total gross proceeds raised and the number of IPOs brought to the market by each broker over the previous 3.5 years. Stat. Broker Rep is the static broker reputation measure. It is based on the total gross proceeds and number of IPOs raised and brought to the market by each broker for the period from January 1999 to December 2008. N is the number of IPOs. Difference Difference Warrants = 0 Warrants = 1 Total Sample in Means in Medians Mean Median Mean Median Mean Median (p-value) (p-value) Underpricing (%) 19.73 10.38 15.95 9.06 18.09 10.00 0.21 0.72 MAIR (%) 19.67 10.36 15.9 9.1 18.03 9.97 0.22 0.79 Issue Price (£) 0.94 0.85 0.66 0.45 0.82 0.65 0*** 0*** Inv. Issue Price 3.46 1.18 5.82 2.25 4.48 1.54 0*** 0*** Age (years) 2.75 0.67 2.41 0.70 2.60 0.69 0.44 0.41 Public Float (%) 31.84 29.84 29.97 28.46 31.03 29.35 0.25 0.50 Retain. Owner. (%) 68.16 70.16 70.03 71.54 68.97 70.65 0.25 0.50 GP (£millions) 10.6 6.22 8.47 4.31 9.7 5.48 0.07* 0*** Secondary Proceeds 10.06 0 4.97 0 7.85 0 0*** 0*** (%) MC (£millions) 35.7 22.1 28.7 18.4 32.7 20.7 0.07* 0.01** Volatility (20 days, 2.34 1.55 2.54 1.97 2.43 1.76 0.38 0.04** %) Recursive Vol (126 3.21 2.24 5.85 2.84 4.35 2.47 0.25 0.02** days, %) Vol (FTSE AIM 0.67 0.54 0.66 0.53 0.67 0.53 0.77 0.87 index, %) Return (FTSE AIM 0.86 1.93 0.43 1.68 0.68 1.93 0.65 0.86 index, %) Commission (%) 3.26 3.44 4.08 4.27 3.61 3.80 0*** 0*** Value of Warrants 2.72 1.25 (%) Total Broker Comp. 3.26 3.44 6.8 5.83 4.8 4 0*** 0*** (%) Dyn. Broker Rep. 0.045 0.028 0.029 0.015 0.038 0.019 0*** 0*** (GP) Dyn. Broker Rep. (N 0.032 0.033 0.025 0.024 0.029 0.03 0*** 0*** IPOs) Stat. Broker Rep. 0.036 0.025 0.026 0.015 0.032 0.02 0*** 0*** (GP) Stat. Broker Rep. (N 0.034 0.031 0.029 0.024 0.032 0.028 0*** 0.03** IPOs) N 288 221 509 ***, ** and * indicate statistically significant at 1%, 5%, 10% significance levels

Table 3 Correlation Matrix This table shows the pairwise correlations matrix of the independent variables that will be used in the two stage probit model. Public Float (Public_Float) is the ratio of the total number of shares sold in the IPO divided by the outstanding shares. MC (MC) is the natural logarithm of the market capitalisation of the company. Secondary Proceeds (Secondary_Proceeds) is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Br. Reput. (Broker_Rep. GP and N of IPOs) is the reputation of the broker based on the gross proceeds and number of IPOs raised and advised by each broker over the previous 3.5 years. Dummy N equal Br. (Dummy_N_Br) is a dummy variable that takes the value of one if the broker and the Nomad are the same firm and zero otherwise. Age (Age) is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). Vol (FTSE AIM index) (Vol_(FTSE AIM index)) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date. Ret (FTSE AIM index) (Ret_(FTSE AIM index)) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. In brackets we report the names of the variables as they appear in the rest of the tables. Public Secondary Br. Reput. Br. Reput. (N Dummy N Vol (FTSE Ret (FTSE AIM MC Age Float Proceeds (GP) IPOs) equal Br. AIM index) index) Public 1 Float MC -0.07 1 Secondary 0.15 0.2 1 Proceeds Br. Reput. 0.08 0.16 0.17 1 (GP) Br. Reput. 0.05 0.08 0.1 0.71 1 (N IPOs) Dummy N 0.01 0.24 0.1 0.22 0.27 1 equal Br. Age -0.1 0.11 0.17 -0.01 -0.03 0.07 1 Vol (FTSE AIM 0.01 0.02 0 -0.03 0.03 -0.13 -0.04 1 index) Ret (FTSE AIM 0.02 0.05 -0.03 -0.05 0.04 0.04 0 -0.43 1 index)

Table 4 Descriptive statistics for the 242 warrants issued to the brokers Size of warrants is measured as (number of shares that can be purchased under the warrant)/(IPO shares). Value of warrants is obtained from the CEV model and is measured as a percentage of gross proceeds. Exercise/offer price is calculated as (price at which the warrant can be exercised)/(offer price). Time to expiration (years) is measured as the number of years between the date of listing on AIM and the expiration date of the warrant. Lock-in (years) is the time period during which the warrant cannot be exercised. Min and Max are the minimum and maximum values respectively. N is the number of warrants offered. Panel A: All warrants Size of Value of Exercise/offer Time to expiration Lock-in

warrant (%) warrant (%) price (years) (years) mean 7.3 2.5 1.02 4 1.07 median 4.65 1.14 1 3.63 1 Min 0.14 0 0.39 1 0. 5 Max 74.7 67 2.33 10 2 N 242 242 242 242 26 Panel B: Warrants with no lock-in period (N=216) mean 7.63 2.64 1.01 3.82 median 4.98 1.18 1 3 Panel C: Warrants with lock-in period (N=26) mean 4.37 1.29 1.08 5.43 1.07 median 2.97 0.74 1 5 1 Panel D: Warrants with an exercise price equal to the offer price (N=214) mean 7.67 2.7 1 3.96 median 4.98 1.23 1 3 Panel E: Warrants with an exercise price lower than the offer price (N=7) mean 4.34 2.35 0.7 5.94 median 3.21 2.29 0.75 5 Panel F: Warrants with an exercise price higher than the offer price (N=21) mean 4.28 0.43 1.36 3.67 median 2.92 0.018 1.25 4 In Panel A, the maximum size of the warrants is 74.4% of the shares offered in the IPO because one company (Mos International) issued 9,687,500 shares and a warrant to subscribe for 7,239,375 ordinary shares.

Table 5 Probit regression model for the contract equation The dependent variable is a dummy variable that takes the value of one if the firm issues compensation warrants to its broker(s) and zero otherwise. Public_Float is the ratio of the total number of shares sold in the IPO divided by the outstanding shares. MC is the natural logarithm of the market capitalisation of the company. Secondary_Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Broker_Rep. (GP and N of IPOs) is the reputation of the broker based on the gross proceeds and number of IPOs raised and advised by each broker over the previous 3.5 years. Dummy_N_Br is a dummy variable that takes the value of one if the broker and the Nomad are the same firm and zero otherwise. Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). Vol_(FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date multiplied by 100. Ret_(FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. The variables public float, secondary proceeds, broker rep., vol and ret (FTSE AIM index) are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. Model 1 Model 2 marginal marginal coef p-value coef p-value effect effect Intercept 2.27* 0.06 2.62** 0.03 Public_Float -0.20 0.56 -0.08 -0.24 0.48 -0.09 MC -0.14** 0.03 -0.06 -0.15** 0.02 -0.06 Secondary_Proceeds -0.64* 0.08 -0.25 -0.67* 0.07 -0.26 Broker_Rep. (GP) -2.47** 0.05 -0.97 Broker_Rep. (N of IPOs) -0.80** 0.02 -0.31 Dummy_N_Br -0.42*** 0.01 -0.16 -0.36** 0.02 -0.14 Age -0.01 0.93 -0.001 -0.01 0.90 -0.001 Vol_(FTSE AIM index) 0.12 0.59 0.05 0.12 0.59 0.05 Ret_(FTSE AIM index) 0.66 0.37 0.26 0.72 0.33 0.28 Yearly_Dummies Yes Yes % correct predictions 70% 70% Pseudo R square 0.09 0.1 N 509 509 ***, ** and * indicate statistically significant at 1%, 5%, 10% significance levels

Table 6 Second-stage regression estimates of underpricing and total broker compensation Underpricing is the first-day return and is calculated as (closing price – issue price)/issue price. Total Broker Compensation is the summation (Commission + Warrant Value)/Gross Proceeds. Public_Float is the ratio of the total number of shares sold in the IPO divided by the outstanding shares. MC is the natural logarithm of the market capitalisation of the company. Secondary_Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Broker_Rep. (GP) is the reputation of the broker based on the gross proceeds raised by each broker over the previous 3.5 years. Dummy_N_Br is a dummy variable that takes the value of one if the Broker and the Nomad are the same firm and zero otherwise. Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). Vol_(FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date multiplied by 100. Ret_(FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. IMR is the inverse Mills ratio which is used to adjust for selectivity bias. For the IPOs that issue warrants the IMR is defined as -φ(ψ)/Φ(ψ) and for those which they do not is φ(ψ)/(1-Φ(ψ)) respectively. φ(ψ) is the standard normal density function and Φ(ψ) is the standard normal cumulative distribution function. The variables public float, secondary proceeds, broker rep., vol and ret (FTSE AIM index) are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. Dependent Variable: Dependent Variable:

Underpricing Total broker Compensation Contracts Contracts with Contracts without Contracts with without Warrants Warrants Warrants Warrants Equation 1 Equation 2 Equation 1 Equation 2 p- p- coef p-value coef p-value coef coef value value Intercept 0.8*** 0.01 -0.65 0.4 0.39** 0.03 0.14** 0.01 - Public_Float -0.04 0.66 0 -0.10** 0.02 -0.003 0.54 0.45*** - MC -0.04** 0.04 -0.05 0.21 -0.03 0.22 0.03 0.01** Secondary_Proceeds -0.12 0.3 -0.07 0.63 -0.08 0.48 -0.01 0.28

Broker_Rep. (GP) -1.25** 0.01 -0.02 0.98 -0.37 0.42 -0.06* 0.1 - Dummy_N_Br -0.04 0.50 0.03 0.02** Age -0.04** 0.03 -0.03 0.4 -0.01 0.27 0.00 0.87 Vol_(FTSE AIM 0.11** 0.06 0.16** 0.05 0.02 0.39 0.003 0.27 index) Ret_(FTSE AIM 1.12*** 0 0.35 0.24 0.11 0.35 0.02* 0.05 index) IMR -0.21 0.12 0.44* 0.08 -0.17 0.48 -0.04* 0.09

Yearly_Dummies Yes Yes Yes Yes

R square 0.25 0.14 0.13 0.28 N 221 288 221 288 ***, ** and * indicate statistically significant at 1%, 5%, 10% significance levels

Table 7 Comparison of the actual costs with the estimated costs had the alternative compensation contract been used This table compares the average underpricing and total broker compensation costs with the estimated costs had the alternative contract been used. Underpricing is the first-day return and is calculated as (closing price – issue price) / issue price. Total Broker Comp. is the total broker compensation and is calculated as: (Commission + Warrant Value)/Gross Proceeds. Total Costs is the summation: (Underpricing + Total Broker Comp.). N is the number of observations. Average cost estimates for the 221 IPOs that Average cost estimates for the 288 IPOs

issued warrants to brokers that did not issue warrants to brokers Estimated cost if Difference Estimated cost if Difference Actual Actual warrants had not been in means warrants had been in means cost cost issued to brokers p-value issued to brokers p-value Underpricing 15.95 22.12 0*** 19.73 16.92 0.24 (%) Total Broker 6.79 3.49 0*** 3.26 6.11 0*** Comp. (%) Total Costs 22.74 25.61 0.1* 22.99 23.03 0.99 (%) N 221 221 288 288

***, ** and * indicate statistical significance at the 1%, 5% and 10% significance levels.

Appendix

Table 1 Probit regression model for the contract equation The dependent variable is a dummy variable that takes the value of one if the firm issues compensation warrants to its broker(s) and zero otherwise. Retain._Owner. is the level of ownership retained by insiders of the IPO firm after the IPO. MC is the natural logarithm of the market capitalisation of the company. Secondary_Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Broker_Rep. (GP) is the reputation of the broker based on the gross proceeds raised by each broker over the previous 3.5 years. Dummy_N_Br is a dummy variable that takes the value of one if the broker and the Nomad are the same firm and zero otherwise. Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). Vol_(FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date multiplied by 100. Ret_(FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. The variables retain. owner., secondary proceeds, broker rep., vol and ret (FTSE AIM index) are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. Model 1 Model 2 marginal marginal coef p-value coef p-value effect effect Intercept 2.07* 0.08 2.38** 0.04 Retain. Owner. 0.20 0.56 0.08 0.24 0.48 0.09 MC -0.14*** 0.03 -0.06 -0.15** 0.02 -0.06 Secondary_Proceeds -0.64* 0.08 -0.25 -0.67* 0.07 -0.26 Broker_Rep. (GP) -2.47** 0.05 -0.97 Broker_Rep. (N of IPOs) -0.80** 0.02 -0.03 Dummy_N_Br -0.42*** 0.01 -0.16 -0.36** 0.02 -0.14 Age -0.01 0.93 -0.003 -0.01 0.90 -0.004 Vol_(FTSE AIM index) 0.12 0.59 0.05 0.12 0.59 0.05 Ret_(FTSE AIM index) 0.66 0.37 0.26 0.72 0.33 0.28 Yearly_Dummies Yes Yes % correct predictions 70% 70% Pseudo R square 0.09 0.1 N 509 509 ***, ** and * indicate statistically significant at 1%, 5%, 10% significance levels

Table 2 Comparison of actual costs with the estimated costs had the alternative compensation contract been used This table compares the average MAIR and total broker compensation costs with the estimated costs had the alternative contract been used. MAIR is the market adjusted first day return and is calculated as (first day return – FTSE AIM all share index return at the day the IPO took place). Total Broker Comp. is the total broker compensation and is calculated as: (Commission + Warrant Value)/Gross Proceeds. Total Costs is the summation: (MAIR + Total Broker Comp.). N is the number of observations. Average cost estimates for the 221 IPOs Average cost estimates for the 288 that issued warrants to brokers IPOs that did not issue warrants to brokers Actual Estimated Difference in Actual Estimated Difference cost cost if means cost cost if in means warrants had p-value warrants p-value not been had been issued to issued to brokers brokers MAIR 15.9 22.15 0*** 19.67 16.79 0.24 Total Broker Comp. 6.79 3.42 0*** 3.26 6.11 0*** (%) Total Costs (%) 22.69 25.57 0.1* 22.93 22.9 0.98 N 221 221 288 288 ***, ** and * indicate statistical significance at the 1%, 5% and 10% significance levels.

Chapter 3

Warrants in Underwritten IPOs

Abstract

We examine the use of non-cash compensation in a regulatory environment that is very different from that of the US. Our results show that, though warrant-issuing UK IPO firms are riskier, they are underwritten by reputable brokers. In addition, brokers appear to time the issuance of warrants because they include them as part of their compensation package mainly when the market is doing well. Furthermore, the probability of issuing warrants is higher for firms that are cash constrained. Interestingly, warrant issuers are still able to minimise their total costs of going public, even under a very light regulatory setting underlying non-cash compensation. They incur an underpricing and a total broker compensation of 23.3% and 5.6% respectively. These costs would have been 40.5% and 3.46% had warrants not been used. The results also show that, on average, brokers enhance their underwriting fees by about 75% as a result of the warrants being part of the compensation package.

1 Introduction

Dunbar (1995) finds that firm commitment initial public offerings (IPOs) in the United States (US) minimise their costs of going public by issuing warrants as part of the compensation package to the underwriters. Dunbar (1995) also suggests that the National Association of Securities Dealers (NASD) should relax their non-cash compensation regulations (i.e. warrants) as they are unnecessarily restrictive. In 2004 NASD actually relaxed some of these regulations. Our study examines the use of non-cash compensation in underwritten IPOs outside the US market and tries to answer the following two questions: In an institutional setting that is very different from that of the US do IPO firms still minimise their costs of going public? In an institutional setting in which there are no regulations restricting the underwriters’ compensation, why do financial intermediaries include warrants in their compensation package and do not get paid all the fees in cash?

In this paper, we focus on the issue of warrants as part of the brokers’ compensation package in underwritten IPOs listed in the United Kingdom (UK). The UK institutional setting is very different from that of the US. While, in the US, the use of warrants is constrained by regulatory requirements regarding the exercise price, the lock-in period and the minimum value of warrants, the Alternative Investment Market (AIM) of the London Stock Exchange (LSE) carries none of these regulatory constraints. AIM, therefore, offers a perfect laboratory in which to study the unfettered use of non-cash compensation to IPO brokers.

We study only the issue of warrants to the brokers because they explicitly guarantee that they will subscribe/purchase shares for which they are not able to procure enough investors. So, the brokers are essentially the underwriters of the IPOs. In addition, brokers play a pivotal role in the IPO process on AIM12 as they are responsible for advising IPO firms on the pricing of shares, assessing the level of investor’s interest and providing information on market and trading related issues (London Stock Exchange, 2007b).

We include only underwritten IPOs in our sample because we want to compare our results with those of previous studies. The empirical literature that examines the use of warrants as part of the underwriters’ compensation is limited to firm commitment offerings in the US market. Although, firm commitment offerings may have differences when compared to underwritten IPOs, in both cases the underwriting obligation is the same as the underwriters

12 We do not provide any detailed description of the AIM market and the role of the broker during the flotation process in order to avoid repetition with the previous chapter.

69 have to buy any unsold IPO shares. We do not include IPOs listed in the Main Market of the LSE because only 4% of them issue warrants to their underwriters. In contrast, this figure is much higher for AIM IPOs in which nearly one third of them grant non-cash compensation to their underwriters.

The contribution of our paper is three-fold. First, to the best of our knowledge, this is the first study to examine the non-cash compensation of the brokers on underwritten IPOs in an environment with almost no regulatory constraints on the issuance of warrants. Second, our study examines and finds supportive evidence of hypotheses that have never been tested in the existing empirical literature. More specifically, we test whether companies issue warrants to their brokers because they are cash constrained. Finally, our results have important policy implication for the market regulators.

Dunbar (1995) supports that the NASD should relax the regulations underlying non-cash compensation because they are unnecessarily restrictive. Our findings further suggest that in an environment with almost no regulations underlying the use of warrants, firms are still able to minimise their costs of going public. More specifically, warrant IPO firms incur an average underpricing of 23.3% and a total broker compensation of 5.6%. These costs would have been 40.5% and 3.46% respectively had they not issued warrants. Overall, if we add up the two costs (underpricing + total broker compensation) then IPO firms incur actual total costs of 28.9%, but would have been almost double, 43.96%, had they chosen not to grant warrants. The total brokers’ compensation is made up of an average underwriting commission of 3.2% and an average value of warrants worth 2.4% of the gross proceeds. This implies that brokers enhance their compensation fees by 75% (2.4%/3.2%) via the use of warrants as part of their underwriting fees.

Our regression analysis shows that companies that are cash constrained usually issue compensation warrants. In addition, brokers appear to time the issuance of warrants because they include them as part of their compensation package mainly when the market is doing well. The two aforementioned arguments (cash constrained IPOs and timing of issuance) may also explain why brokers do not get paid all their fees in cash, but also include warrants in their compensation. Our results also show that warrant-issuing IPOs are underwritten by reputable brokers. This is in contrast to Dunbar’s (1995) finding, who reports that warrant- issuing firms are underwritten by less reputable financial intermediaries. Furthermore, the

70 probability of the use of warrants is higher for riskier firms, which are younger, smaller and have a higher aftermarket standard deviation of returns (20 days).

The remainder of this paper is organised as follows. Section 2 discusses the literature on the use of warrants as part of the compensation paid to the underwriters. Section 3 compares the regulations underlying the non-cash compensation between the UK and US markets. Section 4 provides our hypotheses, while Section 5 describes our data. Section 6 explains our methodology. Section 7 presents our results and Section 8 concludes.

2 Literature review

The empirical literature that examines the use of warrants as part of the underwriters’ compensation is limited to firm commitment offerings in the US. Barry et al. (1991) find that warrants are mainly used in smaller, younger and riskier IPOs that are difficult and costly to market. They also find, in line with the circumvention hypothesis, that investment banks, in order to avoid the binding regulatory limits on underwriting compensation, use warrants instead of charging the issuers higher cash fees. The acceptance of warrants as part of the compensation package for underwriters in the US may, inadvertently, have been encouraged by the fact that the NASD’s pricing formula undervalues warrants when compared to the Black and Scholes and Constant Elasticity Variance (CEV) models. This is mainly due to the fact that the model does not take into account the volatility of the IPO shares. This means that, whenever underwriters bring risky issues to the market, they include warrants in their compensation, instead of charging the issuers a high cash fee that would violate/exceed NASD guidelines (circumvention theory). Furthermore, Barry et al. (1991) report that warrant IPOs have lower offer prices and exhibit higher underpricing than the non-warrant ones. This is consistent with Logue’s (1973) argument that underwriters have the incentive to underprice the offering by setting the offer price too low. In this way the value of warrants will increase because warrants align the underwriters’ compensation with the aftermarket price performance. In addition, Barry et al. (1991) report that the total costs of going public can be as much as 30% of the gross proceeds of the offering because underwriters’ compensation is higher for the warrant IPOs sample when compared to the non-warrant one. However, Barry et al. (1991) fail to control for sample selection bias in their analysis.

Dunbar (1995), who examines US firm commitment offerings during the period 1980 - 1983 and takes into account the self-selection bias in his sample, finds that, for issuers who use

71 warrants, the total costs of raising capital are lower than they would have been if warrants had not been used. Moreover, his results support the cost minimisation hypothesis, according to which issuers choose the type of contract that minimises their costs. Thus, underwriter warrants are chosen because they are considered a credible signal that the offering will not be overpriced (underwriter certification). This means that investors will require a smaller discount on the new issue, reducing the underpricing of the IPO and consequently the total costs of going public.

Ng and Smith (1996) use a two-stage logit model to account for self-selection. They find evidence that issuers select contracts that maximise their net proceeds. The total underwriter costs would have been much higher had the issuers not used warrants. That is to say, net proceeds would have been lower if warrants had not been used. Ng and Smith (1996) also find evidence in support of the certification hypothesis since less well established underwriters, who lack reputational capital, certify the offer by accepting warrants as part of their compensation. In this way underwriters mitigate the information asymmetry problem that the issue may be overpriced, because their own compensation is tied up to the aftermarket price performance. Moreover, consistent with previous studies, Ng and Smith (1996) show that warrants are mainly used by small and risky companies that have significant growth opportunities. Overall, the authors suggest that certification has a much greater effect on the decision to use warrants than circumvention.

Bae and Jo (2007) report that underwriter compensation warrants are used by firms in order to signal their future growth opportunities. More specifically, Bae end Jo (2007) find that the abnormal returns from the announcement of seasoned equity offerings (SEOs) with warrant based compensation are significantly less negative than those of SEOs with cash based compensation. This implies that when warrants are issued to the underwriters as part of their compensation package, during an SEO, they signal positive information of the issuing company’s future prospects.

3 Non-cash and broker/underwriter compensation regulations between the UK and the US markets

Table 1 reveals that the regulatory set-up regarding non-cash (i.e. warrants) and total underwriting compensation are very different in the LSE (both Main and AIM markets) and the US stock exchanges. While, in the LSE, there are very few rules concerning the issuance of warrants, the US stock exchanges prescribe a number of regulatory requirements relating

72 to non-cash compensation. This implies that on AIM we can conduct a unique study on the issue of warrants in an environment in which there are almost no regulations underlying their use. FINRA13 Rule 5110, previously known as NASD Rule 271014, sets out the rules on the characteristics and use of warrants as part of underwriters’ compensation packages in US IPOs (Garner and Marshall, 2010).

According to Schedule One of the AIM rules for Nominated Advisers (London Stock Exchange, 2007a), neither the Nomad nor any of its partners (i.e. the broker) can be, either individually or collectively, a substantial shareholder in a company they advise. These advisers cannot hold 10% or more of the company’s share capital, taking into account any warrants. However, in the US market there is no such regulation. Furthermore, the warrant exercise period can be up to five years in the US (Barry et al., 1991, Ng and Smith, 1996) but there is no such requirement in the LSE.

The underwriter in the US cannot exercise the warrants within the first 180 days after the IPO date15. In the AIM market there is no lock-in period and warrants can be exercised from the first day of trading.16 In both the LSE and the US markets, there is no pre-set minimum requirement relating to the exercise price of the warrants. Since 2004, the minimum value of warrants in the US has been set at 0.2% of the gross proceeds for each amount of securities that is up to 1% of the securities being offered to the public (National Association of Securities Dealers, 2004). There is no such regulatory requirement in the UK.

