Essays on the determinants and costs of corporate offerings

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

2013

Marius Christoph Ziegan

Manchester Business School CONTENTS

Abstract ...... 5 Declaration ...... 6 Copyright Statement ...... 7 Acknowledgements ...... 8

Chapter 1 Introduction

1.1. Overview of empirical studies...... 9 1.2. Thesis structure ...... 14 References ...... 16

1 Chapter 2 Does influence convertible issuance?

Abstract ...... 20 2.1. Introduction ...... 21 2.2. Literature review and hypotheses development...... 23 2.2.1. issuance as a tool to reduce agency and adverse selection costs...... 23 2.2.2. Corporate governance as a tool to reduce agency and adverse selection costs...... 25 2.2.3. The interdependency between governance mechanisms...... 26 2.2.4. Hypotheses...... 27 2.3. Data ...... 28 2.3.1. Security issues...... 28 2.3.2. Proxies for corporate governance quality...... 30 2.3.3. Firm-specific control variables...... 33 2.3.3.1. Proxies for -related financing costs...... 34 2.3.3.2. Proxies for -related financing costs...... 34 2.3.3.3. Proxies for general financing costs...... 35 2.3.3.4. Macroeconomic control variables...... 35 2.4. Empirical results ...... 36 2.4.1. The choice between convertible and straight debt ...... 37 2.4.2. The choice between convertible debt and seasoned equity...... 39 2.4.3. Detailed analysis of blockholder categories ...... 40 2.4.4. returns around convertible debt announcements...... 41 2.5. Conclusion ...... 43 References ...... 45

2 Chapter 3 The signaling content of security offering proceeds

Abstract...... 71 3.1. Introduction...... 72 3.2. Research hypotheses ...... 74 3.3. Data on security issues...... 75 3.4. Research methodology...... 76 3.4.1. Controlling for the endogeneity of issue size...... 76 3.4.2. Firm- and issue-specific characteristics...... 79 3.5. Empirical results ...... 82 3.5.1. First stage results on security offering size...... 82 3.5.2. Endogeneity of issue size and security offering announcement returns...... 84 3.5.3.1. Seasoned equity ...... 84 3.5.3.2. Convertible debt ...... 86 3.5.3.3. Straight debt ...... 86 3.5.4. Post issue earnings and use of proceeds ...... 87 3.6. Conclusion ...... 89 Appendix ...... 91 References...... 92

3 Chapter 4 The costs of raising capital: New evidence

Abstract ...... 111 4.1. Introduction...... 112 4.2. Hypotheses development ...... 116 4.3. Data and methodology...... 118 4.3.1. Sample selection...... 118 4.3.2. Switching regression model ...... 119 4.3.3. Counterfactual analysis...... 121 4.3.4. Determinants of underwriter spreads for different security and distribution types ...... 121 4.4. Empirical results...... 125 4.4.1. Average underwriter spreads and total direct costs...... 125 4.4.2. Descriptive statistics on security choice determinants...... 127 4.4.3. Determinants of the choice between different security types ...... 128 4.4.4. Determinants of underwriter spreads...... 129 4.4.5. Counterfactual analysis...... 133 4.5. Conclusion...... 135 Appendix...... 137 References...... 138

Chapter 5 Conclusion

5.1. Summary of results ...... 164 5.2. Implications and suggestions for future research...... 166 References ...... 168

This thesis contains 51,661 words including title page, tables, and footnotes.

4 Abstract

The University of Manchester Marius Christoph Ziegan Doctor of Philosophy (PhD) Essays on the determinants and costs of corporate security offerings September 2013

This thesis presents three essays on the determinants and costs of corporate security offerings. The essays contribute to an ongoing debate in the literature on what determines firms’ security choice by examining the following issues: “Does corporate governance influence convertible debt issuance?”; “The signaling content of security offerings proceeds”; and “The costs of raising capital: New evidence.”

In the first essay, we explore the influence of corporate governance on firms’ choice between equity, convertible debt and straight debt. For a sample of Western European corporate security offerings between 1999 and 2010, we find that firms with weaker firm- and country-specific corporate governance are more likely to issue convertible debt. They thus use convertible debt as a substitute for corporate governance, which is confirmed by a more favorable stock price reaction to convertible debt announcements by firms with weaker corporate governance. Moreover, these results suggest that corporate governance is a significant determinant of firms’ security choice.

The second essay examines the determinants and signaling content of security offering proceeds, controlling for the endogeneity of issue size. For a sample of US equity, convertible debt and straight debt offerings between 1999 and 2011, the findings show that stockholders can partly predict issue size by analyzing firms’ funding needs and financing costs. We find that stockholders use predicted issue sizes of equity and convertible offerings as signals of growth opportunities, whilst larger than predicted issue sizes signal issuer overvaluation. For straight debt issues, we find that unpredicted issue sizes have a positive impact on announcement returns, which is consistent with them serving as a signal of growth opportunities. Further analysis of firms’ actual uses of predicted and unpredicted offering proceeds confirms these interpretations. The results shed light on previous inconsistent findings on the impact of issue size on security offering announcement returns.

The final essay examines the magnitude and determinants of direct issuance costs, controlling for firms self-selecting into different security classes, namely equity, convertible bonds, and straight bonds, and flotation methods, namely non-shelf, shelf and 144a. For a recent sample of US corporate security offerings between 1999 and 2011, findings show that the magnitude of direct issuance costs has decreased over the last decade. These costs are higher for equity than straight bond offerings and of intermediate magnitude for convertible bond offerings. Within each security class, costs are larger for non-shelf than 144a offerings, which again have larger direct issuance costs than shelf offerings. Finally, underwriter spreads are directly related to underwriter effort on due diligence, pricing and selling, and direct issuance costs are truncated by firms’ self-selection into particular security types.

5 Declaration

I, Marius Ziegan, 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.

6 Copyright Statement 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.

7 Acknowledgements

Many people have supported me and have been very helpful during the time it took me to write this thesis, for which I cannot thank them enough. Foremost, I would like to thank my two supervisors Doctor Marie Dutordoir and Professor Norman Strong. I was certainly most fortunate to have you