On AIM, there is no regulation setting a maximum or minimum limit for brokers’ compensation. The amount of commission17 and annual retainer18 fees the broker is paid depends on its initial agreement with the issuer. This applies to both underwritten and non- underwritten offerings. In contrast to the LSE, in the US stock exchanges underwriter

13 FINRA stands for Financial Industry Regulatory Authority and is the largest independent securities regulator in the US market. 14 NASD and New York Stock Exchange (NYSE) regulations were consolidated under the FINRA in July 2007. FINRA adopted most of NASD Rule 2710 as FINRA Rule 5110 on 16 July 2008. 15 There are some exceptions to the lock-in period reported in FINRA Rule 5110. For instance, if underwriters hold an aggregate amount of the issuer’s securities less than or equal to 1% of the securities being offered, then they are not subject to the lock-in period. 16 The only restriction in the AIM market, according to Schedule One of the AIM rules for Nominated Advisers, is that neither the Nomad nor any of its partners can conduct any transactions on the securities of the company during any close period (the close period usually refers to the two months preceding the publication of the company’s annual results). 17 Fee paid to the brokers for procuring investors for the IPO shares and is expressed as a percentage of the gross proceeds. 18 Fee paid annually for retaining an adviser on an ongoing basis.

73 compensation is expected to be “fair and reasonable”. More specifically, the maximum compensation guidelines (including warrants) are different for firm commitment and best- effort offerings (Notice 92 – 53, FINRA Manual) and vary according to the risk assumed by the underwriter (firm commitment or best-effort offerings) and inversely to the gross proceeds (FINRA Rule 5110). For instance, if the money raised from the IPO is $25 million then the maximum proposed compensation for the underwriter is 7.29% or 6.68% of the gross proceeds for firm commitment and best-effort offerings respectively (Notice 92 – 53, FINRA Manual).

However, the SEC and the NASD in the US have relaxed some of the guidelines on non- cash compensation, bringing them closer to those of the LSE. More specifically, before 1996, the exercise price for underwriter warrants was usually set at 20% above the issue price because most state security (“blue sky”) laws in the US had this requirement (Barry et al., 1991, Ng and Smith, 1996). However, in 1996, the Congress amended Section 18 of the Securities Act of 1933 so that securities listed on the NYSE, the AMEX and the NASDAQ National Market 19 were exempted from state “blue sky” laws (National Securities Improvement Act of 1996). Securities listed on the NASDAQ Capital Market were also made exempt from the “blue sky” requirements on 24 March 2007.20 In addition, the NASD made some amendments to Rule 2710 (the Corporate Financing Rule) that became effective from March 2004. One of these was the abolishment of the requirement that the exercise price of warrants should be at least equal to the issue price of the offering (National Association of Securities Dealers, 2004). In addition, the NASD reduced the lock-in period for exercising the warrants from one year to 180 days. Another amendment ensured that warrants issued to underwriters as compensation could exceed the limit of 10% of the gross proceeds. This is consistent with Dunbar’s (1995) suggestion that the NASD should relax its guidelines as the 10% limit unnecessarily restricted underwriters’ ability to certify the issue through the use of warrants.

4 Hypotheses

In the existing literature there have been suggested three main theories that try to explain the issue of warrants to underwriters and these are the following: the circumvention (Barry et

19 The NASDAQ National Market is now called the NASDAQ Global Market, including the NASDAQ Global Select Market. 20 https://listingcenter.nasdaqomx.com/Show_Doc.aspx?File=FAQsInitial.html

74 al., 1991), certification (Booth and Smith, 1986) and cost minimisation (Dunbar, 1995). The circumvention hypothesis is not applicable on AIM as there is no maximum limit regulation related to the broker’s compensation. However, prior literature has failed to take into account the cash that a company has available during the flotation process as a potential explanation of the use of warrants. It may be the case that warrant issuing firms may simply not have enough cash in their balance sheets (cash constrained) and this is why they pay part of their underwriting fees through the issue of warrants. This explanation does not exclude the three aforementioned theories because an IPO firm may be cash constrained and chooses to pay part of its underwriting fees with warrants, but simultaneously these warrants can certify that the issue is not overpriced because they link the underwriting compensation with the aftermarket performance of the company’s stock price. So, our first hypothesis is the following:

Hypothesis 1: The probability of issuing warrants will be higher for cash constrained companies.

It is puzzling why brokers include warrants in their compensation package and not instead get paid all their fees in cash in an institutional setting in which there is no regulation that restricts their cash fees. Although the lack of sufficient cash may provide an explanation, it may also be the case that brokers have the ability to time the issuance of warrants and ask for non-cash compensation when the market is doing well. The rationale is that if the market is doing well then it is likely that the company’s stock price will increase in the aftermarket and the value of warrants will also increase. This is due to the fact that the value of warrants is aligned with the aftermarket performance of the stock price. So, if the price of the company’s shares increases, then the value of warrants will also rise, leading to a higher total compensation for the broker. This argument formulates our next hypothesis:

Hypothesis 2: The probability of issuing warrants will be higher when the market is doing well.

Ng and Smith (1996) and Dunbar (1995) found that companies minimise their costs of going public by using non-cash compensation. The main factor that leads to this cost minimisation is the lower underpricing of the issue. If insiders can credibly send a signal to the market that they are not selling overpriced securities, then investors are likely to require a lower level of underpricing (Dunbar, 1995). One way to achieve that is to compensate the underwriters with warrants. As we mentioned before, the exercise of warrants is directly dependent on the

75 aftermarket stock price performance. If the issue is overpriced, the stock price will drop, the value of the warrants will decrease and, consequently, the underwriters’ compensation will be lower. If the issue is underpriced, then the opposite will occur. Certification through warrants should be more valuable for smaller and riskier firms, which are characterised by greater informational asymmetries, because insiders may be better informed about the true value of the companies than outside investors (Dunbar, 1995). So, if the cost minimisation hypothesis holds then the issuers choose the contract that minimises their costs. So, our third hypothesis is the following:

Hypothesis 3: IPO firms choose the contract that minimises their costs.

5 Data

Our data include all non-financial IPOs listed on AIM over the period from June 1995 to December 2010. After filtering the spreadsheet provided by the LSE to exclude all financial industry listings, the initial sample consists of 1,262 firms. 902 of these 1,262 offerings are not underwritten whereas 10 are partly underwritten (less than 50% of the shares sold are underwritten). When we exclude them, the remaining sample consists of 350 underwritten listings. 3 of them which are categorised as non-financial firms they actually have SIC codes in the range 6000-6999 (financial companies). In addition, 6 of them were listed to other stock exchanges before listing on AIM, so they are not real IPOs.21 When we omit them we are left with 341 underwritten IPOs, which is approximately 27% of all AIM listings. This is in stark contrast to the US market, in which 65.5% of the IPOs are underwritten (Ritter, 1987). We exclude 13 unit IPOs, 4 IPOs due to the unavailability of admission documents, and 8 IPOs that issued warrants to other advisers (i.e. Nomad). This leaves us with a final sample of 316 underwritten IPOs.

We use a sample of 57 out of the 316 IPOs, listed over the period from June 199522 to December 1998, to construct the dynamic broker’s reputation (based on the gross proceeds or the number of IPOs in the 3.5 years before the focal IPO occurs). Thus the remaining sample contains the 259 underwritten IPOs that took place between January 1999 and December 2010.

21 For example, Petmin Limited conducted an IPO on 20 December 2006 but had already been listed on the Johannesburg Stock Exchange in 1986. Similarly, Tricorn Group plc was previously listed on OFEX (renamed the PLUS Market). The aforementioned IPOs are excluded from out sample. 22 The first underwritten IPO was admitted to AIM on 27 June 1995.

The stock price data are extracted from DataStream, Thomson One Banker and Bloomberg. The data regarding the broker’s compensation (commission and warrant characteristics such as shares underlying the warrants, exercise price and time to expiration), issue price, gross proceeds, market capitalisation, date of incorporation, total assets, revenues and cash available the year prior to the IPO are collected from the admission documents.

6 Methodology23

The decision to issue warrants may be a non-random one because IPO firms may choose to self-select the contract that is more favourable for them. So, it may be an endogenous choice made by the decision maker (IPO firm) (Li and Prabhala, 2007). As a result OLS regressions do not yield consistent estimates. In order to control for this selection bias we make use of the two stage endogenous switching model, which is an extension of the baseline Heckman (1979) self-selection model.

We employ a ‘what-if’ type of analysis because we only observe the first day return/underpricing and total broker compensation for companies that issued warrants, but we cannot observe what would the first day return/underpricing and total broker compensation be had the same company chosen not to issue warrants (counterfactual). The first stage regression (selection model) is a probit model that takes the value of one if the companies issue warrants to their brokers and zero otherwise. This model is the following:

퐼푊 = 훾0 + 훾1퐴𝑔푒 + 훾2푆푡푎푛푑푎푟푑_퐷푒푣𝑖푎푡𝑖표푛

+ 훾3푃푢푏푙𝑖푐_퐹푙표푎푡 + 훾4푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠

+ 훾5퐶푎푠ℎ/퐺푃 + 훾6푀퐶 (1)

+ 훾7퐵푟표푘푒푟_푅푒푝. +훾8푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 훾9푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥) + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠 + 휀

where 퐼푊 is a dummy variable that takes the value of one if the firm issues compensation warrants to its broker(s) and zero otherwise, Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age), Standard_Deviation is the standard deviation of the company’s returns over 20 days in the aftermarket multiplied by 100, Public_Float is the ratio of the total number of shares sold in

23 In order to avoid repetition with Chapter 2, we briefly describe the methodology in this part.

77 the IPO divided by the outstanding shares, Secondary_Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds), Cash/GP is the cash and cash equivalents, available the year prior to the IPO, divided by the gross proceeds, MC is the natural logarithm of the market capitalisation and Broker_Rep. (GP and N of IPOs) is the reputation of the broker based on the gross proceeds and number of IPOs raised and advised by each broker over the previous 3.5 years. All reputation measures are dummy variables that take the value one if the IPO is underwritten by one of the 10% most reputable brokers and zero otherwise. Vol_(FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date multiplied by 100, Ret_(FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date, Yearly_Dummies are the yearly dummies for the examined period and 휀 is the stochastic error term.

Some of the independent variables in equation (1) may also have an effect on the underpricing and total broker compensation. The estimated values of the independent variables are then used to generate the inverse Mills ratio (IMR), which is defined differently for the IPOs that issue warrants and for those that do not. The IMR is included in the second stage of the Heckman procedure in which we have two regression equations for the variable of interest conditional on the choice made in the first stage. The second stage equations (OLS regressions) are the following:

푈푛푑푒푟1 = 훽0 + 훽1퐴𝑔푒 + 훽2푆푡푎푛푑푎푟푑_퐷푒푣𝑖푎푡𝑖표푛

+ 훽3푃푢푏푙𝑖푐_퐹푙표푎푡 + 훽4푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠 + 훽5푀퐶

+ 훽6퐵푟표푘푒푟_푅푒푝. +훽7푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥) (2)

+ 훽8푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥) + 훽9퐼푀푅1

+ 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠 + 푢1

푈푛푑푒푟2 = 휃0 + 휃1퐴𝑔푒 + 휃2푆푡푎푛푑푎푟푑_퐷푒푣𝑖푎푡𝑖표푛

+ 휃3푃푢푏푙𝑖푐_퐹푙표푎푡 + 휃4푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠 + 휃5푀퐶

+ 휃6퐵푟표푘푒푟_푅푒푝. +휃7푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥) (3)

+ 휃8푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥) + 휃9퐼푀푅2

+ 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠 + 푢2

where 푈푛푑푒푟1 and 푈푛푑푒푟2 are the first-day returns (underpricing) for the IPOs that issue warrants to their brokers and those that do not respectively and are calculated as: (closing price – issue price)/issue price. The independent variables in equations (2) and (3) have already been explained above. The only difference is that equation (2) includes only the observations of the companies that issue warrants to their brokers whereas equation (3) includes those of the firms that do not.

The independent variables included in 푈푛푑푒푟1 and 푈푛푑푒푟2 models can also be identical with those of the 퐼푊. It is not necessary to apply any exclusion restrictions in the second stage regressions because they are not critical in the Heckman selection model, as this model is identified by the nonlinearity of the IMRs. So, the second stage models are valid even without any exclusion restrictions (Golubov et al., 2012, p. 304).

In practice we only observe either 푈푛푑푒푟1 or 푈푛푑푒푟2, for each IPO, based on the outcome of 퐼푊. The selection bias arises from the non-zero covariance between 휀 from equation (1) and 푢1 and 푢2 from equations (2) and (3). The self-selection regression model solves the aforementioned problems by allowing the error in equation (1) to be correlated with the errors in equations (2) and (3), so that unobserved or missing variables in the binary outcome equation (1) are can also affect the underpricing. Parameters β and 휃 cannot be estimated directly by using OLS because this will generate inconsistent estimates since the expectation of 퐼푊 does not have a zero mean (u1 and ε may be correlated).

This is why in the second stage we estimate equations (2) and (3) by OLS, adding one additional regressor to adjust for the potential non-zero expectation of the errors. This regressor is the IMR, which allows equations (2) and (3) to be estimated consistently using OLS (Lee, 1978, Heckman, 1979). The IMR is defined differently for the firms that issue warrants and for those that do not (Dunbar, 1995).

The same methodology is used to examine the relation between the use of warrants and total broker compensation by replacing the two underpricing equations with two total broker compensation equations. In this case, equations (2) and (3) will include variables that affect the total broker compensation when warrants are or are not used. The second stage OLS regressions for the total broker compensation are the following:

푇표푡푎푙 퐵푟표푘푒푟 퐶표푚푝푒푛푠푎푡𝑖표푛 1 훽0 + 훽1퐴𝑔푒

= + 훽2푆푡푎푛푑푎푟푑_퐷푒푣𝑖푎푡𝑖표푛

+ 훽3푃푢푏푙𝑖푐_퐹푙표푎푡

+ 훽4푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠

+ 훽5퐶푎푠ℎ/퐺푃 + 훽6푀퐶 (4)

+ 훽7퐵푟표푘푒푟_푅푒푝. +훽8푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 훽9푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 훽10퐼푀푅1 + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠

+ 푢1

푇표푡푎푙 퐵푟표푘푒푟 퐶표푚푝푒푛푠푎푡𝑖표푛 2 휃0 + 휃1퐴𝑔푒

= + 휃2푆푡푎푛푑푎푟푑_퐷푒푣𝑖푎푡𝑖표푛

+ 휃3푃푢푏푙𝑖푐_퐹푙표푎푡

+ 휃4푆푒푐표푛푑푎푟푦_푃푟표푐푒푒푑푠

+ 휃5퐶푎푠ℎ/퐺푃 + 휃6푀퐶 (5)

+ 휃7퐵푟표푘푒푟_푅푒푝. +휃8푉표푙_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 휃9푅푒푡_(퐹푇푆퐸 퐴퐼푀 𝑖푛푑푒푥)

+ 휃10퐼푀푅1 + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠

+ 푢2

Where 푇표푡푎푙 퐵푟표푘푒푟 퐶표푚푝푒푛푠푎푡𝑖표푛1 and 푇표푡푎푙 퐵푟표푘푒푟 퐶표푚푝푒푛푠푎푡𝑖표푛2 are calculated as: (Commission + Warrant Value)/Gross Proceeds, for the IPOs that issue warrants to their brokers and for those that do not. The independent variables in equations (4) and (5) have already been explained above. The only difference is that equation (4) includes only the observations of the companies that issue warrants to their brokers whereas equation (5) includes those of the firms that do not.

The independent variables that are included in the probit model (equation (1)) are expected to have an effect on the total costs of going public (underpricing and total broker compensation). The literature suggests that riskier firms with greater ex ante uncertainty are expected to issue warrants to their underwriters (Barry et al., 1991, Dunbar, 1995, Jain and Kini, 1999, Ng and Smith, 1996). For this reason, we include three variables as proxies for ex ante uncertainty: the age of the firm, the standard deviation of returns for the first 20 days following the official listing and the percentage of gross proceeds raised from the selling of existing shares in the IPO (secondary proceeds).

The last aforementioned variable (secondary proceeds) can affect the issuance of warrants in two different ways. It may have a positive relationship with the probability of issuing warrants because the selling of secondary shares in the IPO will send a negative signal to the market regarding insiders’ valuation for the company. As a result, the firms will likely issue warrants to the brokers. This is due to the fact that the brokers can certify the quality of the firm by including warrants in their compensation package and directly link their compensation with the aftermarket performance of the stock price. But, the secondary proceeds may also have a negative effect in the probability of issuing warrants as the selling shareholders can have a greater bargaining power in lobbying for lower listing costs (Dunbar, 1995, Logue and Lindvall, 1974). The rationale is that if the shares are concentrated in the hands of very few shareholders, as is usually the case in smaller firms, then the selling of existing shares will send a positive signal to the market that the company is willing to expand the number of investors that they hold their shares, and is not dependent on a small number of existing shareholders. As a result, insiders can bargain with the broker to pay lower compensation fees.

Burton et al. (2006, p. 678, 688) also report that smaller companies try to obtain a listing on a stock exchange because they want to broaden their shareholder base, since prior to listing there is a very limited number of dominant shareholders. This implies that when the existing shareholders sell shares this may be considered a positive signal for the market, since they are willing to give up some of their holdings and broaden the company’s investor base. Also, Brau et al. (2007) report that the selling of secondary shares in the IPO is not a negative signal for the market. Instead, it is possibly driven by liquidity reasons and not by opportunistically selling overpriced shares in the flotation. This is why when the existing shareholders give up part of their holdings during the flotation, they can achieve a better

81 bargaining power in lobbying for paying lower compensation to their advisers (Dunbar, 1995, Logue and Lindvall, 1974). The reason for achieving lower compensation may be that that the broker(s) will need to put less effort to attract investors for the IPO shares since the existing shareholders have already sent a positive signal to the market.

We also control for the size of the offering as larger IPOs tend to exhibit economies of scales (some parts of the underwriting costs are fixed) (Ng and Smith, 1996). Also, as shown in Notice 92 – 53, FINRA Manual, the underwriting compensation as a percentage of gross proceeds decreases when the issue size increases. In addition, Habib and Ljungqvist (2001) and Loughran and Ritter (2002) report that the gross proceeds and the public float can have an effect on the underpricing as well. This is why we also include the public float in our two stage model

One possible explanation for the use of warrants may be that companies do not have sufficient cash flow at their disposal to pay the broker fees. To address this, we include as an explanatory variable the cash and cash equivalents the company has the year before the IPO, divided by the gross proceeds. It may also be reasonable to suggest that the brokers may include warrants in their compensation package when the market return is high. The rationale is that if the market is doing well (the market/index volatility and return are low and high respectively) then most probably the stock price in the aftermarket will increase. As a result, the value of warrants will increase and consequently the broker compensation (including warrants) will also rise. So, the brokers may be able to time the issuance of warrants and ask for non-cash compensation when the market return is high. We control for the market environment by calculating the volatility and return of the FTSE AIM all share index for two months prior to the IPO.

Barry et al. (1991) and Dunbar (1995) use the Carter and Manaster (1990) ranking to capture the underwriter’s reputation. However, this ranking cannot be used in the case of the LSE since there are no tombstone announcements of equity offerings. Thus, we make use of two different measures of broker reputation, the dynamic and the static reputations. The dynamic reputation is constructed based on the total gross proceeds raised or the number of IPOs advised by each broker for the 3.5 years before the flotation (June 1995 to December 1998), divided by the total gross proceeds raised and number of IPOs listed during the same period. Our sample starts from the middle (June) and not the beginning of the year1995 because the first underwritten IPO took place on AIM on 27 June 1995. The 3.5 years time period is

82 rolled-over up until 2010. For example, during the first 3.5 years that are used to construct our reputational measure (June 1995-December 1998) there were 57 underwritten IPOs. There are two reasons for choosing a period of 3.5 years prior to the IPO to construct the dynamic broker reputation. First, it takes some time for underwriters to build a good reputation. Second, some reputable underwriters may choose not to underwrite any issues in a depressed market in order to avoid damaging their reputation with a poor IPO (Goergen et al., 2006). From 1995 to 2010 there were 64 brokers that advised companies on how to conduct an IPO on the AIM market. The vast majority of these brokers are active only on the UK market. This is why we do not consider a global reputational measure, as global offers usually are not listed on AIM, but on the Main Market of the LSE.

The static reputation, which is used as an alternative to the Carter and Manaster (1990) ranking in the US (Fang, 2005, Megginson and Weiss, 1991), is based on the total gross proceeds raised or the number of IPOs brought to the market throughout the examined period. The logic for using it is that underwriters are repeated players in the market and their survival and future income depend directly on their reputation. For this reason, reputable underwriters will be very selective about the IPOs that they bring to the market throughout their life (Fang, 2005) and will avoid sponsoring overpriced IPOs.

In order to maintain the same number of observations in our probit model (equation (1)) for all different measures of broker reputation, we construct the static reputation from 1999 to 2010 and not from the end of June 1995. However, instead of using the broker reputation as a continuous measure, we discretise it into a binary classification (Fang, 2005). The dummy variable takes the value of one if the IPO is underwritten by one of the top 10% most reputable brokers and zero otherwise.

In the underpricing second stage regression we do not include the variable cash and cash equivalents (as a percentage of gross proceeds). The reason is that we make use of this variable only to examine whether companies that do not have enough cash at their disposal will instead use warrants to pay the brokers’ compensation. So, we expect this variable to have an effect on the compensation paid to the brokers by the company. In all regressions yearly dummies are included.

We employ Cox’s constant elasticity variance model (CEV) to value the warrants.24 We follow Dunbar’s (1995) approach and use the first day’s closing price of the underlying stock rather than its offer price. Barry et al. (1991) calculate the average standard deviation for all the stocks that exist in the CRSP database for a time period of 126 days prior to the offer. Thus, we also compute the average standard deviation of the FTSE AIM all share index across the 126 days before the offering. Barry’s et al. (1991), Dunbar’s (1995), Ng’s and Smith’s (1996) and our measure of volatility have one disadvantage. Seasoned companies are quite different and inherently less volatile than newly listed ones (Barry et al., 1991, Boehme and Çolak, 2012, Clarkson and Thompson, 1990, Ibbotson, 1975). This is why we use two alternative measures of volatility, based on 20 and 126 days of company-specific returns in the aftermarket. We use the Bank of England base rate in the month of the offering as a measure of the risk-free rate.

7 Results

7.1 Descriptive statistics - univariate analysis

In our sample 86 out of the 259 underwritten offerings issue warrants to their brokers and 173 do not (Table 2). Thus, 33.2% of the underwritten IPOs on AIM grant warrants to their brokers. Barry et al. (1991) find that only 17.4% of firm commitment IPOs issue warrants to their underwriters during the period from January 1983 to May 1987, whereas Dunbar (1995) reports a figure of 37.9% in his sample of IPOs listed in the US market during the period from January 1980 to August 1983. Ng and Smith (1996) report that, during the period from 1981 to 1988, 11% of firm commitment seasoned equity offerings (SEOs) issue warrants to their investment banks.

Table 2 reports the descriptive statistics of the warrant and non-warrant subsamples and the total sample of underwritten offerings for the period from 1999 to 2010. The figures show that the IPOs that issue warrants to their brokers as part of the compensation package are more heavily underpriced than their non-warrant counterparts (23.3% vs. 14%). The figures also show that the warrant issuers are younger at the time of listing than the non-warrant issuers (1.89 vs. 4.53 years). The difference in means is statistically significant at conventional significance levels.

24 Due to the fact that the methodology of valuing the warrants is described in details in the previous chapter of this thesis and in order to avoid repetition, we only briefly describe the methodology in this part.

Furthermore, the firms in the warrant IPO group sell fewer shares (public float of 33%) at a lower issue price (£0.83) and have a higher standard deviation of returns (3%) compared to those in the no-warrant group, which have a public float of 42%, an issue price of £1.29 and a standard deviation of 2%. In addition, companies that issue warrants to their brokers have a lower market capitalisation (£43.8 mil. vs. £63.8 mil.) and raise less money in gross proceeds (£11.7 mil. vs. £27.2 mil.). Moreover, the companies that issue warrants to their brokers have less total assets (£5.5 mil.), revenues (£4.3 mil.) and cash (£1.01 mil.) than the no-warrant issuers (£27.1 mil., £30.7 mil. and £3.53 mil. respectively) in the year prior to the IPO. These findings are consistent with Barry et al. (1991), Dunbar (1995), Ng and Smith (1996) and Jain and Kini (1999), who report that issuers who grant warrants as part of their compensation package to the underwriters are on average smaller, have a higher aftermarket standard deviation, offer the shares at lower issue prices and, in general, are riskier and more difficult to market.

The figures reported in Table 2 also reveal that IPOs that issue warrants pay a lower commission (3.2% vs. 3.6%) to their brokers than those in the no-warrant group. The differences in both the means and the medians are statistically significant at the 10% level. These findings are not consistent with those of Barry et al. (1991), who report that underwriter cash compensation is significantly higher for the warrant group of companies. Our results also show that when warrants are included in the compensation package, then the total broker compensation (commission plus value of warrants) is significantly higher for the warrant IPO group (5.6%) than the no-warrant group (3.6%). The differences in both the means and the medians are statistically significant at the 1% level. This finding is consistent with Dunbar (1995), who also reports that the mean underwriter compensation is significantly higher for the companies that issue warrants to their investment banks. This significant difference in the total broker compensation is due to the fact that the value of warrants is 2.4% of the gross proceeds or 75% of the commission. Thus, the brokers increase their fees by 75% when they include warrants in their compensation package.

The firms in the warrant IPO group are more likely to be underwritten by reputable brokers than those in the no-warrant group. This is in stark contrast to Barry et al. (1991), Dunbar (1995), Bae and Jo (2007) and Jain and Kini (1999), who find that firms that issue warrants to their underwriters are underwritten by less reputable investment banks. A possible explanation for the very limited involvement of the less reputable brokers in the underwriting of riskier IPOs that issue compensation warrants is that these brokers may have neither a

85 broad distributional network of institutional investors to whom they can sell the IPO shares, nor the capital to absorb any unsold shares. On the other hand, reputable brokers also advise risky IPOs because they may have a broad network of institutional investors (Fang 2005). One potential reason that our results are different from the aforementioned studies may be due to the unique role of the broker on the AIM market. A company listed on AIM must have a broker at all times. Brokers play a crucial role for the survival of the IPO firm post- admission as they are responsible for assisting with further corporate finance and fundraising activities (i.e. rights issues, mergers), informing for relevant market issues (change in economic environment) and advising the company on how to maintain good investor relations (London Stock Exchange, 2007b). As a result, a risky firm may do well and grow significantly in the future under the guidance of a reputable broker. If this happens, then the value of its stock price will rise, the value of warrants will increase and the compensation of the brokers will become higher.

In addition, the volatility and return of the FTSE AIM all share index are higher for the warrant group IPOs when compared to those of the no-warrant group (0.94% vs. 0.76% and 4.35% vs. -0.37% respectively). This implies that the IPOs, which issue warrants to the brokers, are usually listed on AIM when the market volatility and return are high. Table 3 reports the correlation matrix of the variables that will later be used in the two stage Heckman model.

Due to the fact that our results related to the brokers reputation, which underwrite IPOs that issue warrant to them, are very different from those reported in previous literature, we provide descriptive statistics for the bottom and top 10% of brokers (Table 4). The brokers are ranked according to their market share (gross proceeds raised from the IPOs each broker has advised on as a percentage of the total gross proceeds from all IPOs). About 64.4% of all warrants (56 out 8725) are issued by companies that are underwritten by the top 10% of brokers, whereas the equivalent percentage for companies advised by the bottom 10% of brokers is only about 3.5% (3 out of the 87 IPOs that issued warrants to their brokers). For instance, Collins Stewart (the most reputable broker based on market share) advised 11.97% of all IPOs (31 IPOs) and raised 17.41% of the total gross proceeds (£994.1 mil.). Out of the 31 IPOs underwritten by Collins Stewart, 17 of them (or 54.84%) issued warrants to the

25 In this table, the total number of companies that have issued warrants to their brokers is 87, and not 86 as shown in Table 2, because one IPO had 2 brokers and issued warrants to both of them.

86 broker. So, in total, Collins Stewart alone underwrote 19.54% (17 out of 87) of all the IPOs that issued warrants.

7.2 Warrant characteristics

Table 5 reports the characteristics of the 87 warrants issued to the brokers as part of their compensation package. Panel A of the table shows that the average size and value of warrants (expressed as a percentage of the number of shares issued and the gross proceeds) are 6.7% and 2.4% respectively. The warrants have a maximum (minimum) size and value of 133.2% and 25.1% (0.2% and 0%) respectively. For the valuation of the warrants, we measure the volatility of the FTSE AIM all-share index over the 126 days before the offering. The value of the warrants is 2.4% and enhances the brokers underwriting compensation package by about 75% (value of warrants as a percentage of the underwriting commission). Barry et al. (1991) report an average size and value of warrants of 7.9% and 3.92% respectively, for issues that raised $10 million or above in the IPO. Ng and Smith (1996) find a warrant value of 5.67% for SEOs. The average ratio of exercise price to offer price is almost equal to 1 (1.02) and the ratio has a minimum and maximum value of 0.55 and 1.62 respectively. The warrants have an average life span of 3.9 years, with a minimum and maximum of 1 and 21 years respectively.

Panel B of the table reveals that 63 IPO companies granted warrants to brokers without any lock-in agreements. These warrants can be exercised from the first day of trading. However, this is not the case in the US as the FINRA Rule 5110 imposes a minimum lock-in period of 180 days. The figures in Panel C show that 24 IPOs issued warrants with a lock-in period, the average lock-in period among this subsample being 0.81 years, and the minimum and maximum being 0.25 and 1 year respectively (Panel A). Panel D shows that 76 IPOs offered warrants to brokers at an exercise price equal to the offer price. The finding that, on average, the exercise price of the warrants is almost equal to the offer price is in contrast to Barry et al. (1991), who find a ratio of 1.205, with 96% of the warrants in their sample having an exercise price equal to or greater than 120% of the offer price. Their findings can, to a large extent, be explained by US state “blue sky” laws on securities.

Furthermore, Panel F shows that only 8 warrants (just over 9%) have an exercise price above the offer price, the average across these 8 being an exercise price 26% higher than the offer price. In addition, 3 (about 3.5%) IPOs issued warrants, shown in Panel E, offered at an exercise price below the issue price (at an average of 20% below the issue price). Thus, the

87 warrants of these 3 IPOs were already in the money prior to listing. This was not permitted in the US prior to 2004 (NASD Rule 2710). In Panel A the warrants have an average time to expiry of 3.9 years. This is shorter than the 4.9 years reported by Barry et al. (1991). In addition, our data shows a maximum expiry period of 21 years, whereas, in the US, according to FINRA Rule 5110, the expiry period is restricted to 5 years (Barry et al., 1991, Ng and Smith, 1996). To summarize, in an environment with fewer regulatory constraints underlying the non-cash compensation, as is the case on AIM, warrants can even be issued at an exercise price lower than the issue price, with no lock in agreement and an expiry period much greater than 5 years.

The characteristics of the warrants issued by both underwritten and non-underwritten offerings are similar as there is no regulation underling non-cash compensation on AIM. The only difference is that 43.4% of the non-underwritten IPOs issue warrants to their brokers whereas the equivalent percentage for the underwritten offerings is 33.3%. One possible explanation may be the fact that non-underwritten IPOs are smaller and riskier when compared to the underwritten. This is why we would expect a higher percentage of the non- underwritten IPOs to grant warrants to their brokers.

7.3 Test of the cost minimisation hypothesis

Table 6 reports the results of the reduced-form probit model in which the dependent variable is a dummy that takes the value of one when warrants are used and zero otherwise. Two different probit models are presented. The independent variables in these models are exactly the same, apart from the broker reputation. Model 1 includes the reputation based on the gross proceeds raised in the IPO, and model 2 includes that based on the number of IPOs advised by the broker, respectively. From Table 6 it is evident that riskier firms are more likely to issue compensation warrants to their brokers. More specifically, firms that are younger, have a higher aftermarket standard deviation of returns, lower public float, less cash, smaller size, raise less money from the selling of existing shares, are underwritten by more reputable underwriters and the return of the FTSE AIM index is high for the two months prior to the IPO date have a higher probability of issuing warrants as part of the package used to compensate brokers for their services. Dunbar (1995) also finds that riskier companies are more likely to grant warrants to their underwriters.

Our results show that the coefficient of the public float is negative and is significantly different from zero. This implies that the probability of using warrants is inversely related to

88 the proportion of shares offered at the time of listing. This finding is not consistent with the positive and statistically significant results reported by Dunbar (1995) for his probit model. In addition, the percentage of the money raised from the selling of existing shares in the IPO is significantly negatively related with the dependent variable. This is in accordance with Dunbar’s (1995) and Logue’s and Lindvall’s (1974) findings according to which the more shares insiders are selling the higher their bargaining power is in lobbying for lower listing costs. The market capitalisation and the standard deviation of returns have a negative and positive relationship respectively with the dependent variable. This means that the larger the firm and the less risky it is, the lower the probability of issuing warrants will be. In addition, the cash and cash equivalents (as a percentage of gross proceeds) the companies have at their disposal the year prior to the IPO have a negative effect on the probability of issuing warrants. So, companies that are cash constrained are more likely to issue warrants. This implies that a possible explanation of why brokers include warrants in their compensation packages may be that companies do not have enough cash to meet the brokers’ costs (hypothesis 1).

The broker reputation variable has a positive and statistically significant effect in the dependent variable. Thus, the probability of issuing warrants is higher for companies that are underwritten by brokers that are more reputable. This is in contrast to the finding of Dunbar (1995), who reports that the probability of issuing warrants is higher for firms that are underwritten by less reputable underwriters. One possible explanation for our result is that brokers make an explicit commitment to buy the shares of the offering if they are not able to sell them (underwritten IPOs). This implies that less reputable brokers will, on average, avoid bringing riskier IPOs to the market because they may not be able to procure enough investors for the company’s shares. On the other hand, more reputable brokers may bring riskier companies, that issue warrants, to the market, as well as less risky companies, because they have a wide network of institutional investors (Fang, 2005) to which they can sell the shares.

The return of the FTSE AIM all share index has a positive and statistically significant effect on the probability of issuing warrants, which means that the higher the return on the market, the higher the likelihood of issuing warrants. This is consistent with the argument that brokers may be able to time the issuance of warrants as they include them in their compensation package mainly when the market is doing well (hypothesis2). If the FTSE AIM index is increasing then most probably the firm’s stock price will also increase in the

89 aftermarket and the compensation of the brokers (including warrants) will consequently increase because the value of warrants is linked with the aftermarket performance of the company’s share price.

The results obtained from the first-stage (reduced-form) probit regression are used to construct the IMR. In the second-stage regression, the variables underpricing and total broker compensation are regressed on the IMR and on the independent variables, separately for the two IPO groups, those with warrants and those without. The use of the IMR adjusts for the selectivity problem that arises from the fact that we can only observe the contracts used by the issuers but cannot observe what would have happened if the alternative contract (i.e. warrant or no-warrant) had been used. In Table 7, we present the second-stage regressions based on model 1 of Table 6.

The results reported in Table 7 show that three of the IMRs coefficients are statistically significant, which suggests that there is selectivity bias, and without them OLS regressions would yield biased and inconsistent estimates. For the companies in the no-warrant group (good quality companies), underpricing is significantly positively related to the standard deviation of returns and the volatility and return of the FTSE AIM all share index, which implies that investors require greater underpricing when the firm is riskier and when the market volatility and return are high. In addition, the underpricing is negatively related to the public float and broker’s reputation, which means that the more shares sold in the IPO and the higher the reputation of the broker is, the lower the underpricing will be. For the same group of companies (no-warrant group) the total broker compensation is significantly positively related to the standard deviation of returns and significantly negatively related to the money raised from the selling of existing shares, cash available the year prior to the IPO, market capitalisation and the volatility of the FTSE AIM index for the two months prior to listing. Thus, the lower the standard deviation is, the higher the cash the company has the year prior to the IPO, the more money raised from the selling of existing shares, the larger the company is and the higher the volatility of the FTSE AIM index, the lower the broker compensation will be. Booth and Smith (1986), Ng and Smith (1996) and Krigman et al. (2001) also report the existence of economies of scale in underwriting fees.

For the warrant IPO group, the underpricing is significantly negatively related to the age, public float, money raised from secondary shares and market capitalisation of the company, and positively related to the standard deviation of returns, broker’s reputation, volatility and

90 return of the FTSE AIM index. Moreover, the total broker compensation, for the same group of companies, increases if the companies have a higher standard deviation of returns, broker reputation and return of the FTSE AIM index. The broker’s compensation has a negative relationship with the age, public float, secondary proceeds, cash and market capitalisation of the IPO. The aforementioned results for the warrant IPO group imply that the riskier the firm is the higher the underpricing and total broker compensation will be.

The coefficients in Table 7 are used to estimate what the underpricing and total broker compensation would have been had the alternative contract (warrants vs. no warrants) been used. We multiply the coefficient estimates from the second-stage regressions (Table 7) with the independent variables. Then, we compare these values with the actual underpricing and total broker compensation, as reported in Table 8.

For the IPOs that issue warrants to their brokers the mean actual underpricing is 23.3% but would have been 40.5% if warrants had not been used. This means that these companies would have experienced a much higher underpricing had warrants not been issued. This finding is consistent with that of Dunbar (1995), who reports a mean underpricing of 23.3% for companies that issue warrants, and a figure of 36.4% had warrants not been used.

For the same group of companies (the warrant IPO group), the total broker compensation is 5.6% but would have been 3.46% if warrants had not been used. The results suggest that if these companies had not issued compensation warrants they would have paid a lower fee to their brokers. Dunbar (1995) also finds qualitatively similar results.

Adding up the two aforementioned costs (underpricing and total broker compensation) we observe that IPOs that issue warrants incur a total cost of 28.9% (23.6% + 5.6%). This figure would have been much higher, 43.96% (40.5% + 3.46%), had the firms not issued warrants. So, companies are able to minimise their total costs of going public by issuing warrants to their brokers (cost minimisation, hypothesis 3). This is due to the fact that warrants are a credible signal that the issuers are not selling overpriced securities. As a result, investors require a smaller discounting in the offer price, reducing the underpricing cost for the IPO firm (23.3% vs. 40.5%).

For the no-warrant group, the mean actual underpricing is 14% but it would have been 24.89% had they issued warrants. This result is different from that of Dunbar (1995), who finds that, for companies that do not issue warrants, the underpricing is unaffected by their

91 use. As far as the total underwriter compensation is concerned, our results show that companies pay a compensation of 3.6% and would have paid 4.87% had they issued warrants. This finding is consistent with that of Dunbar (1995). If we add up the two aforementioned costs, then companies that do not issue warrants incur total costs of 17.6% and this figure would have been almost double, 29.76%, had they chosen to issue warrants. This is why the no warrant IPO group firms only pay cash to their brokers (hypothesis 3).

Overall, it is evident from our results that companies that make use of the non-cash compensation minimise their costs of going public, as their counterparts do in the US. Dunbar (1995) suggests that the NASD (renamed FINRA) should relax the warrant regulations, as they are unnecessarily restrictive. More specifically, according to Dunbar (1995) the 10% maximum limit underlying the amount of warrants that can be offered to the underwriters as part of their compensation restricts their ability to certify the offering price. In 2004 NASD actually relaxed its non-cash compensation regulations and abolished the aforementioned requirement. Our findings further suggest that even in an environment where there are almost no regulations underlying the non-cash compensation, IPO firms are still able to choose the contract that minimises their total costs of going public.

7.4 Robustness tests

In order to test the robustness of our finding we conduct the following tests:

Instead of the first day return (underpricing), we make use of the market adjusted first day return (MAIR) which is calculated as (first day return – FTSE AIM all share index return at the day the IPO took place). From Table 2 it is evident that the first day return is similar to the MAIR. In addition, instead of public float, we use the variable retained ownership which is the level of ownership retained by the insiders of the IPO firm after the IPO. According to the unit IPO literature companies with lower managerial ownership have a higher probability of including warrants in their IPOs (agency costs hypothesis, Schultz, 1993). Due to the fact that it is difficult to obtain data on the shareholdings of managers, previous papers make use of the retained ownership. Table 2 reports that the retained ownership is significantly different between the IPOs that issue compensation warrants and those that do not (67% vs. 58%). This means that insiders in warrant IPOs keep a higher level of ownership when compared with those of the no-warrant IPOs.

Then, we replicate our two stage Heckman analysis and, apart from the two aforementioned variables, we also use two dummies: a hot market dummy and a cash constrained dummy. The hot market dummy takes the value of one if the IPO occurs during the period 2004-2005 and zero otherwise. This is due to the fact that 40% of the IPOs took place during these two years (Yung et al., 2008). The reason that we use a hot market dummy is that it has been widely reported in the literature that there is a connection between underpricing and IPO volume (Pastor and Veronesi, 2005, Lowry, 2003, Jenkinson and Ljungqvist, 2001 and Lowry and Schwert, 2002). The cash constrained dummy takes the value of one if the company is cash constrained and zero otherwise. More specifically, in order to categorise a company as cash constrained we create average industry benchmarks. For every industry we calculate the average (median) cash (as a percentage of gross proceeds) the companies have available the year prior to the IPO. If a company within an industry has cash (as a percentage of gross proceeds) below the average (median) industry benchmark then it is considered cash constrained.

From Table 1 in the Appendix the retained ownership variable is statistically significant at 1% level and has a positive sign which implies that the higher the ownership retained by insiders the higher the probability of issuing warrants to the brokers. This is also consistent with the Secondary_Proceeds variable according to which the more proceeds raised from the selling of existing shares the lower the probability of issuing warrants. Logue and Lindvall (1974) report that if insiders are selling shares at the IPO then they can lobby for paying less fees to their underwriters. Also, the cash dummy variable (Table 1) is statistically significant and has a positive sign which means that firms that have cash below their industry average benchmark have a 21% higher probability to issue warrants to their brokers. Similar results are obtained if we substitute the average industry benchmark cash constrained dummy with the median one. This finding is consistent with that in Table 6 according to which the less cash the companies have at their disposal the year prior to the IPO the more likely they are to issue compensation warrants. The hot market dummy is not statistically significant (Table 1 in Appendix) in both probit models. Overall our results remain qualitatively the same as the firms that issue warrants to their brokers minimise their total costs of going public. This decrease in costs is mainly driven by a reduction in MAIR as companies that issue warrants incur a MAIR of 23.3%, but would have incurred 33.1% had they not issued them (Table 2 in Appendix).

In addition, we make use of the rest three variables that also capture the broker’s reputation (look at Table 2 in the main part of the analysis). Also, we compute the dynamic reputation based on the three and four years prior to the IPO. Furthermore, we calculate the return and volatility of the FTSE AIM all share index for the three months prior to listing. Moreover, we redo the same analysis by including the variable ‘cash/GP’ in the first day return/underpricing second stage regression (Table 7). The results obtained are qualitatively the same, as the firms that issue warrants to their brokers are able to minimise their total costs of going public (first day return/underpricing and broker’s compensation).

8 Conclusion

This study examines the use of compensation warrants in an institutional setting that is very different from that of US, as it has almost no regulations underlying the use of non-cash compensation. To the best of our knowledge, this is the first empirical analysis on the use of compensation warrants in underwritten offerings on the AIM market of the LSE. One of the main findings is that brokers appear to time the issuance of warrants because they include them as part of their compensation package mainly when the market is doing well and the market return is high. This result is in line with our second hypothesis according to which if the market is doing well then it is likely that the share price of the company will increase in the aftermarket and the value of warrants will also increase. In addition, we find supportive evidence of our first hypothesis which has never been tested in the existing empirical literature; firms that are cash-constrained usually issue warrants to their brokers. This implies that companies that do not have enough cash in their balance sheets may issue warrants to pay part of the underwriting fees. Moreover, companies that issue warrants are younger and riskier and are more likely to be underwritten by reputable brokers. This contradicts Dunbar’s (1995) and Barry’s et al. (1991) results who find that firms that issue compensation warrants are underwritten by less reputable investment banks.

The companies reduce their total costs of going public (underpricing and total broker compensation) by almost half (28.9% vs. 43.96%) by issuing warrants to their brokers. So, IPO firms choose he contract that minimises their costs of going public (hypothesis 3). The results also reveal that, on average, the brokers enhance their underwriting compensation by about 75% as a result of accepting warrants as part of the compensation package.

Consistent with our results, Dunbar (1995) finds that companies that issue warrants minimise their costs of going public. In addition, he suggests that the NASD should relax the

94 regulations underlying non-cash compensation as they are unnecessarily restrictive. Our findings show that, in an environment in which there are almost no regulations related to non-cash compensation, issuers are still able to minimise their costs of going public through the use of warrants, as their counterparts in the US market do.

References

Bae, S., Jo, H., 2007. Underwriter warrants, underwriter reputation, and growth signalling. Review of Quantitative Finance and Accounting 29, 129-154.

Barry, C. B., Muscarella, C. J., Vetsuypens, M. R., 1991. Underwriter warrants, underwriter compensation, and the costs of going public. Journal of Financial Economics 29, 113-135.

Boehme, R., Çolak, G., 2012. Primary market characteristics and secondary market frictions of stocks. Journal of Financial Markets, 15, 286–327.

Booth, J. R., Smith, R. L., 1986. Capital raising, underwriting and the certification hypothesis. Journal of Financial Economics 15, 261-281.

Brau, J. C., Li, M., Shi, J., 2007. Do secondary shares in the IPO process have a negative effect on aftermarket performance? Journal of Banking & Finance, 31, 2612-2631.

Burton, B., Helliar, C., Power, D., 2006. Practitioners’ Perspectives on the IPO Process and the Perils of Flotation. The European Journal of Finance 12, 671-692.

Carter, R., Manaster, S., 1990. Initial Public Offerings and Underwriter Reputation. The Journal of Finance 45, 1045-1067.

Clarkson, P. M., Thompson, R., 1990. Empirical estimates of beta when investors face estimation risk. The Journal of Finance 45, 431-453.

Dunbar, C. G., 1995. The use of warrants as underwriter compensation in initial public offerings. Journal of Financial Economics 38, 59-78.

Fang, L. H., 2005. Investment Bank Reputation and the Price and Quality of Underwriting Services. The Journal of Finance 60, 2729-2761.

Garner, J. L., Marshall, B. B., 2010. The non-7% solution. Journal of Banking & Finance 34, 1664-1674.

Goergen, M., Khurshed, A., Mudambi, R., 2006. The Strategy of Going Public: How UK Firms Choose Their Listing Contracts. Journal of Business Finance & Accounting 33, 79- 101.

Golubov, A., Petmezas, D., Travlos, N. G., 2012. When It Pays to Pay Your Investment Banker: New Evidence on the Role of Financial Advisors in M&As. The Journal of Finance 1, 271-311.

Habib, M. A., Ljungqvist, A. P., 2001. Underpricing and entrepreneurial wealth losses in IPOs: theory and evidence. Review of Financial Studies 14, 433-458.

Heckman, J. J., 1979. Sample selection bias as a specification error. Econometrica 47, 153- 161.

Ibbotson, R. G., 1975. Price performance of common stock new issues. Journal of Financial Economics 2, 235-272.

Jain, B. A., Kini, O., 1999. On investment banker monitoring in the new issues market. Journal of Banking & Finance 23, 49-84.

Jenkinson, T., Ljungqvist, A., 2001. Going Public: The Theory and Evidence on How Companies Raise Equity Finance. Clarendon Press, Oxford.

Krigman, L., Shaw, W. H., Womack, K. L., 2001. Why do firms switch underwriters? Journal of Financial Economics 60, 245-284.

Lee, L.-F., 1978. Unionism and Wage Rates: A Simultaneous Equations Model with Qualitative and Limited Dependent Variables. International Economic Review 19, 415-433.

Li, K., Prabhala, N. R., 2007. Self-Selection Models in Corporate Finance. Handbook of Corporate Finance: Empirical Corporate Finance, 47-62.

Logue, D. E., Lindvall, J. R., 1974. The Behavior of Investment Bankers: An Econometric Investigation. The Journal of Finance, 29, 203-215.

London Stock Exchange, 2010. A guide to listing on the London Stock Exchange. http://www.londonstockexchange.com/companies-and-advisors/main- market/documents/brochures/gudetolisting.pdf

London Stock Exchange, 2007a. AIM Rules for Nominated Advisers. http://www.londonstockexchange.com/companies-and-advisors/aim/publications/rules- regulations/aim-rules-for-nominated-advisers.pdf

London Stock Exchange, 2007b. Joining AIM: A Professional Handbook. http://www.faegre.com/files/22472_Joining%20AIM%202.pdf

Lowry, M., 2003. Why does IPO volume fluctuate so much? Journal of Financial Economics 67, 3–40.

Lowry, M., Schwert, G., 2002. IPO market cycles: bubbles or sequential learning? Journal of Finance 57, 1171–1200.

Loughran, T., Ritter, J. R., 2002. Why Don't Issuers Get Upset About Leaving Money on the Table in IPOs? Review of Financial Studies 15, 413-444.

Megginson, W. L., Weiss, K. A., 1991. Venture Capitalist Certification in Initial Public Offerings. The Journal of Finance 46, 879-903.

National Association of Securities Dealers, 2004. Amendments to the Corporate Financing Rule. http://www.finra.org/web/groups/industry/@ip/@reg/@notice/documents/notices/p003258 .pdf

Ng, C. K., Smith, R. L., 1996. Determinants of Contract Choice: The Use of Warrants to Compensate Underwriters of Seasoned Equity Issues. The Journal of Finance 51, 363-380.

Pastor, L., Veronesi, P., 2005. Rational IPO waves. Journal of Finance 60, 1713–1757.

Ritter, J. R., 1987. The costs of going public. Journal of Financial Economics 19, 269-281.

Schultz, P. 1993. Unit initial public offerings: a form of staged financing. Journal of Financial Economics 34, 199-229.

Yung, C., Gonul, C., Wei, W., 2008. Journal of Financial Economics 89, 192-208.

Table 1 Non-cash (i.e. warrants) and total underwriting compensation regulations: London Stock Exchange vs. US Stock Exchanges This table reports the main differences in the regulations underlying non-cash (i.e. warrant) and total underwriting compensation between the London Stock Exchange (LSE) and the US stock exchanges. Gross proceeds denotes the money raised from the IPO. NA means not applicable. NASD is the National Association of Securities Dealers. The regulations underlying the warrant characteristics and the underwriting compensation are the same in the Alternative Investment Market (AIM) and Main Market of the LSE, except for the amount of warrants that can be issued. The NASD and New York Stock Exchange (NYSE) made amendments to Rule 2710 (Corporate Financing Rule) that became effective from 22 March 2004. NASD and NYSE regulations were consolidated under the Financial Industry Regulatory Authority (FINRA) on 30 July 2007. FINRA adopted most of NASD Rule 2710 as FINRA Rule 5110 on 16 July 2008. London Stock US Stock Exchanges Exchange US Stock Exchanges (after March 2004) Rules (Main and AIM (before March 2004) markets) The amount of warrants issued must be less than 10% of the company’s The amount of share capital on AIM securities (i.e. (London Stock warrants) issued to 1. Amount of warrants Exchange, 2007a). The underwriters as part of offered to adviser (broker amount of warrants their compensation NA or Nomad) or underwriter. issued cannot exceed cannot exceed 10% of 20% of the company’s the shares issued to issued share capital in the public (NASD the Main Market Rule 2710). (London Stock Exchange, 2010). 5 years (NASD Rule 5 years (FINRA Rule 2. Exercise period. NA 2710). 5110).

1 year (NASD Rule 180 days (FINRA Rule 3. Lock-in period. NA 2710). 5110).

Equal to the offer 4. Min. exercise price. NA price (NASD Rule NA 2710).

0.2% of gross proceeds for 1% amount of 5. Min. value of warrants. NA NA warrants (FINRA Rule 5110)

Varies according to Varies according to 6. Max. limit in total gross proceeds and gross proceeds and risk underwriting NA risk assumed (NASD assumed (FINRA Rule compensation. Rule 2710). 5110).

Table 2 Comparison between warrant and no-warrant IPO groups First day return/Underpricing is the 1st day return and is calculated as (closing price – issue price)/issue price. MAIR is the market adjusted first day return and is calculated as (first day return – FTSE AIM all share index return at the day the IPO took place). Public Float is the shares sold in the IPO divided by outstanding shares. Retain. Owner. is the level of ownership retained by insiders of the IPO firm after the IPO. Standard Deviation is the standard deviation of returns for 20 days in the aftermarket. Age is the number of years from incorporation to flotation on AIM. Gross Proceeds (million £) is the money raised from the flotation, measured as the offer price times the total number of shares offered. Secondary Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Market Cap. (million £) is the market capitalisation of the company. Total Assets, Revenues and Cash are extracted from the admission document and relate to figures for the year prior to the IPO. Commission is the money paid to the brokers for procuring subscribers and buyers for the new and selling shares respectively. The Value of Warrants is obtained from the CEV model and is measured as a percentage of gross proceeds. Total Broker Comp. is the summation of the commission and the value of warrants. Dyn. Broker Rep is the dynamic broker reputation and is measured based on the total gross proceeds raised and the number of IPOs brought to the market by each broker over the previous 3.5 years. Value of Warrants/Commission (%) is the Value of Warrants divided by the Commission and is expressed as a percentage. Stat. Broker Rep is the static broker reputation measure. It is based on the total gross proceeds and number of IPOs raised and brought to the market by each broker for the period from January 1999 to December 2010. Vol (FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date. Ret (FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. N is the number of IPOs. Warrants = 1 Warrants = 0 Total Sample (N Difference Difference in

(N = 86) (N = 173) = 259) in Means Medians Mean Median Mean Median Mean Median (p-value) (p-value) Underpricing (%) 23.3 10.0 14.0 8.0 17.0 9.0 0.06* 0.29 MAIR (%) 23.3 10.5 14.1 7.86 17 8.8 0.06* 0.3 Issue Price (£) 0.83 0.74 1.29 1.1 1.14 1 0*** 0*** Public Float (%) 33.0 33.0 42.0 37.0 39.0 35.0 0*** 0*** Retain. Owner. (%) 67 67 58 63 61 65 0*** 0*** Standard Deviation (%) 3.0 2.0 2.0 1.0 2 0.02.0 0*** 0*** Age (years) 1.89 0.462 4.53 0.71 3.78 0.671 0.02** 0.13 Gross Proceeds (mil. £) 11.7 6.34 27.2 14 22.1 10.8 0*** 0*** Secondary Proceeds 5.65 0 18.36 0 14.14 0 0*** 0*** (%) Market Cap. (mil. £) 43.8 21.1 63.8 40.4 57.1 33.6 0.07* 0*** Total Assets (mil. £) 5.5 1.5 27.1 7.9 19.9 4.8 0*** 0*** Revenues (mil. £) 4.3 0.4 30.7 5.4 21.9 2.4 0*** 0*** Cash (mil. £) 1.01 0.15 3.53 0.76 2.7 0.51 0.03** 0*** Commission (%) 3.2 3.2 3.6 3.5 3.5 3.5 0.06* 0.07* Value of Warrants (%) 2.4 0.8 Total Broker 5.6 4.7 3.6 3.5 4.3 4.0 0*** 0*** Comp.(%) Value of Warrants/Commission 75 25 (%) Dyn. Broker Rep. 0.091 0.076 0.066 0.035 0.074 0.039 0.07* 0*** (gross proceeds) Dyn. Broker Rep. (N 0.088 0.084 0.073 0.031 0.078 0.039 0.18 0.03** IPOs) Stat. Broker Rep. 0.081 0.061 0.054 0.043 0.063 0.043 0*** 0*** (gross proceeds) Stat. Broker Rep. (N 0.086 0.116 0.059 0.042 0.068 0.054 0*** 0*** IPOs) Vol (FTSE AIM index, 0.94 0.64 0.76 0.55 0.82 0.58 0.03** 0.16 %) Return (FTSE AIM 4.35 4.77 -0.37 1.46 1.2 2.14 0.02** 0.07* index, %) ***, ** and * indicate statistical significance at the 1%, 5% and 10% significance levels.

100

101

Table 3 Correlation Matrix This table provides the pairwise correlation matrix of the independent variables that will be used in the two stage probit model. Age (Age) is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). SD (Standard_Deviation) is the standard deviation of the company’s returns over 20 days in the aftermarket. Public (Public_Float) is the ratio of the total number of shares sold in the IPO divided by the outstanding shares. Second (Secondary_Proceeds) is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Cash (Cash/GP) is the cash and cash equivalents, available the year prior to the IPO, divided by the gross proceeds. MC (MC) is the natural logarithm of the market capitalisation. Rep (GP and NIPOs) (Broker_Rep.) is the reputation of the broker based on the gross proceeds and number of IPOs raised and advised by each broker over the previous 3.5 years. All reputation measures are dummy variables that take the value one if the IPO is underwritten by one of the 10% most reputable brokers and zero otherwise. Vol (Vol_(FTSE AIM index)) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date. Ret (Ret_(FTSE AIM index)) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. In brackets we provide the names of the variables as they appear later in the two stage model. Age SD Public Second Cash MC Rep (GP) Rep (NIPOs) Vol Ret Age 1 SD -0.09 1 Public -0.02 -0.18 1 Second 0.25 -0.10 0.21 1 Cash 0.15 0.04 -0.11 0.17 1 MC 0.13 -0.11 -0.03 0.23 0.04 1 Rep (GP) -0.05 0.05 0.14 -0.10 0.04 -0.26 1 Rep (NIPOs) -0.07 0.05 0.13 -0.08 0.09 -0.22 0.90 1 Vol -0.12 0.17 -0.17 -0.08 -0.05 -0.06 -0.03 -0.07 1 Ret -0.06 0.20 0.05 0.02 -0.04 -0.09 0.02 0.00 -0.31 1

102

Table 4 Top and bottom 10% of brokers Number of IPOs is the number of IPOs brought to the market by the given broker. Gross proceeds (£ million) is the money raised from all the IPOs each broker advised. % of IPOs is the number of IPOs the broker advised, expressed as a percentage of the total number of IPOs. % gross proceeds is the gross proceeds each broker raised, expressed as percentage of the total gross proceeds. % of IPOs issued warrants is the number of IPOs that granted warrants to the broker, divided by the total number of IPOs that the broker advised. % of the total number of IPOs that issued warrants is the number of warrants the broker received divided by the total number of IPOs that issued warrants. The brokers are ranked according to the gross proceeds raised from the flotation. % of the total Number % of Gross number Number of IPOs % of % gross IPOs Brokers proceeds of IPOs of IPOs issued IPOs proceeds issued (mil. £) that warrants warrants issued warrants Bottom 10% of brokers Nabarro Wells & Co. 1 0.071 0 0.39 0.001 0 0 Credo Capital 1 0.304 1 0.39 0.01 100 1.15 Hoodless Brennan & 1 0.36 1 0.39 0.01 100 1.15 Partners Astaire & Partners 1 1.081 1 0.39 0.02 100 1.15 Brown Shipley 1 1.374 0 0.39 0.02 0 0 Securities Top 10% of brokers Altium Capital 11 278.2 2 4.25 4.87 18.18 2.30 Seymour Pierce 39 346.9 19 15.06 6.08 48.72 21.84 Numis Securities 14 547.9 7 5.41 9.60 50.00 8.05 Evolution Securities 30 774.4 11 11.58 13.56 36.67 12.64 Collins Stewart 31 994.1 17 11.97 17.41 54.84 19.54

103

Table 5 Descriptive statistics for the 87 warrants issued to the broker(s) Size of warrants is measured as (number of shares that can be purchased under the warrant)/(IPO shares). Value of warrants is obtained from the CEV model and is measured as a percentage of gross proceeds. Exercise/offer price is calculated as (price at which the warrant can be exercised)/(offer price). Time to expiration (years) is measured as the number of years between the date of listing on AIM and the expiration date of the warrant. Lock-in (years) is the time period during which the warrant cannot be exercised. Min and Max are the minimum and maximum values respectively. N is the number of warrants offered. Panel A: All warrants Size of Value of Exercise/offer Time to expiration Lock-in

warrant (%) warrant (%) price (years) (years) mean 6.7 2.4 1.02 3.9 0.81 median 3.7 0.8 1 3 1 Min 0.2 0 0.55 1 0.25 Max 133.2 25.1 1.62 21 1 N 87 87 87 87 24 Panel B: Warrants with no lock-in period (N=63) mean 5.5 2.3 1.01 4 median 3.7 0.8 1 3 Panel C: Warrants with lock-in period (N=24) mean 9.7 2.4 1.03 3.65 0.81 median 3.6 1.0 1 3.5 1 Panel D: Warrants with an exercise price equal to the offer price (N=76) mean 7.2 2.6 1 3.91 median 3.8 0.9 1 3 Panel E: Warrants with an exercise price lower than the offer price (N=3) mean 3.0 1.4 0.8 5.33 median 2.7 1.8 0.91 5 Panel F: Warrants with an exercise price higher than the offer price (N=8) mean 3.1 0.4 1.26 3.25 median 2.4 0.3 1.2 3 In Panel A, the maximum size of the warrants is 133.2% of the shares offered in the IPO because one company (Alltracel Pharmaceuticals plc) issued 825,843 shares and a warrant to subscribe for 1,100,000 ordinary shares. The maximum time to expiration of the warrants issued is 21 years because one company (CNG Travel Group plc) issued warrants that could be exercised up to 21 years post-admission.

104

Table 6 Probit regression model for the contract equation The dependent variable is a dummy that takes the value of one if the firm issues compensation warrants to its broker(s) and zero otherwise. Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). Standard_Deviation is the standard deviation of the company’s returns over 20 days in the aftermarket and is multiplied by 100. Public_Float is the ratio of the total number of shares sold in the IPO divided by the outstanding shares. Secondary_Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Cash/GP is the cash and cash equivalents, available the year prior to the IPO, divided by the gross proceeds. MC is the natural logarithm of the market capitalisation. Broker_Rep. (GP and N of IPOs) is the reputation of the broker based on the gross proceeds and number of IPOs raised and advised by each broker over the previous 3.5 years. All reputation measures are dummy variables that take the value one if the IPO is underwritten by one of the 10% most reputable brokers and zero otherwise. Vol_(FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date and is multiplied by 100. Ret_(FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. The variables standard deviation, public float, secondary proceeds, cash/gp, vol and ret (FTSE AIM index) are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. Model 1 Model 2 marginal marginal coef p-value coef p-value effect effect Intercept 3.14* 0.08 3.09* 0.09 Age -0.19* 0.06 -0.06 -0.17* 0.09 -0.06 Standard_Deviation 0.11*** 0.01 0.04 0.11*** 0.01 0.04

Public_Float -2.16*** 0 -0.72 -2.16*** 0 -0.72

Secondary_Proceeds -0.98* 0.07 -0.33 -1.00* 0.06 -0.33

Cash/GP -1.22** 0.03 -0.41 -1.29** 0.03 -0.43 MC -0.19* 0.06 -0.06 -0.19* 0.06 -0.06

Broker_Rep. (GP) 0.54*** 0.01 0.18

Broker_Rep. (N of 0.53*** 0.01 0.18 IPOs) Vol_(FTSE AIM 0.28 0.17 0.09 0.33 0.1 0.11 index) Ret_(FTSE AIM 1.67** 0.03 0.56 1.74** 0.02 0.58 index) Yearly_Dummies Yes Yes

% correct predictions 77% 77%

Pseudo R square 0.25 0.25 N 259 259 ***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

105

Table 7 Second-stage regression estimates of the underpricing and total broker compensation First day return/Underpricing is the first-day return and is calculated as (closing price – issue price)/issue price. Total Broker Compensation is the summation (Commission + Warrant Value)/Gross Proceeds. Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). Standard_Deviation is the standard deviation of the company’s returns over 20 days in the aftermarket and is multiplied by 100. Public_Float is the ratio of the total number of shares sold in the IPO divided by the outstanding shares. Secondary_Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Cash/GP is the cash and cash equivalents, available the year prior to the IPO, divided by the gross proceeds. MC is the natural logarithm of the market capitalisation. Broker_Rep. (GP) is a dummy variables that takes the value one if the IPO is underwritten by one of the 10% most reputable brokers (based on the gross proceeds over the previous 3.5 years) and zero otherwise. Vol_(FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date and is multiplied by 100. Ret_(FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. IMR is the inverse Mills ratio which is used to adjust for selectivity bias. The variables standard deviation, public float, secondary proceeds, cash/gp, vol and ret (FTSE AIM index) are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. Dependent Variable: Dependent Variable:

First day return/Underpricing Total Broker Compensation Contracts with Contracts without Contracts with Contracts without Warrants Warrants Warrants Warrants Equation 1 Equation 2 Equation 1 Equation 2 p- coef p-value coef p-value coef p-value coef value Intercept 3.6*** 0 -0.36 0.44 0.53*** 0 -0.001 0.96 Age -0.24*** 0 0.02 0.41 -0.03*** 0 0.001 0.25

Standard_Deviation 0.12*** 0 0.04** 0.01 0.01*** 0 0.002* 0.09

Public_Float -1.47*** 0 -0.28* 0.05 -0.23*** 0 -0.001 0.9

Secondary_Proceeds -1.09** 0.01 0.12 0.13 -0.13** 0.01 -0.01** 0.03

Cash/GP -0.10** 0.05 -0.01* 0.06 MC -0.28*** 0 -0.002 0.94 -0.03*** 0 -0.002* 0.05

Broker_Rep. (GP) 0.54*** 0 -0.08* 0.1 0.06*** 0.01 -0.001 0.72

Vol_(FTSE AIM 0.22** 0.02 0.13** 0.04 0.02 0.14 -0.01* 0.06 index) Ret_(FTSE AIM 2.24*** 0 0.99*** 0 0.2*** 0 -0.01 0.56 index) IMR -1.45*** 0 0.45*** 0 -0.15*** 0.01 0.01 0.47

Yearly Dummies Yes Yes Yes Yes

R-square 0.49 0.35 0.36 0.30 N 86 173 86 173 ***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

106

Table 8

Comparison of actual costs with the estimated costs should the alternative compensation contract have been used This table compares the average underpricing and total broker compensation costs with the estimated costs should the alternative contract has been used. Underpricing is the first-day return and is calculated as (closing price – issue price) / issue price. Total Broker Compensation is the summation (Commission + Warrant Value)/Gross Proceeds. N is the number of observations. Average cost estimates for the 86 IPOs that Average cost estimates for the 173 IPOs

issued warrants to brokers that did not issue warrants to brokers

Difference Difference Estimated cost if in means Estimated cost if in means Actual warrants had not Actual warrants had been cost been issued to cost issued to brokers brokers p-value p-value

Underpricing 23.3 40.5 0*** 14 24.89 0 *** (%)

Total Broker 5.6 3.46 0*** 3.6 4.87 0 *** Comp. (%)

N 86 86 173 173

***, ** and * indicate statistical significance at the 1%, 5% and 10% significance levels.

107

Appendix

Table 1 Probit regression model for the contract equation The dependent variable is a dummy that takes the value of one if the firm issues compensation warrants to its broker(s) and zero otherwise. Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). Standard_Deviation is the standard deviation of the company’s returns over 20 days in the aftermarket and is multiplied by 100. Retain. Owner. is the level of ownership retained by insiders of the IPO firm after the IPO. Secondary_Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Cash is a dummy variable that takes the value of one if the company is cashed constrained and zero otherwise. MC is the natural logarithm of the market capitalisation. Broker_Rep. (GP and N of IPOs) is the reputation of the broker based on the gross proceeds and number of IPOs raised and advised by each broker over the previous 3.5 years. All reputation measures are dummy variables that take the value one if the IPO is underwritten by one of the 10% most reputable brokers and zero otherwise. Vol_(FTSE AIM index) is the volatility of the FTSE AIM all share index during the period two months prior to the IPO date and is multiplied by 100. Ret_(FTSE AIM index) is the return of the FTSE AIM all share index during the period two months prior to the IPO date. Hot_Market is a dummy variable that takes the value of one if the IPO occurs during the period 2004-2005 and zero otherwise. The variables standard deviation, retain. owner., secondary proceeds, cash/gp, vol and ret (FTSE AIM index) are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. Model 1 Model 2 marginal marginal coef p-value coef p-value effect effect Intercept 0.68 0.71 0.63 0.73 Age -0.21** 0.04 -0.07 -0.20* 0.06 -0.07 Standard_Deviation 0.13*** 0.00 0.04 0.12*** 0.00 0.04 Retain. Owner. 2.11*** 0.00 0.71 2.11*** 0.00 0.71 Secondary_Proceeds -0.94* 0.06 -0.32 -0.96** 0.05 -0.32 Cash 0.69*** 0.00 0.21 0.72*** 0.00 0.21 MC -0.20* 0.06 -0.07 -0.21** 0.05 -0.07 Broker_Rep. (GP) 0.46** 0.02 0.16 Broker_Rep. (N of 0.47** 0.02 0.16 IPOs) Vol_(FTSE AIM 0.18 0.32 0.06 0.21 0.23 0.07 index) Ret_(FTSE AIM 0.65 0.30 0.22 0.71 0.26 0.24 index) Hot_Market -0.04 0.90 -0.01 0.01 0.98 0.00 Yearly_Dummies Yes Yes % correct predictions 0.76% 0.76% Pseudo R square 0.24 0.24 N 259 259 ***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

108

Table 2 Comparison of actual costs with the estimated costs should the alternative compensation contract have been used This table compares the average MAIR and total broker compensation costs with the estimated costs should the alternative contract has been used. MAIR is the market adjusted first day return and is calculated as (first day return – FTSE AIM all share index return at the day the IPO took place). Total Broker Compensation is the summation (Commission + Warrant Value)/Gross Proceeds. N is the number of observations. Average cost estimates for the 86 IPOs Average cost estimates for the 173 IPOs that

that issued warrants to brokers did not issue warrants to brokers Estimated cost Difference Difference Estimated cost if Actual if warrants had in means Actual in means warrants had been cost not been issued cost p-value issued to brokers p-value to brokers MAIR (%) 23.3 33.1 0.08* 14.1 16 0.29 Total Broker 5.6 3.7 0*** 3.6 3.9 0.52 Comp. (%) N 86 86 173 173 ***, ** and * indicate statistical significance at the 1%, 5% and 10% significance levels.

109

Chapter 4

The IPO when-issued market

Abstract

We examine the IPO when-issued/conditional market of the London Stock Exchange which has a regulatory setting that is very different from that of other developed markets. Our results show that the decision to have a when-issued market affects the setting of the offer price. For companies that have a conditional trading the actual offer price is £3.4 but would have been 54% lower (£1.55) had these firms not had a when-issued market. So, investors actually pay a ‘rent’, through a higher offer price, in order to acquire shares of companies that will be traded in the when-issued market. In addition, companies that are larger, less risky, with higher future growth opportunities and underwritten by more reputable underwriters are more likely to have a when-issued market. Furthermore, the when issued market appears to be highly informative for investors and affects the volume in the first day of trading.

110

1 Introduction

Benveniste and Spindt (1989) report that underwriters pay ‘informational rents’ to investors because they provide them with truthful private information about the value of the company, through their orders during the subscription period. These ‘rents’ are paid in the form of allocating underpriced shares to investors. However, Aussenegg et al. (2006) document that when there exists a when-issued market for initial public offerings (IPOs), which takes place simultaneously with the subscription period, underwriters do not pay any ‘rents’ to investors. This is due to the fact that investors’ private information is revealed to the underwriters through their trades in the when-issued market, as the prices within this market are publicly observable. However, there is one important question that is still not addressed in the existing academic literature; apart from the underwriters, is there any case in which investors also have to pay ‘rents’? More specifically, in an institutional setting in which the when-issued market commences after the subscription period and the allocation of shares have been completed, do investors have to pay any ‘rents’ to the underwriters in order to acquire the IPO shares?

The when-issued market is a market in which trading in the company’s shares commences a few days prior to the security’s official admission on the stock exchange. All trades occurring during the when-issued market are conditional on the fact that the security will be admitted to the stock exchange. If this does not happen then all the when-issued trades will be void and will not be settled. Our paper is focused on the conditional trading26 on the Main Market of the London Stock Exchange (LSE) for four important reasons; the first one is that the number of companies which conduct an IPO through a when-issued market has increased substantially in the more recent years. More specifically, the percentage of IPOs that had a when-issued market during the year 1996 was only 6.56%, but this percentage soared to more than 50% from 2002 onwards. The second reason is that the regulatory setting of the conditional trading in the United Kingdom (UK) is very different from that of other developed markets around the world. The third reason is that the when-issued dealing has attracted a lot of attention recently. According to The Financial Times (2012) it is not clear whether the conditional trading can benefit investors or pose a risk for them as liquidity within this market is invariably low. And the forth reason is that, to the best of our knowledge,

26 The terms when-issued/conditional/grey market/trading/dealing are used interchangeably in this paper.

111 this is the first empirical paper that examines the when-issued market on the LSE over a long period of time, January 1996 to December 2012.

In the UK the conditional trading is not taking place over the counter, but is instead closely regulated and monitored by the LSE. In order for a company to have a when-issued market it has to satisfy a number of requirements set by the LSE (London Stock Exchange, 2012a and 2012b), such as there should be sufficient demand and liquidity for the security, taking into consideration the size of the issuer. In addition, the when-issued market commences only after the allocation of shares is completed and the offer price is determined. These are only some of the differences that the UK’s conditional trading has when compared to those of other European and non-European grey markets.

The contribution of our paper is threefold. First, we study what the determinants of the when- issued market are. This is the first paper that examines why some companies have a when- issued market and others do not. Second, our results have important policy implications for market regulators as the when-issued dealing in the UK has a very different institutional setting when compared to that of other markets. Third, we shed light on a highly under- researched when-issued market of one of the major stock exchanges in the world, the LSE.

The key findings of our paper are the following; the decision to have a when-issued market affects the setting of the offer price. More specifically, for companies that have a conditional trading the actual offer price is £3.4 but would have been 54% lower (£1.55) had these firms not had a when-issued market. So, investors actually pay a ‘rent’ through a higher offer price in order to acquire the company’s shares and would have paid a lower price if this company did not have a when-issued market. The reason is that the conditional trading has informational value for investors and offers them certain advantages, such as price formation before the commencement of the unconditional/aftermarket trading and earliest entry/exit price. As a result, investors know whether they purchased overpriced or underpriced securities and can enter or exit their trades from the very first day of the when-issued market, without having to wait until the commencement of the unconditional trading. Furthermore, for companies that do not have a when-issued market the actual offer price is £1.7 but would have been £3.14 had they had a conditional trading. So, under the LSE’s when-issued institutional setting the Securities and Exchange Commission’s (SEC’s) argument,

112 according to which the grey market for IPOs is not allowed in the United States (US) because it will lead to a lower offer price27, does not hold.

In addition, companies which have a when-issued market have very different characteristics from those that do not. Firms that are larger, less risky, with higher future growth opportunities and underwritten by more reputable underwriters are more likely to have a when-issued market. This is in contrast to Dorn’s (2009) findings who reports that grey and non-grey market companies have similar characteristics. In fact, in Dorn’s study companies that have a conditional trading in Germany are slightly smaller than those that do not. Also, the when-issued market affects the volume in the first day of trading. For companies that have a conditional dealing the volume of trading is approximately 14% higher than those that do not. Moreover, the UK’s when-issued market appears to be highly informative for investors as it facilitates price formation ahead of the unconditional trading.

The remainder of this paper is organised as follows. In Section 2 we discuss the literature review on the when-issued market and the hypotheses tested. Section 3 provides a detailed explanation of the UK’s regulatory setting underlying the when-issued market and compares it with that of other developed markets. In Section 4 we provide details of our data. In Section 5 we discuss our methodology, while in Section 6 we present our results. Section 7 has the conclusion.

2 Literature review & hypotheses

As far as the existing literature on the IPO when-issued trading is concerned, it is focused on the German market because it is considered the most active European grey market (Cornelli et al., 2006). More specifically, Dorn (2009) uses proprietary data of grey market retail investors’ trades of a German retail broker during the period August 1999 to May 2000 and finds that retail buyers pay a hefty premium in the when-issued market relative to the immediate aftermarket. The fact that individual investors are willing to overpay implies that sentiment is driving their trading decisions. Dorn (2009) also reports that the poor performance of retail investors in the conditional market continues even after the crash of 2000, which means that sentiment investors are still active in this market even during periods

27 Paragraph II.F of the Securities and Exchange Act Release No. 38067 (20 December 1996).

113 of poor performance. In addition, he documents that IPOs that are aggressively bought by retail investors in the grey market experience low aftermarket returns, even after controlling for IPO characteristics. This is consistent with the argument that retail sentiment initially pushes the prices in the aftermarket above their fundamental values.

Cornelli et al. (2006) use grey market prices as a proxy for small/retail investor valuations and examine whether their irrational behaviour drives aftermarket prices. They find that high prices during the grey market, which is an indication of overoptimism among small/retail investors, are a very good predictor of the first day price observed in the aftermarket, whereas low prices during the grey market, which is an indication of pessimism of small investors, are not. The economic significance of their results is evident from the fact that the overoptimism among grey market investors can increase aftermarket prices up to 40.5%. In addition, they report that there exist higher levels of trading volume in the aftermarket when prices in the grey market are high. This is due to the fact that bookbuilding investors sell their shares to retail investors only when the latter are overoptimistic and grey market prices are high. Moreover, IPOs with high grey market prices experience long run reversals.

Löffler et al. (2005) provide a very detailed description of the German pre-IPO (or grey) market. They find that pre-IPO quotes are highly informative and good proxies for the first day closing price in the aftermarket. In addition, this pricing accuracy in the grey market gradually increases as the pricing error between the last day quote in the pre-IPO trading and the first day closing price in the exchange is only 0.47%. Moreover, Löffler et al. (2005) report that the pre-IPO return explains a large part of the underpricing which cannot be explained by other variables that have been suggested in the existing empirical literature. Their results cast doubts on the winner’s curse problem (Rock, 1986) and information acquisition models (Benveniste and Spindt, 1989) that have been suggested as possible explanations of the IPO underpricing.

Aussenegg et al. (2006) also provide a very detailed description of the grey market in Germany and they document that there is no partial adjustment phenomenon28 in the Neuer Markt, which is in contrast to other studies that were conducted in the US (Hanley, 1993).

28 IPOs that have positive revisions in the offer price, due to favourable information revealed during the subscription period, experience higher underpricing. The underwriters only partially adjust the price upwards, due to the arrival of this information. This is termed in the literature as ‘partial adjustment phenomenon’.

114

Aussenegg et al. (2006) find that the informational role of the subscription period only exists before the commencement of the grey market. Once the when-issued trading starts, bookbuilding is not a source of costly information for the pricing of the IPO. This is due to the fact that investors’ private information is revealed through the when-issued prices and underwriters do not need to pay any informational ‘rents’ to gather this information.

In this study we examine four hypotheses, in which three of them have never been tested in the existing empirical literature due to the different institutional setting of the UK’s when- issued market. In contrast to the grey market of other major stock exchanges (Cornelli et al., 2006), in the UK the offer price is determined before the commencement of the when-issued market. This means that the information collected by the underwriters during the subscription period is incorporated in the offer price and disclosed to the market before the commencement of the when-issued dealing. So, we would expect the conditional trading prices in the UK to be very good proxies of the market price in the first day of unconditional trading. This argument formulates our first hypothesis which is the following:

Hypothesis 1: The when-issued market prices are unbiased estimators of the closing price in the first day of unconditional trading.

In Table 1 (explained in details in the next part) we documented that the when-issued market is regulated by the LSE and companies should satisfy certain requirements in order to have a conditional trading. One of them is that the security should be sufficiently liquid, taking into consideration the size of the issue. So, we would expect the probability to have a when- issued market to be positively related to the size of the company. Another requirement is that there should be sufficient demand for the security during the when-issued dealing. As a result, we should expect companies with higher future growth opportunities and underwritten by more reputable underwriters to have a when-issued market. According to the certification hypothesis (Shiller, 1989) investors decide on whether to invest in an IPO based on the quality of the underwriters, as they are aware of the informational asymmetries that exist between the issuers and themselves, and expect the underwriters to certify the quality of the IPO (Dorn, 2009, Fang, 2005).

In addition, firms in which the existing shareholders are selling shares at the IPO may be less likely to have a when-issued market. The reason is that the selling of existing shares by

115 insiders may send a negative signal to the market about the future value of the company. This means that the IPO will attract less demand from investors, so it will not satisfy one of the when-issued dealing requirements. As a result we would expect the probability to have a conditional trading to be negatively related with the percentage of gross proceeds raised from the selling of existing shares in the IPO. All in all, based on the when-issued dealing requirements, we expect larger and less risky companies to have a when-issued market. So, our second hypothesis is the following:

Hypothesis 2: The probability to have a when-issued market is higher for large companies that have high future growth opportunities, are underwritten by more reputable underwriters and raise less money from the selling of existing shares as a percentage of their gross proceeds. Or in other words riskier companies are less likely to have a when-issued market.

It is true that some of the companies postpone their listing on the stock exchange because there is a high volatility in the market. So, it may be the case that companies have a when- issued dealing simply because the market volatility is high. The rationale is that due to the informational value of the when-issued market (explained in details in the next part), the IPOs that have a conditional trading may attract a higher demand from investors than those that do not and consequently the listing on the stock exchange will not be postponed for a later date, even if there is a high volatility in the market. This argument formulates our next hypothesis:

Hypothesis 3: The probability of having a when-issued dealing will be positively related to the volatility that exists in the market.

Under the UK’s institutional setting for the when-issued market the offer price is set first and then the when-issued market commences. However, the decision to have a conditional trading is made at least ten business days prior to commencement of the when-issued market (Figure 1, explained in details in the next part), as the company’s broker has to submit a draft when-issued dealing application form to the LSE. This implies that the decision to have a when-issued market precedes the setting of the offer price. In addition, according to the LSE, the conditional trading has certain advantages for investors (price formation before the commencement of the unconditional trading and earliest opportunity to agree on an exit or entry price of the IPO) (London Stock Exchange, 2012b). If the companies and their

116 underwriters are aware of the informational value that the when-issued market has for investors then they will not offer it for free, but they will instead charge investors some ‘rents’, by selling the IPO shares at a higher issue price. Investors may be willing to pay these ‘rents’ because they know the advantages that the when-issued market can offer them. So, we may observe the offer price for companies that have a when-issued market to be higher than what it would have been had these companies not have a conditional trading.

In addition, investors are informed about the IPO offer price and whether they are allocated shares before the commencement of the when-issued dealing. As a result, in contrast to other European markets, if these investors start selling shares in the when-issued market they are not short sellers because the allocation of shares has already been completed. In the US the SEC prohibits the IPO grey market because short sales during this market may lead to a lower offer price. But, under the LSE’s when-issued dealing regulatory setting does the SEC’s argument still hold? If the advantages of the when-issued market, which are documented by the LSE, really exist then we would expect the decision to have a when- issued market to have a positive effect on the offer price. In other words, in contrast to the SEC’s argument, we would expect companies that do not have a when-issued market to set a higher offer price had they had a conditional trading. So, our forth hypothesis is the following:

Hypothesis 4: The decision to have a when-issued market affects the setting of the offer price.

The volume in the first day of trading may be higher for companies that have a conditional trading when compared to those that do not as the when-issued market has certain advantages for investors. However, it may be argued that if larger companies usually have a when-issued market, then the higher volume, which may be observed in the first day of trading, may simply be due to the larger size of the firm. So, is the volume in the first day of trading affected by the existence of a when-issued market, when taking into account the size of the company? Based on the aforementioned argument our last hypothesis is the following:

Hypothesis 5: The volume in the first day of trading will be higher for companies that have a when-issued market when compared to that of companies that do not.

3 When-issued Dealing

117

In the UK there is an active when-issued market, in which the shares of companies that will conduct an IPO are already traded a few days prior to the firm’s official admission on the stock exchange. Under the Rules 1530 - 1532 of the London Stock Exchange, the when- issued dealing is a ‘period of dealing with deferred settlement’ (London Stock Exchange, 2012a and 2012b). All transactions that are taking place during the when-issued period are conditional on the fact that the security will be admitted to trading on the stock exchange. The settlement of these transactions will not take place until the listing of the security on the exchange has occurred. If the security is not listed then all the when-issued dealing trades will be void and will not be settled.

In the UK market the when-issued trading is an on exchange activity conducted under the LSE’s electronic order book or off book, bilaterally between the LSE’s member firms. If the when-issued transactions are executed off the LSE’s order book then they should be reported to the stock exchange. This is why investors can observe the when-issued prices on the LSE’s website29.

In order for companies to have a when-issued market they should satisfy the following requirements (London Stock Exchange, 2012b):

 there must be a fair and orderly market for the securities,  there must be sufficient liquidity, taking into consideration the size of the offer,  there must be settlement in electronic form and  there must be sufficient demand for the security during the when-issued period.

Figure 1 shows the timeline before, during and after the when-issued period on the LSE.

Stage 1: At least ten business days before the commencement of the when-issued trading (Day T-13) the stock exchange must receive a draft when-issued dealing application form and a draft IPO prospectus.

29 This is why the UK’s conditional trading is also known as ‘when-issued market’, which means on exchange trades, whereas a ‘grey market’ means off exchange/over the counter trades.

118

Stage 2: At least two business days before the commencement of the when-issued dealing (Day T-6) the stock exchange must receive the final when-issued trading application form.

Stage 3: The last day that investors can subscribe for the IPO shares is usually the one before the commencement of the when-issued trading (Day T-4). During this day the stock exchange must receive confirmation that the prospectus has been approved by the relevant authority, when share allocation will take place and an indicative offer price. During the same day (Day T-4) the stock exchange will inform market participants of the proposed timetable for the commencement of the when-issued and unconditional trading.

Stage 4: At Day T-3 (point B), the company should announce the offer price and notify investors whether they were allocated IPO shares prior to the commencement of the when- issued market. As a result, the conditional trading commences only after the allocation of shares has been completed, the offer price has been announced and the exchange has collected all relevant and operational approvals (London Stock Exchange, 2012b). Usually the conditional trading lasts for three business days, but the exchange will consider shorter or longer trading periods on a case by case basis.

Stage 5: At Day T (Point A) the unconditional trading commences and this is the first day of the secondary market.

According to the LSE the conditional trading offers investors two important advantages (London Stock Exchange, 2012b):

 Facilitates price formation before the commencement of the secondary/unconditional trading.  Offers them the earliest opportunity to agree on an entry/exit price of the IPO.

3.1 Differences of the when-issued dealing among the UK and other developed markets

The UK’s when-issued dealing institutional set up is very different from that of other developed markets. In this part of the paper although we compare the UK’s conditional trading to that of Australia’s and Hong Kong’s, we mainly focus on the comparison between

119 the UK’s and Germany’s conditional trading (see Table 1) for three important reasons; first Germany has the most active European grey market; second almost all existing academic literature is focused on the German conditional trading and third the grey market in other European countries, such as France, Spain, Switzerland, etc., is conducted in a similar way to that of Germany’s (Cornelli et al., 2006). In the US the grey market for IPOs is prohibited (Regulation M, Rule 105).

One major difference between the UK and German when-issued markets is that the former has to follow the LSE’s rules, whereas the latter is conducted over the counter, by independent brokers, who quote bid-ask spreads and investors take a long or short position based on their expectations. In Hong Kong the grey market also takes place over the counter, whereas in Australia it is regulated by the exchange. The fact that companies have to satisfy certain requirements in the UK in order to have a when-issued market may explain potential differences in the characteristics between the group of firms that have a when-issued market and those that do not. For instance, according to hypothesis 2, we would expect larger companies to have a when-issued market because one of the LSE requirements is that firms that have a conditional trading should be sufficiently liquid, taking into account their size.

Consequently, the lack of such requirements, may also explain the similarity of the characteristics of the two groups of IPOs, as is the case of Germany. The size of companies that have a grey market and those that do not is similar because every company that wants to have a grey market actually does. Aussenegg et al. (2006) report that 94% of the IPOs that were conducted in Germany have a grey market and Dorn (2009) documents that the characteristics between the two groups of IPOs are quite similar.

A non-regulated/over the counter conditional market may also explain the fact that in Germany there are IPOs that were withdrawn within the grey market (Cornelli et al., 2006), whereas this has never occurred on the when-issued trading of the LSE. So, in the UK, all when-issued IPOs are eventually listed on the LSE. Although there is no when-issued IPO that failed to be admitted in the unconditional trading, the conditional and unconditional trading periods are different. Firstly, the risk of an IPO being withdrawn within the conditional trading still exists as the IPO shares are traded on a when-issued basis. In addition, the volume of trading between the two periods (conditional vs. unconditional trading) is significantly different because the price formation takes place at day one of the conditional

120 trading as the when-issued market has certain advantages for investors that do not exist in the unconditional trading period (provide empirical evidence in later parts).

As far as the subscription period is concerned, in the UK and Australia investors usually can subscribe for the IPO shares up until the last day before the conditional trading starts, whereas in Germany until the last day of the grey market (Aussenegg et al., 2006, Löffler et al., 2005). This means that in Germany the subscription period takes place simultaneously with the grey market, whereas in the UK it stops the day before the when-issued trading starts. In addition, in the UK the offer price is determined just before the when-issued dealing commences whereas in Germany is set the last day of the grey market. In the Australian Stock Exchange (ASE) the price is also determined before the commencement of the conditional trading (ASE Operating Rules). According to our first hypothesis the when- issued market in the LSE may be highly informative for investors as they know exactly what is the offer price before the commencement of the when-issued market. In addition, the LSE’s conditional trading may be more informative than that of Frankfurt’s Stock Exchange (FSE) because the offer price in FSE is determined the day before the commencement of the unconditional trading.

Another difference is related to the allocation of shares. In the UK investors are notified whether they are allocated shares just before the when-issued dealing starts, whereas in Germany this takes place in the last day of the grey market (Aussenegg et al., 2006, Löffler et al., 2005). On the ASE the allocation also takes place before the conditional trading commences. This implies that investors on the LSE know exactly how many shares they are allocated before the start of the when-issued trading. In contrast, investors on the FSE, make trades on the grey market on the expectation that they will be allocated shares. As a result, those who are selling on the grey market and are not allocated shares are effectively short sellers. As Schnigge AG, a major broker in the German conditional market, reports on its website, grey market short sellers are usually institutional investors who expect to be allocated shares in the primary market (Aussenegg et al., 2006). The SEC prohibits a grey market for IPOs in the US based on the argument that short sales within this market can lead to a lower offer price. But, under the UK’s institutional setting, investors who are selling shares within the when-issued market already know how many shares they were allocated. So, they are not short-sellers. As a result, the SEC’s argument may not hold under the UK’s when-issued regulatory framework.

121

Moreover, on the LSE the when-issued dealing commences only when the allocation of shares is completed and the offer price is announced (London Stock Exchange, 2012b). In contrast, in Germany the grey market trading usually begins when the subscription period commences (Aussenegg et al., 2006, Löffler et al., 2005). So, in the German market the subscription period occurs simultaneously with the grey market. As a result, the underwriter observes the information from the subscription period and the grey market bid-ask quotes, and based on this information it sets the offer price. Dorn (2009), Cornelli et al. (2006) and Aussenegg et al. (2006) report that it is mainly retail investors that trade in the grey market. This means that when underwriters set the offer price they also take into account indications of interest from noise (retail) investors who are looking for short term profits. As a result the offer price may not represent the ‘true’ price of the company based on its fundamentals. According to Cornelli et al. (2006), only when the offer price is announced (last day of the grey market), the information that the underwriter gathers from the subscription period is revealed to all investors. However, in the UK this information is disclosed much earlier as the offer price is announced first and then the when-issued dealing commences. This means that investors in the UK not only know the information from the subscription period earlier than those in Germany, but also can extract further information by observing actual when- issued dealing prices on the stock exchange. As a result, the when-issued market in the UK may be highly informative for the market price in the first day of the unconditional trading. Furthermore, the when-issued trading on the LSE usually lasts three days whereas that on the FSE lasts five to seven days (Löffler et al., 2005, Aussenegg et al., 2006, Dorn, 2009). On the ASE it usually lasts 4 to 6 days.

4 Data

We conduct our study on the Main Market of the LSE and the data includes all non-financial IPOs that took place during the period January 1996 to December 2012. The reason that our analysis starts from 1996 onwards is because the percentage of IPOs that have a when-issued market before this year is very small (less than 6%). In addition, our study is focused only on the IPOs that are taking place on the Main Market for two important reasons. The first one is that companies listed on the Alternative Investment Market (AIM) of the LSE have very different characteristics from those listed on the Main Market. More specifically, the Main Market is targeting large and well established companies that should satisfy stricter

122 admission and disclosure requirements, whereas the AIM market is mainly attracting smaller, younger and riskier companies that raise less money from their flotation (London Stock Exchange, 2013). The second reason is that the percentage of companies that have a when- issued market on AIM is much smaller than that on the Main Market (less than 4% vs. 34%).

From our data we exclude all financial IPOs, 74 IPOs that issued Global Depository Receipts on the LSE, 3 IPOs for which the prospectuses are not available, 2 IPOs for which we cannot find the stock price data and 30 companies that were categorised as IPOs, but were actually listed on other stock exchanges before30. So, our final sample consists of 341 non-financial IPO firms. 116 of them have a when-issued trading whereas the remaining 225 do not.

The information of whether a company has a when-issued market is collected from the IPO prospectuses, as the firms have to explicitly disclose whether a conditional trading will take place. In addition, the offer price, gross proceeds, market capitalisation, secondary shares sold in the IPO, total assets, revenues, date of incorporation, commission paid to the underwriters and the number of the syndicate are also collected from the IPO prospectuses. The stock price data, such as closing, opening, bid and ask daily prices, even within the when-issued market period, are extracted from Datastream, as these prices are publicly available. The data for the IPO daily volume and the levels of the FTSE all share index are collected from Datastream.

5 Methodology

5.1 When-issued market price accuracy

We examine the accuracy of the when-issued market prices by following and extending Löffler’s et al. (2005) approach. We make use of the pricing errors which are defined as the percentage difference between the when-issued market closing prices and the closing price in the first day of unconditional trading. More specifically the when-issued price error is defined as:

30 For instance, according to the LSE, Exillon Energy conducted an IPO on 17 December 2009. However, this company was admitted to the PLUS market for the trading of its global depository receipts on 16 February 2009 (p. 178 of its prospectus) and was withdrawn from PLUS on 20 August 2009 (p. 189 of its prospectus). So, Exillon Energy is not a real IPO, since it was listed on the PLUS market before, and this is why it is excluded from our sample.

123

When-issued price error = (푃퐶푃 − 푃푊퐼푖)/푃푊퐼푖 (1)

where 푃퐶푃 = closing price in the first day of unconditional trading, and

푃푊퐼푖 = the closing price in day i of the when-issued market.

We also calculate the offer price error which is defines as:

Offer price error = (푃퐶푃 − 푃푂푃)/푃푂푃 (2)

where 푃푂푃 = offer price.

In order to evaluate whether the when-issued market prices or the offer price are better proxies of the closing price in the first day of unconditional trading we compare their corresponding errors.

In addition, the Mincer-Zarnowitz test for unbiasedness is used in order to examine whether the when-issued market prices are unbiased estimators of the true price, which is approximated by the closing price in the first day of unconditional trading (Löffler et al., 2005). We run the following Ordinary Least Squares (OLS) regression:

푃퐶푃 = 푎 + β푃푊퐼푖 + 휀푖 (3)

The when-issued closing prices are unbiased estimators of the closing price in the first day of the unconditional trading if a = 0 and β = 1 (joint null hypothesis). We also test whether the offer price is an unbiased estimator by substituting the PWIi with the POP in regression (3). We extent Löffler’s et al. (2005) approach by splitting the IPOs that have a when-issued market into two groups, those that have underpricing and those that have overpricing or zero return in the first day of the unconditional trading.

5.2 Determinants of the when-issued market and how the conditional trading affects the offer price?

124

The decision to have a when-issued market may be a non-random one and may also have an effect on the setting of the offer price. It may be the case that IPO firms choose to self-select into their preferred choices. So, the decision to have a conditional trading may be an endogenous choice made by the decision maker (IPO firm) (Li and Prabhala, 2007). This is true as companies rarely make decisions randomly (Hamilton and Nickerson, 2003). As a result, if the IPOs that have a when-issued market are not a random subset of the whole population then OLS regressions do not yield consistent estimates. In order to take this selection bias into account we use the two stage Heckman (1979) selection model.

More specifically, we use the endogenous switching model, which is an extension of the baseline Heckman self-selection model. We make use of a ‘what-if’ type of analysis because we only observe the offer price for companies that have a when-issued market, but we cannot observe what would the price be had the same company chosen not to have a when-issued market (counterfactual). The endogenous switching approach consists of a binary choice equation which models the decision to have a when-issued market. This selection model (first stage) is the following:

∗ 퐼푖 = 푍푖훾 + 휀푖 (4)

Vector 푍푖 includes all observable independent variables that may affect the decision to have a when-issued market. Some of these variables may also have an effect on the setting of the offer price. Vector 훾 includes all parameters that need to be estimated and 휀푖 is the error term. The selection model is estimated by using a probit regression in which the dependent ∗ variable 퐼푖 is equal to one if the firm has a when-issued market and zero otherwise.

∗ ∗ 퐼푖 = 1 iff 퐼푖 > 0, and 퐼푖 = 0 iff 퐼푖 ≤ 0 (5)

The estimated value of 푍푖훾̂ is then used to generate the inverse Mills ratio (IMR), which is defined differently for the IPOs that have a when-issued market and for those that do not. The IMR is included in the second stage of the Heckman procedure in which we have two regression equations for the variable of interest conditional on the choice made in the first stage. The second stage equations are the following:

125

푦1푖 = 푋푖훽1 + 푢1푖 (6)

푦2푖 = 푋푖휃2 + 푢2푖 (7)

Equation (6) is the offer price equation for the companies that have a when-issued market, and equation (7) is the offer price equation for those that do not. In practice, we only observe either 푦1푖 or 푦2푖, for each IPO, based on the outcome of 퐼푖:

푦푖 = 푦1푖 iff 퐼푖 = 1, and 푦푖 = 푦2푖 iff 퐼푖 = 0 (8)

푦푖 includes the offer prices and 푥푖 includes the independent variables that affect the issue price when there exists a when-issued market or when there is not respectively.

The selection bias arises from the non-zero covariance between 휀푖 from equation (4) and

푢1푖 and 푢2푖 from equations (6) and (7). If we try to estimate equations (6) and (7) by OLS then we make the assumption that all factors that affect both the decision to have a when- issued market and the setting of the offer price are observable and included in the regressions. But, this is rarely the case as there may be variables that are not observed by the researcher, which will consequently lead to potential endogeneity problems.

The self-selection regression model solves the aforementioned problems by allowing the error in equation (4) to be correlated with the errors in equations (6) and (7), so that unobserved or missing variables in the binary outcome equation (4) are allowed to also affect the offer price. Parameters β1 and 휃2 cannot be estimated directly by using OLS because this will generate inconsistent estimates since the expectation of 푦1푖 does not have a zero mean

(u1 and ε are correlated).

126

This is why in the second stage we estimate equations (6) and (7) by OLS, with one additional variable added into each regression to adjust for the potential non-zero expectation of the errors. This additional regressor is the IMR, which allows equations (6) and (7) to be estimated consistently using OLS (Lee, 1978, Heckman, 1979). The IMR is defined as:

휑(푍푖훾) 퐼푀푅1 = − for IPOs that have a when-issued market. Φ(푍푖훾)

휑(푍푖훾) 퐼푀푅2 = for IPOs that do not have a when-issued market and 1− Φ(푍푖훾)

휑 is the standard normal density function and Φ is the standard normal cumulative distribution function. If at least one of the IMRs in equations (6) and (7) is statistically significant then this implies that there is self-selection and we have to use the two stage Heckman model. If none of the IMRs is statistically significant then OLS estimates are not affected by selection bias (Golubov et al., 2012, p. 291).

In this study, the analytical form of the first stage probit31 selection model is the following:

퐼푊퐼 훾0 + 훾1푆푒푐표푛푑푎푟푦 + 훾2퐼푛푣_푃푟𝑖푐푒 + 훾3퐴𝑔푒 + 훾4푇퐴

= + 훾5퐵표표푘_푡표_푀푎푟푘푒푡 + 훾6퐷푢푚푚푦_(푃푙푎푐𝑖푛𝑔)

+ 훾7퐷푢푚푚푦_(푇푒푐ℎ푛표푙표𝑔푦) (9)

+ 훾8푈푛푑푒푟푤푟𝑖푡푒푟_푅푒푝. +훾9푅푒푡_(퐹푇푆퐸 퐴푙푙 푠ℎ푎푟푒 𝑖푛푑푒푥)

+ 훾10푉표푙_(퐹푇푆퐸 퐴푙푙 푠ℎ푎푟푒 𝑖푛푑푒푥) + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠 + 휀

where 퐼푊퐼 is a dummy variable that takes the value of one if the IPO has a when-issued market and zero otherwise, Secondary is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds), Inv_Price is the inverse of the offer price and is calculated as: (1/offer price), Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age), TA is the total assets of the firm the year prior to the IPO and is calculated as the natural logarithm of the total assets, Book_to_Market is the book value divided by the market value of equity, Dummy_(Placing) is a dummy

31 Tucker (2010) reports that the calculation of the IMR requires the numerator and denominator functions to be normally distributed. This is why we should make use of a probit, and not a logit model.

127 variable that takes the value of one if the company raises money through a placing and zero otherwise, Dummy_(Technology) is a dummy variable that takes the value of one if the IPO is a technology company and zero otherwise, Underwriter_Rep. (GP and N of IPOs) is the reputation of the underwriter based on the gross proceeds and number of IPOs raised and advised by the underwriter during the 2 years before the IPO respectively, expressed as a percentage of the total gross proceeds and the total number of IPOs raised and occurred during these two years and is multiplied by 100, Ret_(FTSE All share index) is the return of the FTSE all share index during the period two months prior to the ten days before the commencement of the first day of trading, Vol_(FTSE All share index) is the volatility of the FTSE all share index during the period two months prior to the two days before the commencement of the first day of trading and is multiplied by 100, Yearly_Dummies are the yearly dummies for the examined period and 휀 is the stochastic error term.

The second stage OLS regressions are the following:

푂푃1 = 훽0 + 훽1푆푒푐표푛푑푎푟푦 + 훽2퐴𝑔푒 + 훽3푇퐴 + 훽4퐵표표푘_푡표_푀푎푟푘푒푡

+ 훽5퐷푢푚푚푦_(푇푒푐ℎ푛표푙표𝑔푦) (10) + 훽6푈푛푑푒푟푤푟𝑖푡푒푟_푅푒푝. +훽7푉표푙_(퐹푇푆퐸 퐴푙푙 푠ℎ푎푟푒 𝑖푛푑푒푥)

+ 훽8퐼푀푅1 + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠 + 푢1

푂푃2 = 휃0 + 휃1푆푒푐표푛푑푎푟푦 + 휃2퐴𝑔푒 + 휃3푇퐴 + 휃4퐵표표푘_푡표_푀푎푟푘푒푡

+ 휃5퐷푢푚푚푦_(푇푒푐ℎ푛표푙표𝑔푦) (11) + 휃6푈푛푑푒푟푤푟𝑖푡푒푟_푅푒푝. +휃7푉표푙_(퐹푇푆퐸 퐴푙푙 푠ℎ푎푟푒 𝑖푛푑푒푥)

+ 휃8퐼푀푅2 + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠 + 푢2

Where 푂푃1 and 푂푃2 are the offer prices for IPOs that have a when-issued market and those that do not respectively. The independent variables in equations (10) and (11) have already been explained above. Equation (10) includes the observations only for the companies that have a when-issued market whereas equation (11) includes those for the firms that do not have a when-issued market.

128

The endogenous switching regression model allows us to conduct a counterfactual analysis on the alternative choice which we cannot observe. The counterfactual estimates are computed by multiplying the coefficient estimates obtained from the second stage regressions with the explanatory variables. For instance, in order to calculate the offer price for the IPOs that have a when-issued market had they not have a conditional trading we multiply the coefficient estimates from equation (11) with the independent variables from equation (10). In this way we obtain the offer prices if the alternative scenario had occurred. Then we compare the estimated offer prices with the actual ones and infer whether the decision to have a grey market has any effect on the offer price.

The independent variables that are included in the probit model (equation (9)) are chosen from the previous academic literature and the LSE’s when-issued dealing requirements. One of these requirements is that there must be sufficient demand for the security within the when-issued market. As a result, we would expect riskier companies not to have a when- issued market because there may not be sufficient demand for their shares. Due to this reason we include four variables as proxies for ex ante company specific uncertainty, which are the following: the money raised from selling existing shares in the IPO as a percentage of the total gross proceeds, the inverse of the offer price, the book to market ratio and the age of the firm. We expect the first three variables to have a negative relationship with the probability of having a when-issued market, whereas the age is expected to have a positive relationship with the dependent variable.

In addition, we control for the underwriter reputation because, according to the certification hypothesis (Shiller, 1989), investors decide whether to buy the shares of an IPO based on the quality of the underwriter. As a result the more reputable the underwriter is, the higher the demand will be for the IPO shares. We use four different measures of underwriter reputation. Two of them are based on the gross proceeds and number of IPOs raised and advised by the underwriter during the 2 years before the IPO takes place, and then divided by the total gross proceeds and number of IPOs within these two years’ time period. There are two reasons for choosing a period of two years. Firstly, it takes some time for underwriters to build a good reputation. Secondly, some reputable underwriters may choose not to underwrite any issues in a depressed market in order to avoid damaging their reputation with a poor IPO (Goergen et al., 2006).

129

The remaining two measures of underwriter reputation are constructed based on the gross proceeds and number of IPOs raised and advised throughout the time period 1996 to 2012, and then divided by the total gross proceeds and number of IPOs during the aforementioned period. Dorn (2009), Aussenegg et al. (2006), Fang (2005) and Megginson and Weiss (1991) also used the market share to compute the reputation of the underwriters. The logic for calculating the reputation in this way is that underwriters are repeated players in the market and their survival and future income depends directly on their reputation. For this reason, reputable underwriters will be very selective about the IPOs that they bring to the market throughout their life and will avoid sponsoring overpriced IPOs, for which there may not be sufficient demand by investors.

The LSE’s when-issued dealing requirements also state that the security should be sufficiently liquid, taking into account the size of the issue. We proxy the size of the issue with the total assets that the company has the year before the IPO. Moreover, the when- issued dealing requirements report that the company’s broker has to submit a draft and a final when-issued dealing application forms at least ten and two business days before the commencement of the when-issued trading respectively. In order to capture the market return prior to the start of the when-issued dealing we calculate the return of the FTSE All share index for the two months period before the ten days prior to the commencement of the first day of trading. In addition, we calculate the volatility of the FTSE All share index during the two months period prior to the two days before the start of the first day of trading. Aussenegg et al. (2006) and Löffler et al. (2005) use similar measures.

Cornelli et al. (2006), Aussenegg et al. (2006) and Dorn (2009) report that it is mainly retail investors that trade in the grey market. In order to examine whether companies have a when- issued market because they want to attract demand from retail investors we include a dummy variable that takes the value of one if the company is raising money through a placing and zero otherwise. Placings are only sold to institutional investors, whereas other forms of raising capital (i.e. global offer) sell part of their shares to retail investors. As a result, IPO firms that are only targeting institutional investors (placings) may be less likely to have a when-issued trading. The reason is that the conditional trading has informational value for investors, because any private information that may exist at the time of the IPO will be incorporated in the conditional trading prices. This informational value will mainly be useful to retail investors, who are usually less informed when compared to institutional investors.

130

This implies that IPO companies, which, apart from institutions, are also targeting retail investors, will conduct an IPO through a conditional dealing. So, this dummy is used only as a proxy for retail investor demand. Furthermore, Cornelli et al. (2006) and Dorn (2009) control for technology companies in their empirical analysis. This is why we also include a dummy variable that takes the value of one if the company is a technology company and zero otherwise.

In the second stage regression, in which the dependent variable is the offer price, all the independent variables from the first stage are included except from the inverse of the issue price, the dummy placing and the return of the FTSE All Share index for the two months before the ten days prior to the commencement of the first day of trading. More specifically, we exclude the inverse of the offer price because the dependent variable in the second stage is the offer price. We also exclude the dummy placing because we use it here only as a proxy for retail investor demand. The return of the FTSE index is not included because it is calculated for the two months starting from the tenth day before the commencement of the conditional trading. So, we would expect not to affect the issue price as it does not include the return of the index for the ten days preceding the setting of the offer price.

5.3 Volume in the first day of trading

In order to examine whether the when-issued market has any effect on the volume in the first day of trading we run the following OLS regression.

푌푖 = 푋푖훽 + 휀푖

= 훽0 + 훽1퐷푢푚푚푦푊퐼 + 훽2푆푒푐표푛푑푎푟푦 + 훽3퐼푛푣_푃푟𝑖푐푒 + 훽4퐴𝑔푒

+ 훽5푇퐴/푀퐶 + 훽6퐵표표푘_푡표_푀푎푟푘푒푡 + 훽7퐷푢푚푚푦_푇푒푐ℎ푛표푙표𝑔푦 (12) + 훽8푈푛푑푒푟푤푟𝑖푡푒푟_푅푒푝. +훽9푅푒푡푢푟푛_(퐹푇푆퐸 퐴푙푙 푠ℎ푎푟푒 𝑖푛푑푒푥)

+ 훽10푉표푙푎푡𝑖푙𝑖푡푦_(퐹푇푆퐸 퐴푙푙 푠ℎ푎푟푒 𝑖푛푑푒푥) + 푌푒푎푟푙푦_퐷푢푚푚𝑖푒푠

+ 휀푖

where 푌푖 is the volume in the first day of trading and is calculated as: (Volume in the first day of trading/Outstanding shares), Dummy_(WI) is a dummy variable that takes the value of one if the company has a when-issued market and zero otherwise, TA/MC is the total

131 assets divided by the market capitalisation, Return_(FTSE All share index) and Volatility_(FTSE All share index) are the return and volatility of the FTSE All share index for the month prior to the first day of trading. All the other variables have already been defined before.

The independent variables included in equation 12 are selected from the existing empirical literature. Two of them are of particular importance, the size of the company and the when- issued dummy. The size is captured by the ratio of the total assets divided by the market capitalisation. The reason that we do not use the total assets or the market capitalisation individually as proxies for the size is because the correlation between the two aforementioned variables and when-issued dummy is high.

6 Results

6.1 Descriptive statistics

Table 2 reports the total number of non-financial IPOs and those that had and did not have a when-issued market during the period January 1996 to December 2012. Approximately 34% of all IPOs listed on the Main Market of the LSE had conditional trading during the aforementioned period. This is in contrast to the German grey market in which 94% of the IPOs in Germany have a grey market (Aussenegg et al., 2006) and 95% of the IPOs that had a conditional trading were listed in the Neuer Markt (Dorn, 2009), which was mainly targeting small and medium size, young technology companies (Jenkinson and Ljungqvist, 2001, Löffler et al., 2005, Aussenegg et al., 2006, Dorn, 2009, Vismara et al., 2012) that face less stringent entry requirements in terms of age, size and a record of profitability. Cornelli et al. (2006) report that from June 1997 to July 2002 the percentage of IPOs that had a when- issued dealing on the LSE was only 4.1%. This is due to the fact that they also include IPOs that were admitted to trading on AIM in their sample.

From Table 2 it is evident that during the mid-90s, the percentage of IPOs that had a conditional trading was very small and gradually started to increase. In 1996 and 1997 this percentage was less than 9% for each year and increased to more than 49% during the period 2002 to 2012, reaching its highest level in 2008 and 2011 in which all IPOs listed on the LSE had a conditional trading.

132

Table 3 provides information for the companies that have a when-issued market (Panel A) and compares them with those that do not (Panel B). The when-issued period on average lasts for 3.6 days. Out of all the IPOs that had conditional trading, 61% of them had a when- issued period of 3 days. This is consistent with the LSE’s requirements (London Stock Exchange, 2012b), according to which the when-issued dealing typically lasts for three business days and the stock exchange may also consider other trading periods. In addition, the return in the first day of the when-issued trading is 11.96%.

The bid-ask spread during the when-issued period is very small, approximately 1.6%, which suggests that the conditional trading in the UK is very liquid. This is in stark contrast with that of other developed European markets, in which the bid-ask spread is at least five times higher than that of the UK. For instance, the bid-ask spreads within the grey market for Germany, France, Spain and Austria are 10.2%, 8.3%, 10.3% and 13.8% respectively (Cornelli et al., 2006). So, the argument that institutional investors do not participate in the grey market because the bid-ask spread is very wide may not hold in the case of the UK. Cornelli et al. (2006) report a bid-ask spread of 3% for the LSE, which is double the size of ours. Probably this is due to the fact that they also include AIM IPOs in their analysis, whereas we do not. The volume (shares traded during the when-issued market as a percentage of the outstanding shares) in the first day of the when-issued trading is much higher than the average volume during the remaining days of the conditional trading (19.74% vs. 2.71%). This implies that the vast majority of the trades within the when-issued market take place during the first day. In the German grey market the volume is only 0.48% (Löffler et al., 2005). As a result, the when-issued market is much more liquid in the UK when compared to that of Germany’s. One possible explanation may be the fact that smaller and younger companies usually have a conditional trading in the German market (Cornelli et al., 2006), and these companies are expected to be less liquid, when compared to larger and more established firms, as is the case on the Main Market of the LSE.

6.2 Univariate analysis

Panel B of Table 3 compares the two groups of IPOs, those that have a when-issued market and those without. From this table it is evident that firms with conditional trading have very different characteristics from those that do not. More specifically, when-issued IPOs are

133 younger (5.19 vs. 7.75 years), less risky (inverse of issue price) and better quality IPOs as they sell their shares at higher prices (£3.4 vs. £1.7), when compared to the non when-issued IPOs. According to Fang (2005), Klein and Leffler (1981), Shapiro (1983) and Allen (1984) higher prices are an indicator of superior quality. The reasoning for this argument is that when quality is ex ante unobserved, as is the case for companies that are listed for the first time in the stock exchange, a higher price ensures better quality because the present value of future income exceeds the short term profit made from selling low quality securities at high prices.

Also, the when-issued IPOs are less underpriced as the median (mean) return in the first day of unconditional trading is 7.22% (12.88%), whereas the equivalent for the non when-issued IPOs is 10.8% (16.4%). In the German market the underpricing for the companies that have grey market is more than triple the size of that in the UK, and ranges from 40% to 50% (Aussenegg et al., 2006, Cornelli et al., 2006, Dorn, 2009). Some potential explanations for that may be the different regulatory set up of the when-issued trading between the two markets which may lead to the attraction of companies with very different characteristics (i.e. riskier IPOs). Consistent with this explanation, the vast majority of IPOs that had grey market in the FSE were listed on the Neuer Markt, which was targeting smaller companies that were characterised by more informational asymmetries. So, we would expect them to incur higher underpricing. Another explanation may be the time period used in the German papers which includes the hot market period of the year 2000.

Furthermore, companies that have a conditional trading are much larger (£1440 vs. £96 millions market cap. and £1130 vs. £40 millions total assets), raise more money from the IPO (£416 vs. £38 millions) and have higher revenues (£1340 vs. £63 millions) than those that do not. This is also consistent with the LSE requirements (London Stock Exchange, 2012b) that companies need to satisfy in order to have a when-issued market.

As far as the number of the syndicate is concerned, when-issued IPOs are on average underwritten by 4.3 investment banks whereas the no when-issued ones are underwritten by only 1.42 underwriters. This is again different from the German market in which both groups of IPOs are underwritten by approximately the same number of syndicate (Dorn, 2009). Firms that have a conditional trading in the UK pay a higher commission to their underwriters (3.23% vs. 2.11%) and are underwritten by more reputable investment banks.

134

One potential explanation for this may be the fact that prestigious investment banks underwrite less risky issues (as is the case with the companies that have when-issued trading), obtain higher issue prices and receive higher compensation because they offer high quality services (Fang, 2005, Chemmanur and Fulghieri, 1994).

The volume in the first day of unconditional trading for the when-issued IPOs is much smaller than that for the no when-issued group (1.65% vs. 7.6%). However, as we reported in Panel A of Table 3 the vast majority of the volume takes place in the first day of the when- issued market. If we compare the volume of the first day of the unconditional trading, for the no when-issued IPOs, with that of the first day of the conditional trading, for the when- issued IPOs, then it is more than two times smaller (7.6% vs. 19.74%). If we compare the average volume for the first month in the unconditional trading, excluding that of the first day for the two groups, then the when-issued firms have a significantly higher volume than the no when-issued ones (0.8% vs. 0.46%). So, as we move from the conditional dealing period to the unconditional trading, the volume drops significantly for the when-issued IPOs as most of the trading takes place within the conditional trading period. This pattern is exactly the opposite of that in the German market in which the volume increases in the grey market as we are approaching closer to the unconditional trading date.

The bid-ask spread during the first month in the unconditional trading for the when-issued IPOs is approximately half of that for the IPOs that do not have when-issued trading. This suggests that IPOs with conditional trading have a more liquid aftermarket trading. In addition, the return of the FTSE all share index two months before the ten days prior to the commencement of the first day of trading is significantly different between the two samples (1.2% vs. 2.22%). The money raised from the selling of existing shares in the IPO and the volatility of the FTSE all share index for the two months prior to the two days before the commencement of the first day of trading, are not significantly different from zero for the two IPO groups.

Table 4 provides the correlation matrix of the independent variables that will be used later in the two stage Heckman model and the volume regression. All in all, the companies that have a when-issued market have very different characteristics than those that do not as they are less risky, raise more money, have higher revenues, market capitalisation, total assets, are underwritten by more reputable investment banks and are more liquid.

135

6.3 Informational accuracy of the when-issued market

In order to examine if conditional trading prices are good proxies of the closing price in the first day of unconditional trading we compute pricing errors (equations (1) and (2) in part 6.1 of this study) as reported in Table 5 for the whole sample of IPOs that have a when- issued market (Panel A), those that experience overpricing or zero returns (Panel B) and those that are underpriced (Panel C) in the first day of unconditional trading respectively. From Panel A of Table 5 it is quite evident that the mean pricing errors during the when- issued market are much smaller that the offer price error. For instance, the mean price error in the first day of the when-issued period is only 1.03%, whereas the mean offer price error is much higher (12.88%) and the difference in means between these two errors is statistically significant at 1% significance level. In addition, the mean pricing errors within the when- issued period reduce significantly from the first to the last day of trading (1.03% vs. 0.36%). The pricing error in the last day of the conditional trading period is only 0.36% which implies that the price in the last day of the when-issued period is almost equal to that in the first day of unconditional trading. This reduction in the pricing error, as we reach closer to the first aftermarket trading day, means that the pricing accuracy gradually increases towards the unconditional IPO date. This improvement in the accuracy may be due to two reasons. The first one is that the accuracy of information increases due to the arrival of public information and the second one is that private information gradually incorporates into the when-issued prices (Löffler et al., 2005).

The standard deviation of the when-issued errors is much smaller than that of the offer price error (7.2%, 5.7% and 3.9% vs. 26%) which implies that when-issued prices are not only less biased, but also more efficient estimates of the price in the first day of unconditional trading when compared to the offer price. Also, the pricing errors in the when-issued period are not significantly different from zero whereas the offer price error is. In addition, the null hypothesis that there is no difference between the mean when-issued pricing errors and the mean offer price is rejected as the p-values of the difference in means are zero.

All the aforementioned results indicate that the when-issued prices are significantly more informative than the offer price and are very good proxies for the closing price in the first day of unconditional trading. The when-issued errors, offer price error and standard

136 deviations on the Main Market of the LSE are much smaller than those of the German grey market as reported by Löffler et al. (2005), which implies that the informational accuracy of the UK when-issued market is much higher than that of the German one. One potential explanation may be the fact that in Germany investors can only observe bid and ask quotes and not actual when-issued dealing prices, as is the case in the UK. This is due to the fact that in Germany the offer price is set only in the last day of the grey market, whereas in the UK is determined before the commencement of the when-issued trading. This may also explain the fact that in the German grey market an investor who buys shares within this market and sells them in the first day of unconditional trading would on average lose money as the grey market quotes are significantly higher than the first aftermarket price (Löffler et al., 2005, Dorn, 2009). However, this is not the case on the LSE, in which the when-issued prices are on average smaller than the first aftermarket price.

We extent Löffler’s et al. (2005) analysis by splitting the when-issued IPO sample into those that experience overpricing or zero returns in the first day of unconditional trading (Panel B) and those that are underpriced. The results are similar to Panel A and when-issued prices are better proxies for the first day closing price in the unconditional trading when compared to the offer price, regardless of whether the issue is underpriced, overpriced or has a zero return.

We also test whether the when-issued prices and the offer price are unbiased estimators of the closing price in the first day of unconditional trading (Table 6) by using the Mincer- Zarnowitz test for unbiasedness (equation (3), part 6.1). Our expectation is that only when- issued prices will be unbiased estimators of the ‘true price’ (closing price in the first day of unconditional trading). As expected, from Table 6 it is evident that the null hypothesis of unbiasedness is accepted for all the regressions that include as an independent variable the midpoint and the last prices in the when-issued trading respectively. Although the number of IPOs that are overpriced and have zero return is only 30 firms, we still run OLS regressions for this subsample to test the unbiasedness joint hypothesis. For the whole sample of IPOs even the first day price in the when-issued market is an unbiased estimator of the first day closing price in the aftermarket. This is in contrast to Löffler’s et al. (2005) results according to which the first and midpoint range of the subscription period are not unbiased predictors of the first closing price in the unconditional trading. Moreover the R- squares in our regressions are much higher when compared to those of Löffler’s et al. (2005). From Table 6, the unbiasedness joint hypothesis for the offer price is rejected in all three

137 panels which means that the offer price is not an unbiased estimator of the closing price in the first day of unconditional trading.

The results from Tables 5 and 6 document that the informational asymmetries that exist during an IPO will be much lower for companies that have a conditional trading, as when- issued prices appear to be highly informative. So, any informational disadvantage that uninformed (usually retail) investors may have is substantially reduced through the when- issued market. As a result, retail investors will be more keen to invest in IPOs that have a when-issued trading. According to Löffler et al. (2005), the when-issued market can alleviate the winner’s curse problem.

As we mentioned before the LSE reports that the conditional trading provides investors with two important advantages (London Stock Exchange, 2012b): facilitates price formation and offers investors the earliest opportunity to agree on an entry/exit price of the IPO. In the following two sub-parts of this study (7.3.1 and 7.3.2) we provide evidence of these two advantages.

6.3.1 The when-issued market facilitates price formation

The when-issued market facilitates price formation ahead of the unconditional admission as the vast majority of the trading volume takes place in the first day of the conditional dealing market. As we already reported in Table 3 the volume of trading in the first day of the when- issued market is 19.74%, which is much higher than the average volume during the remaining days of the when-issued market (2.71%) and the volume in the first day of unconditional trading (1.65%). As a result, most of the price formation takes place in the first day of the when-issued market. This is in contrast to the German grey market in which the volume during the when-issued market is smaller than that in the aftermarket (0.48% average volume within the grey market and 0.55% in the 30th day of secondary market) (Löffler et al., 2005).

If investors observe that there is a positive return in the first day of the conditional trading (when-issued price is higher than offer price) then this will also continue in the first day of unconditional trading since most of the price change/formation is taking place within the when-issued market. From Table 4 and 5 we already showed that the when-issued prices are

138 unbiased estimators of the closing price in the first day of the unconditional trading. So, investors know whether the IPO is underpriced or overpriced from the very first day of the when-issued market and do not have to wait until the commencement of the unconditional trading in order to obtain this information. This is very different from the German grey market in which investors who buy shares in the grey market have to wait up until the last day of the conditional trading, when the offer price is announced, to find out if they bought overpriced or underpriced shares.

6.3.2 The when-issued market allows investors the earliest opportunity to agree an entry or exit price

The when-issued market offers investors the earliest opportunity to decide an entry or exit price. For an investor who is allocated shares, he has the option to sell his shares from the very first day of the when-issued market and make an average return of 11.96%, instead of having to wait until the commencement of the unconditional trading in order to do that. This investor can even wait up until the first day of unconditional trading, sell his shares and make an average return of approximately 13%. This is due to the fact that this investor knows that when-issued market prices are very good proxies of the first day price in the unconditional trading. So, if on average IPOs are underpriced in the first day of the when-issued market then this underpricing will also continue in the first day of the unconditional trading. As a result, investors who are allocated shares have the earliest opportunity to decide on an exit price.

But, how about an investor who is not allocated shares? Panel A of Table 7 shows that if an investor buys shares of an underpriced IPO in the first day of the when-issued market and sells them in the first day of unconditional trading then he will make a return of 2.5%. If an investor wants to replicate the same return, but this time by buying the shares in the first day of unconditional trading, then even after two weeks (ten working days) from the commencement of the unconditional trading he will not be able to do it, as the return from the first to the tenth day of the unconditional trading is 1.75%. In the same way, if an investor decides to short sell shares of an overpriced offering in the first day of the conditional trading and then buy them back in the first day of unconditional trading then he will make a return of 3.2% (Panel B). If this investors wants to replicate the same return, but by short selling shares in the first day of the unconditional trading, then even after two weeks from the start

139 of the aftermarket trading he will not be able to achieve it, as the return from the first to the tenth day of the unconditional trading is 2.93%.

As a result, the when-issued market gives investors the earliest opportunity to decide on an entry or exit point because it will take these investors more than two weeks to replicate the return that can be achieved from the when-issued market period. This is mainly due to the fact that most of the price formation/change takes place within the when-issued market, and after that the price does not change a lot.

6.4 Probit regressions

In order to examine why some companies have a conditional trading and others do not we make use of a probit model (Table 8) in which the dependent variable is a dummy that takes the value of one if the IPO has a when-issued market and zero otherwise (equation (4) from part 6.2). To the best of our knowledge this is the first empirical study that looks at the determinants of the when-issued market. In Germany they cannot perform a similar analysis because most of the IPOs have a grey market. Aussenegg et al. (2006) report that 94% of the IPOs taking place on the FSE have a conditional trading.

From Table 8 it is evident that the proceeds from secondary shares, the inverse of the issue price and the book to market ratio have a negative relationship with the dependent variable. This means that the probability of having a when-issued market will be smaller the higher the proceeds raised from the selling of existing shares in the IPO are, the riskier the company is and the less future growth opportunities it has. This is expected as one of the requirements in order to have a when-issued market is that there will be sufficient demand for the security. So, we would expect riskier companies to attract less demand from investors and consequently not satisfy the aforementioned requirement.

Furthermore, the probability of having a when-issued market is also affected by the size of the company (total assets as a proxy for the size). The larger the company the higher the probability of having a when-issued trading. This is consistent with the LSE’s requirement, according to which a security can be traded in the when-issued market if it is sufficiently liquid, taking into account the size of the IPO. Moreover, the more reputable the underwriter, the higher the probability that the IPO will have a when-issued market. The rationale for that

140 is that underwriters are repeated players in the IPO market and they are concerned about their reputation. So, reputable investment banks will avoid bringing to the market overpriced IPOs for which there is not sufficient demand and consequently these IPOs will not satisfy one of the LSE’s requirements in order to have a when-issued market.

In addition, the dummy variable that takes the value of one if the company is raising money through a placing 32 and zero otherwise is statistically significant and has a positive relationship with the dependent variable. One potential explanation for that is that the when- issued trading is highly informative for the quality of the IPO and any private information that may exist at the time of the flotation will be incorporated in the conditional trading prices, thus reducing information asymmetries around the IPO date (Löffler et al., 2005). This release of information is mainly useful to retail investors, who are usually less informed when compared to institutional investors. As a result, companies, which, apart from institutions, are also targeting retail investors, will conduct an IPO through a conditional dealing.

Although the age of the firm has a positive relationship with the dependent variable, it is statistically significant only in Model 1, but not in Model 2. So, we cannot make any inferences for the effect of this variable. The decision to have a when-issued market is not affected by the volatility or the return of the FTSE all share index prior to the IPO. In addition, the dummy variable that takes the value of one if the company is a technology firm has no effect on the dependent variable.

To sum up, companies that are larger, less risky, have higher growth opportunities, are underwritten by more reputable underwriters and want to attract retail investors for their shares are more likely to have a when-issued market. This is in contrast to the German market, in which the vast majority of IPOs that went to the grey market were listed on the Neuer stock exchange, which is a market mainly for smaller and riskier IPOs. This may be explained from the fact that the when-issued market in the UK is regulated by the LSE, and companies need to satisfy certain requirements in order to have a conditional trading, whereas this is not the case in the German grey market.

32 Placings are only sold to institutional investors, whereas other forms of raising capital through an IPO (i.e. global offer) sell part of their shares to retail investors.

141

6.5 Does the decision to have a when-issued market affect the offer price?

In this part we will test whether hypothesis (4) in part 4 holds. In order to do that we make use of the two stage Heckman procedure and take into account any potential self-selection bias that may exist between the decision to have a when-issued market and the setting of the offer price. The results obtained from the first stage (reduced form) probit regressions in Table 8 (equation (4) in part 6.2) are used to construct the IMR. In the second stage regression the natural logarithm of one plus the offer price33 is regressed on the IMR and the independent variables separately for the two IPO groups, those that have when-issued market and those that do not (equations (6) and (7) in part 6.2). The use of the IMR adjusts for the selectivity problem that arises from the fact that we can only observe the companies that have a when-issued market but cannot observe what would have happened if these firms did not have a conditional trading. In Table 9, we report the second-stage regressions based on model 1 of Table 8. To the best of our knowledge this is the first study that examines whether the decision to have a when-market has an effect on the offer price.

The results reported in Table 9 show that one of the two IMRs is statistically significant, which suggests that there is selectivity bias, and without them OLS regressions would yield biased and inconsistent estimates. For the companies that have a when-issued market the total assets and the reputation of the underwriter have a positive relationship whereas the age of the firm has a negative relationship with the issue price respectively. This implies that firms that are larger, underwritten by more reputable underwriters and younger will set a higher issue price. For the companies that do not have a when-issued market the total assets and the technology dummy variable have a positive relationship, whereas the book to market ratio has a negative relationship, with the dependent variable respectively. As a result, for the no warrant IPO group the offer price will be higher if the total assets are higher, the firm is a technology company or the firm has higher future growth opportunities.

The coefficients in Table 9 are used to estimate what the offer price would have been had the alternative choice (have a when-issued market or not) been made. More specifically, the

33 The offer price is expressed in £ and not in pences and this is why we take the logarithm of one plus the offer price, as there are offer prices that have a value less than £1.

142 forecasts are computed as the product of the coefficient estimates from the second-stage regressions (Table 9) and the independent variables. Then, we compare these forecasts with the actual offer price, as reported in Table 10.

For the IPOs that have a grey market the mean actual offer price is £3.4 but would have been 54% lower (£1.55) if the company decided not to have a when-issued market. This means that the conditional trading has an effect on the issue price. As a result, the underwriters do not offer the advantages of the when-issued market to investors for free, but they instead charge them ‘rents’ through a higher issue price. But, why would investors, who were allocated shares, be willing to pay a higher offer price? One potential reason may be the fact that the when-issued market has certain advantages for investors.

For the IPOs that do not have a when-issued market their actual offer price is £1.7 but would have been £3.14 if they would have had a conditional trading. Consistent with our previous argument, underwriters would have charged a higher issue price if the IPOs would have had a when-issued market. So, under the UK’s when-issued institutional setting, the SEC’s argument that the grey market will lead in a reduction of the offer price does not hold.

But, if companies can sell their shares at a higher price because they will have a when-issued market, then why not all of them have a conditional trading? The answer to this question is that the when-issued market is regulated by the LSE and some companies may not satisfy the requirements in order to have a conditional trading (i.e. sufficient demand and liquidity). For example, from Table 3 it is quite obvious that there is a huge difference between the size of the when-issued and no when-issued IPOs (average market capitalisation of £1440million vs. £96million respectively).

Apart from selling the shares at a higher issue price, do companies have a when-issued market in order to avoid litigation risk? In other words, do companies use the conditional trading period as a form of insurance against future litigation? According to the litigation risk hypothesis IPO firms and their underwriters intentionally underprice the IPO shares in order to avoid future liability (Ibbotson, 1975, Tinic, 1988, Ibbotson et al., 1994, Lowry and Shu, 2002). An IPO company that underprices its shares by a greater amount has a lower offer price, which implies that it has a lower probability of being sued. As a result, someone may argue that IPOs have a when-issued trading simply because they want to find out what

143 the demand for their shares is. Base on this demand they will set the offer price at a level which is lower than the expected market value of the securities because this will decrease the risk of future litigation. Although this argument may be applicable for the grey market in other European markets, i.e. Germany, in which the grey market occurs first and then the issue price is set, it does not hold in the case of the UK. As we mentioned earlier, on the LSE the when issued market will commence only when the issue price is determined and the allocation of shares is completed. So, the issue price is already know before the commencement of the conditional trading period.

In addition, the underpricing for the companies that do not have a when-issued market is higher when compared to that of those that have a conditional trading (16.4% vs. 12.88%). But, according to the litigation hypothesis, the underpricing should be higher for the when- issued IPOs. Moreover, to the best of our knowledge, there is no IPO that was withdrawn within the LSE’s when-issued market period. However, as we have already reported before in other European grey markets there exist IPOs that were withdrawn during the grey market and may potentially raise litigation concerns. As a result, we cannot claim that the companies listed on the LSE use the grey market as a form of insurance to avoid litigation risk.

6.6 Volume in the first day of trading

In Table 3 we report that the volume in the first day of trading is 19.74% and 7.6% for companies that have a when-issued market and those that do not respectively. In addition, the size of the companies between the two groups of IPOs is very different, as the firms that have a when-issued market are much larger (in terms of total assets and market capitalisation) than those that do not. So, someone can argue that the difference in the volume of trading simply exists because of the difference in the size of the firms, and consequently the existence of a when-issued market does not have any effect on the volume. In order to examine that we run an OLS34 regression in which the dependent variable is the volume (expressed as a percentage of the outstanding shares) in the first day of trading (Table 11).

34 We initially used the two stage Heckman model but none of the IMRs in the second stage volume regressions was statistically significant. This implies that the OLS estimates are not affected by selection bias. If this is the case then we can run OLS regressions and we can control for the existence of the when-issued dealing by including a dummy variable in the right hand side of equation (12) (Golubov et al., 2012, p. 291).

144

In this regression we control, among other factors, for both the size35 of the firm and the existence of the when-issued market.

From Table 11 it is evident that the when-issued dummy variable, that takes the value of one if there is a when-issued market, and zero otherwise, is statistically significant at 1% level, although we control for the size of the company. The volume in the first day of trading will approximately be 14% higher for companies that have when-issued market when compared to those that do not. As a result, the large difference between the volume in the first day of trading cannot be explained by the difference in the size of the two IPO groups, but by the existence of a when-issued market. The volume is also affected by the future growth opportunities of the companies (book to market ratio) as the higher the future growth opportunities are, the higher the volume will be. Moreover, the volatility that exists in the market the month before the first day of trading has a negative relationship with the volume, which implies that the higher the volatility in the market the lower the volume will be.

6.7 Robustness tests

In order to check the robustness of our results we conduct the following robustness checks:

6.7.1 Heckman two stage model

We replicate the two stage Ηeckman model by using the other three measures of underwriters’ reputation:

1. The reputation based on the number of IPOs advised by the underwriter the previous 2 years before the IPO takes place, divided by the total number of IPOs within these two years’ time period. 2. The reputation based on the gross proceeds and number of IPOs raised and advised by the underwriter throughout the time period 1996 to 2012, divided by the total gross proceeds and number of IPOs during the aforementioned period.

35 The reason that we make use of the variable total assets divided by market capitalisation and not just total assets or market capitalisation is because the two aforementioned variables are correlated with the dummy variable that takes the value of one if there is when-issued market and zero otherwise.

145

The results under all aforementioned measures of reputation are qualitatively the same.

In addition, our findings suggest that there is a negative relationship between the percentage of gross proceeds raised from the selling of existing shares and the probability of having a when-issued market. However, it may be argued that the selling of existing shares in the IPO will only send a negative signal to the market if it exceeds a certain threshold, as companies usually sell a mixture of new and existing shares. In order to examine this argument we create a dummy variable that takes the value of one if the percentage of secondary proceeds is higher than the median of this variable (33.49%) and zero otherwise. Then we substitute the percentage of secondary proceeds with this dummy variable in the probit model and replicate the same analysis. Furthermore, we conduct again the aforementioned analysis by creating a dummy variable, but this time it is based on the mean, and not the median. The results obtained are qualitatively the same. Moreover, instead of calculating the return and volatility of the FTSE all share index for a two month period, we also compute them for one and three months’ period respectively. Once again the results are qualitatively similar.

6.7.2 Volume OLS regression

We make use of the underwriters’ reputation measures described above and we calculate the return and volatility of the FTSE all share index for different time periods (two and three months before the IPO date). The results are qualitatively the same under all different variables. In addition, we also match the IPOs of the two subsamples according to the market capitalisation based on a 20%, 30% and 40% bands (Bodnaruk et al., 2009). The results are qualitatively similar, as the when issued dummy is still statistically significant and has a positive effect on the volume in the first day of trading.

7 Conclusion

In this paper we study the when-issued dealing on the Main Market of the LSE during the period 1996 to 2012. To the best of our knowledge this is the first empirical study that examines what are the determinants of the conditional trading and why some companies have a when-issued market and others do not. In addition, this is the first paper that describes in details and compares the UK’s when-issued regulatory setting with that of other developed markets around the world. One of the main findings is that the decision to have a when-

146 issued market affects the setting of the offer price. More specifically, for companies that have a conditional trading the actual offer price is £3.4 but would have been 54% lower (£1.55) had these firms not had a when-issued market. The reason is that the conditional trading has certain advantages for investors (price formation, earliest entry/exit price) and the underwriters do not offer these advantages for free, but they instead charge investors with ‘rents’ that take the form of a higher offer price. Furthermore, for companies that do not have a when-issued market the actual offer price is £1.7, but would have been £3.14 had they had a conditional trading. So, under the LSE’s institutional setting the SEC’s argument, according to which the grey market for IPOs is not allowed in the US stock exchanges because it will lead to a lower offer price, does not hold.

Moreover, the probability of having a when-issued market is higher for companies that are larger, less risky, with higher future growth opportunities and more reputable underwriters. This is in contrast to the German grey market in which the companies that have a conditional trading are smaller than those that do not (Dorn, 2009). One potential explanation for that may be the fact that the when-issued regulatory framework between the UK and Germany are very different, as the when-issued market in the UK is closely regulated and monitored by the LSE, whereas that in Germany is over the counter/non-regulated. This may also explain the fact that in Germany there are IPOs that were withdrawn within the grey market, whereas this has never happened in the UK as all the IPOs that commenced their trading in the when-issued dealing were successfully listed on the LSE. Also, the when-issued market affects the volume in the first day of trading and for companies that have a conditional dealing the volume of trading will approximately be 14% higher than that of the companies that do not have a when-issued market. In addition, the UK’s when-issued market appears to be highly informative for investors and much more informative than that of Germany’s. One potential reason may be the fact that the offer price and allocation of shares are set and completed before the commencement of the conditional trading, whereas in the German market they take place in the last day of the grey market.

147

References

Allen, F., 1984. Reputation and product quality. The Rand Journal of Economics 15, 311- 327.

Aussenegg, W., Pichler, P., Stomper, A., 2006. IPO pricing with bookbuilding and a when- issued market. Journal of Financial and Quantitative Analysis 41, 829.

Benveniste, L. M., Spindt P. A., 1989. How investment bankers determine the offer price and allocation of new issues. Journal of Financial Economics 24, 343-361.

Bodnaruk, A., Massa M., Simonov A., 2009. Investment banks as insiders and the market for corporate control. Review of Financial Studies 22, 4989-5026.

Chemmanur, T. J., Fulghieri P., 1994. Investment bank reputation, information production, and financial intermediation. The Journal of Finance 49, 57-79.

Cornelli, F., Goldreich D., Ljungqvist A., 2006. Investor sentiment and pre-ipo markets. The Journal of Finance 61, 1187-1216.

Dorn, D., 2009. Does sentiment drive the retail demand for ipos? Journal of Financial and Quantitative Analysis 44, 85.

Fang, L. H., 2005. Investment bank reputation and the price and quality of underwriting services. The Journal of Finance 60, 2729-2761.

Goergen, M., Khurshed, A., Mudambi, R., 2006. The Strategy of Going Public: How UK Firms Choose Their Listing Contracts. Journal of Business Finance & Accounting 33, 79- 101.

Golubov, A., Petmezas, D., Travlos, N. G., 2012. When It Pays to Pay Your Investment Banker: New Evidence on the Role of Financial Advisors in M&As. The Journal of Finance 1, 271-311.

148

Hamilton, B. H., Nickelson, J. A., 2003. Correcting for Endogeneity in Strategic Management Research. Strategic Organisation 1, 51-78.

Hanley, K. W., 1993. The underpricing of initial public offerings and the partial adjustment phenomenon. Journal of Financial Economics 34, 231-250.

Heckman, J. J., 1979. Sample selection bias as a specification error. Econometrica 47, 153- 161.

Ibbotson, R., 1975. Price performance of common stock new issues. Journal of Financial Economics 2, 235–272.

Ibbotson, R., Sindelar, J., Ritter, J., 1994. The market’s problems with the pricing of initial public offerings. Journal of Applied Corporate Finance 7, 66–74.

Jenkinson, T., Ljungqvist, A., 2001. Going public: The theory and evidence on how companies raise equity finance. Oxford University Press on Demand.

Klein, B., Leffler K. B., 1981. The role of market forces in assuring contractual performance. Journal of Political Economy 89, 615-641.

Lee, L.-F., 1978. Unionism and Wage Rates: A Simultaneous Equations Model with Qualitative and Limited Dependent Variables. International Economic Review 19, 415-433.

Li, K., Prabhala, N. R., 2007. Self-Selection Models in Corporate Finance. Handbook of Corporate Finance: Empirical Corporate Finance, 47-62.

Löffler, G., Panther P. F., Theissen E., 2005. Who knows what when? The information content of pre-ipo market prices. Journal of Financial Intermediation 14, 466-484.

London Stock Exchange, 2012a. Rules of the London Stock Exchange. http://www.londonstockexchange.com/traders-and-brokers/rules-regulations/rules-lse.pdf

149

London Stock Exchange, 2012b. When Issued Dealing. http://www.londonstockexchange.com/traders-and-brokers/rules-regulations/when-issued- dealing-guidance.pdf

London Stock Exchange, 2013. A comparison of the eligibility criteria and continuing obligations for the Main Market (Premium and Standard) and AIM. http://www.londonstockexchange.com/companies-and-advisors/main- market/companies/primary-and-secondary-listing/premiumstandardandaimcomparison.pdf

Lowry, M., Shu, S., 2002. Litigation risk and IPO underpricing. Journal of Financial Economics 65, 309–335.

Megginson, W. L., Weiss, K. A., 1991. Venture Capitalist Certification in Initial Public Offerings. The Journal of Finance 46, 879-903.

Rock, K., 1986. Why new issues are underpriced. Journal of Financial Economics 15, 187- 212.

Shapiro, C., 1983. Premiums for high quality products as returns to reputations. The Quarterly Journal of Economics 98, 659-679.

Shiller R. J., 1989. Initial Public Offerings: Investor Behavior and Underpricing. NBER Working Paper 2806.

The Financial Times, 2012. IPOs: outlook brightens for grey market.

Tinic, S., 1988. Anatomy of initial public offerings of common stock. Journal of Finance 43, 789–822.

Tucker, J., 2010. Selection bias and econometric remedies in accounting and finance research. Journal of Accounting Literature 29, 31–57.

Vismara, S., Paleari S., Ritter J. R., 2012. Europe's second markets for small companies. European Financial Management 18, 352-388.

150

Figure 1 Timeline of the when-issued dealing in the UK market

B A Day Day Day Day Day Day Day …. …. .... …. T-13 T-6 T-4 T-3 T-2 T-1 T

Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 when-issued dealing Unconditional period trading/ secondary market

Point B: In the first day of the when- Point A: This is the first issued period (Day T-3) the company day that the should: unconditional trading/secondary  Announce the offer price. market commences (Day  Notify investors of the T). allocation of shares.

After the allocation of shares is completed and the offer price is announced then the when-issued trading commences.

151

Table 1 Regulatory differences in the when-issued/grey market trading between the UK and the German markets. LSE stands for London Stock Exchange and FSE stands for Frankfurt Stock Exchange. In this table we report the institutional/regulatory differences between the two aforementioned markets. ‘When-issued’ market means on exchange trades, whereas ‘grey’ market refers to off exchange/over the counter trades. This is why the conditional trading in the UK is also called ‘when-issued’ dealing, whereas that in Germany is called ‘grey’ market. LSE FSE When-issued dealing takes No Yes place over the counter

Usually the last day before Usually the last day of the End of subscription period the when-issued trading grey market commences

Before the when-issued Setting of the offer price Last day of the grey market trading commences

Before the when-issued Allocation of shares Last day of the grey market trading commences

After the allocation of shares Commencement of the when- Usually when the price range is completed and the offer issued dealing is announced price is announced

Number of days that the Approximately 3 days Approximately 5 to 7 days when-issued dealing lasts

152

Table 2 Number of IPOs during the period January 1996 to December 2012 This table shows the number of non-financial IPOs that took place in the Main Market of the London Stock Exchange during the period January 1996 to December 2012. We also report the percentage of IPOs that have a when-issued market for each of the different years. WI stands for when-issued. % of IPOs with Year IPOs with WI market IPOs with no WI market Total a WI market 1996 4 57 61 6.56 1997 5 51 56 8.93

1998 5 28 33 15.15

1999 9 13 22 40.91

2000 12 45 57 21.05

2001 2 4 6 33.33

2002 7 5 12 58.33

2003 3 2 5 60

2004 13 3 16 81.25

2005 14 3 17 82.35

2006 12 5 17 70.59

2007 15 4 19 78.95

2008 3 0 3 100

2010 6 3 9 66.67

2011 4 0 4 100 2012 2 2 4 50 Total 116 225 341

153

Table 3 Comparison between IPOs that have a when-issued (WI) market and those that do not Panel A reports information only for the when-issued market and Panel B reports information for both groups of companies. WI stands for when-issued. b-a stands for the bid-ask spread. WI days is the number of days that the WI period lasts. Ret (1st day WI) is the return during the first day of the when-issued market and is calculated as (closing price in the 1st day of the WI market – issue price)/issue price. Volume is calculated as: (Volume of trades/Outstanding shares). Volume (1st day WI) is the volume in the first day of the WI market. Average Volume is the average volume in the WI market, excluding that of the first day in the WI market. WI b-a spread 1 is the percentage b-a spread during the WI period and is calculated using the formula (ask price – bid price)/ask price. WI b-a spread 2 is the relative b-a spread during the WI period and is calculated using the formula (ask price – bid price)/(( ask price – bid price)/2). Price (£) is the issue price in £. Ret (1st uncond.) is the return during the first day of the unconditional trading and is calculated as (closing price in the 1st day of unconditional trading – issue price)/issue price. Secondary Proceeds is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). GP and MC (million £) are the money raised from the flotation and the market capitalisation of the company, expressed in million £. Reven.and TA (million £) are the revenues and total assets for the year prior to the IPO, expressed in million £. N of Syndicate is the number of the syndicate that underwrites the offer. Commission (% GP) is the money paid to the syndicate for selling the IPO shares, expressed as a percentage of the GP. Reputation 1 and 2 refer to the reputation of the underwriter based on the number of IPOs it advised and gross proceeds raised during the two years before the IPO took place, expressed as a percentage of the total number of IPOs and gross proceeds advised and raised during these two years. Volume (1st uncond.) is the volume in the first day of unconditional trading. Average Volume (20 days) is the average volume during the 20 days in the unconditional trading, excluding that of the first day of the unconditional trading. b-a spread 1 and 2 refer to the percentage and relative b-a spreads respectively during the period of one month excluding the first day of the unconditional trading. Ret FTSE refers to the return of the FTSE all share index during the period two months prior to the ten days before the commencement of the first day of trading. Vol FTSE is the volatility of the FTSE all share index for the two months prior to the two days before the commencement of the first day of trading. N is the number of IPOs. IPOs with WI Difference IPOs with no WI All IPOs Difference Market in Market (N = 225) (N = 341) in Means (N = 116) Medians Mean Median Mean Median Mean Median p-value p-value Panel A WI Days 3.63 3 Ret (1st day 11.96 5.93 WI, %) Volume (1st day 19.74 10.25 WI, %) Average Volume 2.71 1.67 (%) WI b-a spread 1 1.58 1.07 (%) WI b-a spread 2 1.61 1.08 (%) Panel B Price (£) 3.4 2.3 1.7 1.5 2.28 1.7 0*** 0*** Inverse of Price 0.46 0.43 0.96 0.67 0.79 0.59 0.03** 0*** Ret (1st 12.88 7.22 16.4 10.8 15.2 9.48 0.19 0*** uncond., %) Secondary 32.81 22. 1 37.67 35.48 36.2 33.49 0.2 0.13 Proceeds (%) Age (years) 5.19 2.08 7.75 3.07 6.88 2.57 0.09* 0.08* GP (million £) 416 214 38 25 166 47 0*** 0***

154

Table 3 continues MC (million £) 1,440 553 96 65 554 129 0*** 0*** Reven. (million 1,340 235 63 19 499 30 0.02** 0*** £) TA (million £) 1,130 264 40 13 411 27 0*** 0*** N of Syndicate 4.31 4 1.42 1 2.4 2 0*** 0*** Commission (% 3.23 3 2.11 1.75 2.49 2.25 0*** 0*** GP) Reputation 1 (%) 5.90 4.76 2.36 1.82 3.57 2.45 0*** 0*** Reputation 2 (%) 9.24 5.05 2.3 0.48 4.66 0.82 0*** 0*** Volume (1st 1.65 0.95 7.6 5.9 5.57 2.44 0*** 0*** uncond., %) Average Volume 0.8 0.47 0.46 0.45 0.57 0.45 0*** 0*** (20 days, %) b-a spread 1 (%) 1.87 1.44 3.22 2.54 2.76 2.24 0*** 0*** b-a spread 2 (%) 1.92 1.45 3.51 2.57 2.97 2.27 0.04** 0*** Ret FTSE (%) 1.2 1.41 2.22 2.26 1.88 1.87 0.04** 0.04** Vol FTSE (%) 0.79 0.73 0.77 0.71 0.78 0.72 0.35 0.2

***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

155

Table 4 Correlation Matrix WI (Dummy_(WI)) is a dummy variable that takes the value of one if the company has a when-issued market and zero otherwise. Secondary (Secondary) is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (proceeds from existing shares/gross proceeds). Inv_Price (Inv_Price) is the inverse of the offer price and is calculated as: (1/offer price). Age (Age) is the number of years from incorporation to flotation and is calculated as the natural logarithm of: ln (1+age). TA (TA) is the total assets of the firm the year prior to the IPO and is calculated as the natural logarithm of the total assets. B_T_M (Book_to_Market) is the book value divided by the market value of equity. Placing (Dummy_(Placing)) is a dummy variable that takes the value of one if the company raises money through a placing and zero otherwise. Technology (Dummy_(Technology)) is a dummy variable that takes the value of one if the IPO is a technology company and zero otherwise. Rep. (GP) (Underwriter_Rep. (GP)) is the reputation of the underwriter based on the gross proceeds raised during the 2 years before the IPO, expressed as a percentage of the total gross proceeds raised during these two years. Rep. (N IPOs) (Underwriter_Rep. (N of IPOs)) is the reputation of the underwriter based on the number of IPOs advised during the 2 years before the IPO, expressed as a percentage of the total number of IPOs that took place during these two years. Ret. (Ret_(FTSE All share index)) is the return of the FTSE all share index during the period two months prior to the ten days before the commencement of the first day of trading. Vol. (Vol_(FTSE All share index)) is the volatility of the FTSE all share index during the period two months prior to the two days before the commencement of the first day of trading. TA/MC (TA/MC) is the total assets divided by the market capitalisation. Return and Volatility (Return_(FTSE All share index) and Volatility_(FTSE All share index)) are the return and volatility of the FTSE all share index for the month prior to the first day of trading. In brackets we provide the names of the variables that will be used later in the two stage model and the volume regression. Rep. Rep. (N WI Secondary Inv_Price Age TA B_T_M Placing Technology Ret_(FTSE) Vol_(FTSE) TA/MC (GP) IPOs) WI 1 Secondary -0.07 1 Inv_Price -0.27 -0.11 1 Age -0.11 0.20 -0.01 1 TA 0.64 0.06 -0.37 0.00 1 B_T_M 0.00 0.08 0.08 0.21 0.24 1 Placing -0.72 0.09 0.22 0.14 -0.44 -0.02 1 Technology -0.04 -0.01 -0.03 -0.07 -0.32 -0.11 -0.05 1 Rep. (GP) 0.39 -0.09 -0.18 -0.15 0.30 -0.04 -0.38 0.06 1 Rep. (N 0.36 -0.02 -0.16 -0.14 0.32 -0.08 -0.32 -0.01 0.77 1 IPOs) Ret. -0.11 0.06 -0.04 -0.01 -0.05 0.04 0.14 -0.11 -0.07 -0.05 1 Vol. 0.05 -0.09 -0.03 -0.08 -0.06 -0.07 -0.09 0.36 0.05 0.06 -0.41 1 TA/MC 0.14 -0.01 0.02 0.07 0.57 0.45 -0.12 -0.25 0.02 0.00 -0.03 -0.17 1 Return 0.00 0.03 -0.01 0.01 0.04 0.09 0.04 -0.02 -0.11 -0.02 0.22 0.01 0.02 Volatility 0.06 -0.08 -0.04 -0.04 -0.04 -0.06 -0.11 0.30 0.09 0.07 -0.47 0.87 -0.13 156

Table 4 continues Return Volatility Return 1 Volatility -0.27 1

157

Table 5 Accuracy of the when-issued market prices vs. offer price This table reports the pricing errors of the when-issued market prices and the offer price from the first day closing price of unconditional trading. WI stands for when-issued. The errors are calculated based on the following formulas: Error from first day in the WI = (first day closing price of unconditional trading - first day closing price in the WI)/first day closing price in the WI. Error from midpoint day in the WI = (first day closing price of unconditional trading - midpoint day closing price in the WI)/midpoint day closing price in the WI. Error from last day in the WI = (first day closing price of unconditional trading - last day closing price in the WI)/last day closing price in the WI. Error from offer price = (first day closing price of unconditional trading - offer price)/offer price. The difference in means compares whether each of the mean when-issued price errors is significantly different from the offer price error. All means and standard deviations (sd) are in percentages. Panel A includes all the IPOs that have when- issued dealing. Panel B includes only IPOs that have an overpricing or zero return in the first day of unconditional trading. Panel C includes only IPOs that have an underpricing in the first day of unconditional trading. N is the number of IPOs. Difference in When-Issued Price Errors Offer Price Error Means Mean Sd Mean p- Sd p-value p-value (%) (%) (%) value (%) Panel A: All IPOs First day in the WI 1.03 0.128 7.20 0*** Midpoint day in the 0.58 0.277 5.70 0*** WI Last day in the WI 0.36 0.312 3.90 0*** Offer price 12.88*** 0 26 N 116 Panel B: Overpricing and zero return IPOs First day in the WI -3.2** 0.02 7.10 0.09* Midpoint day in the -0.93 0.21 3.99 0*** WI Last day in the WI -0.96 0.15 3.55 0*** Offer price -5.1*** 0 6.08 N 30 Panel C: Underpriced IPOs First day in the WI 2.5*** 0 6.66 0*** Midpoint day in the 1.10 0.1 6.11 0*** WI Last day in the WI 0.4 0.19 2.78 0*** Offer price 19.15*** 0 27.36 N 86 *, **, *** indicate statistical significance at 10%, 5% and 1% respectively.

158

Table 6 Are when-issued market prices unbiased estimates of the closing price in the first day of unconditional trading? This table reports the Mincer-Zarnowitz test for unbiasedness. We run the following regression: PCP = a + βPWIi + εi. PCP is the true price and here is proxied by the closing price in the first day of unconditional trading. PWIi refers to each of the when-issued market prices or the offer price. Unbiasedness implies that a = 0 and β = 1. We use the Wald test for the joint null hypothesis Ho: a = 0 and β = 1 and the p-values are reported in the last column. The p-values for the individual coefficients are reported in parentheses. We make use of robust standard errors. WI stands for when-issued. Panel A includes all the IPOs that have a when-issued dealing. Panel B includes only IPOs that have an overpricing or zero return in the first day of unconditional trading. Panel C includes only IPOs that have an underpricing in the first day of unconditional trading. N is the number of IPOs. constant coef. R square p-value for Ho: constant=0; coef.=1 Panel A: All IPOs First day price in WI 0.095 0.98*** 0.99 0.226 (-0.282) (0) Midpoint day price in WI 0.095 0.98*** 0.99 0.275 (-0.202) (0) Last day price in WI 0.026 0.99*** 0.998 0.724 (-0.521) (0) Offer Price -0.524 1.33*** 0.871 0*** (-0.118) (0) N 116 Panel B: Overpricing and zero return IPOs First day price in WI -0.07* 0.99*** 0.992 0.181 (0.075) (0) Midpoint day price in WI -0.052 1.01*** 0.999 0.303 (0.129) (0) Last day price in WI -0.026 0.99*** 0.999 0.405 (0.184) (0) Offer Price -0.12*** 0.99*** 0.998 0*** (0.002) (0) N 30 Panel C: Underpriced IPOs First day price in WI 0.159 0.98*** 0.991 0.01*** (0.121) (0) Midpoint day price in WI 0.133 0.97*** 0.992 0.11 (0.116) (0) Last day price in WI 0.045 0.99*** 0.998 0.34 (0.35) (0) Offer Price -0.62* 1.42*** 0.875 0*** (0.086) (0) N 86 ***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

159

Table 7 Returns from the first day of conditional and unconditional trading WI stands for when-issued. Return 1 (from the 1st day of WI) is calculated based on the following formula: (first day closing price of unconditional trading - first day closing price in the WI)/first day closing price in the WI. Return 2 (5 days unconditional) is calculated based on the following formula: (fifth day closing price of unconditional trading - first day closing price of unconditional trading)/first day closing price of unconditional trading. Return 3 (10 days unconditional) is calculated based on the following formula: (tenth day closing price of unconditional trading - first day closing price of unconditional trading)/first day closing price of unconditional trading. All means and standard deviations (sd) are in percentages. Panel A includes IPOs that have an underpricing in the first day of unconditional trading. Panel B includes IPOs that have an overpricing or zero return in the first day of unconditional trading. N is the number of IPOs. Mean p-value (%) Panel A: Underpriced IPOs Return 1 (from the 1st day of WI) 2.5 0*** Return 2 (5 days unconditional) 0.29 0.8 Return 3 (10 days unconditional) 1.75 0.16 N 86 Panel B: Overpriced and zero return IPOs Return 1 (from the 1st day of WI) -3.21 0.02** Return 2 (5 days unconditional) -1.54 0.09* Return 3 (10 days unconditional) -2.93 0.09* N 30 ***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

160

Table 8 Probit regressions for the IPOs with and without when-issued market The dependent variable is a dummy variable that takes the value of one if the IPO has a when-issued market and zero otherwise. Secondary is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Inv_Price is the inverse of the offer price and is calculated as: (1/offer price). Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of one plus age: ln (1+age). TA is the total assets of the firm the year prior to the IPO and is calculated as the natural logarithm of the total assets. Book_to_Market is the book value divided by the market value of equity. Dummy_(Placing) is a dummy variable that takes the value of one if the company raises money through a placing and zero otherwise. Dummy_(Technology) is a dummy variable that takes the value of one if the IPO is a technology company and zero otherwise. Underwriter_Rep. (GP and N of IPOs) is the reputation of the underwriter based on the gross proceeds and number of IPOs raised and advised by the underwriter during the 2 years before the IPO respectively, expressed as a percentage of the total gross proceeds and the total number of IPOs raised and occurred during these two years and is multiplied by 100. Ret_(FTSE All share index) is the return of the FTSE all share index during the period two months prior to the ten days before the commencement of the first day of trading. Vol_(FTSE All share index) is the volatility of the FTSE all share index during the period two months prior to the two days before the commencement of the first day of trading and is multiplied by 100. The variables Inv_Price, Book_to_Market and Underwriter_Rep. are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. Model 1 Model 2 marginal p- marginal Variables coeff. coeff. p-value effect value effect Intercept -10.38*** 0 -9.87*** 0 Secondary -0.99** -0.16 0.04 -1.13** -0.22 0.02 Inv_Price -2.59*** -0.43 0 -2.36*** -0.45 0 Age 0.2* 0.03 0.09 0.17 0.03 0.17 TA 0.61*** 0.10 0 0.6*** 0.12 0 Book_to_Market -1.24** -0.20 0.04 -0.93* -0.18 0.09 Dummy_(Placing) -1.67*** -0.35 0 -1.69*** -0.4 0 Dummy_(Technology) 0.34 0.07 0.37 0.55 0.13 0.15 Underwriter_Rep. (GP) 0.07*** 0.01 0 Underwriter_Rep. (N of IPOs) 0.14*** 0.03 0 Ret_(FTSE All share index) -1.89 -0.32 0.59 -3.87 -0.74 0.27 Vol_(FTSE All share index) 0.45 0.07 0.51 -0.34 -0.07 0.57 Yearly_Dummies Yes Yes % of correct predictions 92% 93% Pseudo R-square (%) 75% 74% N 341 341 ***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

161

Table 9 Second stage regression estimates of the offer price The dependent variable is the natural logarithm of: ln (1+ offer price). Secondary is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (gross proceeds from existing shares/total gross proceeds). Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of: ln (1+age). TA is the total assets of the firm the year prior to the IPO and is calculated as the natural logarithm of the total assets. Book_to_Market is the book value divided by the market value of equity. Dummy_(Technology) is a dummy variable that takes the value of one if the IPO is a technology company and zero otherwise. Underwriter_Rep. (GP) is the reputation of the underwriter based on the gross proceeds raised during the 2 years before the IPO and is multiplied by 100, expressed as a percentage of the total gross proceeds raised during these two years. Vol_(FTSE All share index) is the volatility of the FTSE all share index for the two months prior to the two days of the first day of trading and is multiplied by 100. IMR is the inverse Mills ratio which is used to adjust for selectivity bias. For the IPOs which have a when-issued market the IMR is defined as -φ(ψ)/Φ(ψ) and for those which they do not is φ(ψ)/(1-Φ(ψ)) respectively. φ(ψ) is the standard normal density function and Φ(ψ) is the standard normal cumulative distribution function. The variables Book_to_Market and Underwriter_Rep. are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. IPOs with when-issued IPOs without when-issued

market market coeff. p-value coeff. p-value Intercept 0.23 0.75 0.32 0.27 Secondary 0.37 0.78 0.54 0.4 Age -0.09* 0.07 0.02 0.45 TA 0.04** 0.05 0.04** 0.04 Book_to_Market -0.13 0.52 -0.32*** 0 Dummy_(Technology) -0.08 0.61 0.13* 0.07 Underwriter_Rep. (GP) 0.01* 0.06 0.01 0.14 Vol_(FTSE All share index) 0.29 0.12 -0.04 0.7 IMR 0.13 0.23 0.29*** 0 Yearly_Dummies Yes Yes R-square (%) 31% 15% N 116 225 ***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

162

Table 10 Comparison between the actual and estimated offer price The table compares the average actual offer price (£) with the estimated. For instance, the average actual offer price for the companies that have a when-issued market is £3.4, but it would have been £1.55 if the company did not have a when-issued market. Average offer price for the 116 IPOs that have Average offer price for the 225 IPOs that do

a when-issued market not have a when-issued market Estimated offer Estimated offer Difference in Difference in Actua price if the Actua price if the Means/Median Means/Median l offer company did not l offer company had a s s price have a when-issued price when-issued (p-value) (p-value) market market

Mean 3.4 1.55 0*** 1.7 3.14 0***

Media 2.3 1.54 0*** 1.5 2.66 0*** n

***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

163

Table 11 Does the WI market affect the volume? This table reports the results of an OLS regression in which the dependent variable is the volume in the first day of trading and is calculated as: (Volume in the first day of trading/Outstanding shares). Dummy_(WI) is a dummy variable that takes the value of one if the company has when-issued market and zero otherwise. Secondary is the percentage of gross proceeds raised from the selling of existing shares in the IPO and is calculated as: (proceeds from existing shares/gross proceeds). Inv_Price is the inverse of the offer price and is calculated as: (1/offer price). Age is the number of years from incorporation to flotation and is calculated as the natural logarithm of: ln (1+age). TA/MC is the total assets divided by the market capitalisation. Book_to_Market is the book value divided by the market value of equity. Dummy_(Technology) is a dummy variable that takes the value of one if the IPO is a technology company and zero otherwise. Underwriter_Rep. (GP) is the reputation of the underwriter based on the gross proceeds raised during the 2 years before the IPO, expressed as a percentage of the total gross proceeds raised during these two years. Return_(FTSE All share index) and Volatility_(FTSE All share index) are the return and volatility of the FTSE all share index for the month prior to the first day of trading. The variables Inv_Price, TA/MC, Book_to_Market and Underwriter_Rep. are winsorised at the 1st and 99th percentiles respectively to control for outliers. N is the number of observations. N is the number of observations. Yearly dummies are included in the regressions but are not reported. We make use of robust standard errors. Dependent Variable: Volume coeff. p-value Intercept 0.13*** 0

Dummy_(WI) 0.14*** 0

Secondary 0.059 0.26

Inv_Price 0.0001 0.99

Age -0.006 0.69

TA/MC 0.02 0.93

Book_to_Market -0.1* 0.08

Dummy_(Technology) 0.06 0.45

Underwriter_Rep. (GP) 0.35 0.35

Return_(FTSE All share index) -0.64 0.19

Volatility_(FTSE All share index) -9.19** 0.03

Yearly_Dummies Yes R-square (%) 16% N 341 ***, **, * indicate statistical significance at 1%, 5% and 10% significance levels respectively.

164

Chapter 5

Conclusion

1 Summary of results

This thesis examines the non-cash compensation paid to the underwriters/brokers during the flotation process and the IPO when-issued market in the United Kingdom (UK). In both aforementioned areas there is very limited academic literature and this makes them very interesting topics for research.

In the first essay I examine the issue of non-cash compensation (i.e. warrants) to brokers as part of their compensation package in non-underwritten offerings on the Alternative Investment Market (AIM) of the London Stock Exchange (LSE). To the best of my knowledge, this is the first empirical analysis that examines the use of warrants in non- underwritten offerings. In the US the same study cannot be conducted due to the fact that the vast majority of best effort offerings issue warrants to their advisers. My main finding is that IPO firms are able to make efficient decisions and choose the contract that minimises their costs of going public. More specifically, for companies that issue warrants to their brokers the total costs of going public are 22.74%, but would have been 25.61% had they not issued them. This 2.87% reduction in costs is equivalent to 70.34% of the commission (2.87%/4.08%) paid to the brokers by the company. The main source of this decrease in the costs is the lower underpricing the firm incurs by issuing non-cash compensation to its broker. The actual underpricing is 15.95%, but would have been 22.12% had the firm not granted these warrants. One possible explanation is the fact that companies, by using warrants, can credibly signal to the market that they do not sell overpriced securities. The reason is that warrants link the compensation of the brokers with the aftermarket performance of the stock price. So, if the stock price increases in the aftermarket, the value of warrants will also increase leading to a higher compensation for the brokers. My results shed light on the ongoing debate whether IPO firms overpay their underwriters in order to obtain a listing on the stock exchange. As we already mentioned, this is not the case on AIM as the brokers, by including warrants in their compensation package, minimise their costs of going public.

165

The second essay examines the use of non-cash compensation by underwritten offerings in a market with an institutional setting that is very different from that of the US, in which the existing academic literature has been focused on. While the US has several regulations underlying the use of non-cash compensation, the AIM market of the LSE carries none of them. One of the main findings is that companies which are cash constrained have a higher probability to issue warrants. In addition, brokers appear to have the ability to time the issue of warrants as they include them in their compensation package when the market is doing well. The rationale is that if the market is doing well, then the company’s stock price will increase in the aftermarket. As a result, the value of warrants will also increase and the compensation of the brokers will be higher. The two aforementioned explanations (cash constrained companies and timing of the issue of warrants) may explain why brokers include warrants in their compensation package and not get paid all their fees in cash, as there is no regulation on AIM that restricts their cash compensation. In addition, the total costs of going public for the underwritten IPOs are 28.9, but would have been 43.96% had they not issued warrants. My results are consistent with those of Dunbar (1995) who suggests that, since the underwriters do not force the IPO firms to accept costlier compensation contracts, the National Association of Securities Dealers (NASD) should relax the non-cash compensation regulations as they are unnecessarily restrictive for financial intermediaries. My results further suggest that even in an environment with almost no regulations underlying the non- cash compensation, IPO firms are still able to minimise their floatation costs.

In the third essay I examine the when-issued dealing on the Main Market of the LSE during an extensive period of time, 1996 to 2012. One of the main findings is that the decision to have a when-issued market affects the setting of the offer price. More specifically, for companies that have a conditional trading the actual offer price is £3.4 but would have been 54% lower (£1.55) had these firms not had a when-issued market. The reason is that the conditional trading has certain advantages for investors (price formation, earliest entry/exit price). The underwriters do not offer these advantages for free, but they instead charge investors with ‘rents’, which take the form of a higher offer price. Furthermore, for companies that do not have a when-issued market the actual offer price is £1.7 but would have been £3.14 had they had a conditional trading. So, under the LSE’s institutional setting the Securities and Exchange Commission’s argument, according to which the grey market for IPOs is not allowed in stock exchanges in the United States (US) because it will lead to a lower offer price, does not hold.

166

Moreover, the probability of having a when-issued market is higher for companies that are larger, less risky, with higher future growth opportunities and underwritten by more reputable underwriters. This is in contrast to the German conditional market in which the companies that have a when-issued dealing are smaller than those that do not (Dorn, 2009). One potential explanation may be the fact that the when-issued regulatory framework between the UK and Germany is very different, as in the UK it is closely regulated and monitored by the LSE, whereas that in Germany it takes place over the counter. Also, the when-issued market affects the volume in the first day of trading and for companies that have a conditional dealing the volume of trading will approximately be 14% higher than that of the companies that do not have a when-issued market. In addition, the UK’s when-issued market appears to be highly informative for investors.

2 Future research

My findings are beneficial for a number of different parties, such as investors, market regulators, companies that seek to raise capital through an IPO and financial intermediaries. First, investors will benefit from my research in the when-issued market because I provide evidence that they are charged a higher offer price if they want to buy shares that will be traded in the when-issued market. By being aware of this fact, investors can make more informative decisions because they can compare the benefits of buying securities that will be traded in the when-issued market (price formation, earliest entry/exit price) with the corresponding costs (higher offer prices).

Second, market regulators may find my results very interesting as I provide evidence that even in an institutional setting in which there are almost no regulations underlying the non- cash compensation, IPO firms are still able to minimise their costs of going public by issuing warrants to their financial intermediaries. In addition, based on my findings related to the when-issued dealing, I show that a regulated conditional trading market attracts IPO firms with very different characteristics when compared to those in an over the counter grey market. Moreover, in contrast to the Securities and Exchange Commission’s argument, I report that a regulated when-issued market does not lead to a lower offer price.

Third, my results are beneficial for firms that consider a listing on a stock exchange and their chief financial officers’ (CFOs). The reason is that firms, by issuing warrants to their advisers, are able to minimise their costs of going public. So, companies’ CFOs should be aware that compensation warrants can provide a credible signal to the market that the

167 company’s securities are not overpriced and can consequently lead to a lower discounting of the offer price required by the IPO investors.

Finally, my results are beneficial for financial intermediaries because I document that even less reputable brokers, who lack reputational capital, can certify that the offering will not be overpriced by including warrants to their compensation package. The reason is that warrants align the financial advisers’ compensation with the company’s aftermarket stock price performance in the secondary market and consequently certify that the issue will not be overpriced.

This thesis creates a lot of interesting questions for future research. As far as the use of non- cash compensation is concerned, this thesis and previous academic literature in this area have been focused on whether issuing firms can minimise their costs of going public through the use of warrants. My thesis goes one step further and provides preliminary evidence that underwriters may have the ability to time the issuance of warrants. In other words, underwriters will on average include warrants in their compensation package when the market is doing well. This implies that the value of warrants will probably increase in the aftermarket and consequently the underwriters’ fees will also increase. But, there is still one important gap in the existing non-cash compensation literature. Do the underwriters exercise these warrants? If they do, when do they decide to exercise them and what is happening to the company’s stock price afterwards? It may be the case that underwriters decide to exercise the compensation warrants when the company’s stock price is at its highest level so that they can receive maximum compensation. But, it may also be the case that underwriters do not exercise these warrants because if they do investors may consider this a negative signal for the company’s future growth and may also start selling their holdings causing a huge drop in the stock price. Answers to the aforementioned questions will provide a better understanding of the use of non-cash compensation by IPO issuers.

Further research is also needed in the when issued market area as almost all the existing academic literature has been focused on the non-regulated German grey market. But, as we already reported the volume in the first day of trading and the characteristics of the companies that have a when-issued dealing that is regulated by the stock exchange are very different from those that have a grey market which takes place over the counter. But, are these differences between regulated and non-regulated when-issued markets observed in other stock exchanges around the world? For example, the Australian Stock Exchange also

168 has a regulated conditional trading, similar to that of the LSE. Further research in this area will provide additional evidence on the effect of regulations on the type of companies that are attracted in the when-issued dealing and the volume of trading in the stock exchanges. In addition, all the theories that have been suggested in order to explain the existence of the conditional trading have been based on empirical evidence from non-regulated grey markets. But, do these theories still hold in the case of regulated when-issued markets?

Finding answers to the questions above will provide a better insight on the effect of regulations on the when-issued markets. These answers will also shed light on the actual amount that underwriters gain from the use of warrants as part of their compensation package and whether the non-cash compensation is really beneficial for IPO firms in the long term.

My research also has certain limitations, such as the non-availability of accounting data in the admission documents of the non-underwritten offerings. This restricts the hypotheses that can be tested in order to explain the use of non-cash compensation. For instance, I cannot test whether a plausible reason for the issue of compensation warrants to the brokers in non- underwritten offerings is the cash constrains. In addition, thin trading is very prevalent in the aforementioned IPO group and researchers should take this into consideration when they conduct their analysis.

169

References

Dorn, D., 2009. Does sentiment drive the retail demand for ipos? Journal of Financial and Quantitative Analysis 44, 85.

Dunbar, C. G., 1995. The use of warrants as underwriter compensation in initial public offerings. Journal of Financial Economics, 38, 59-78.

170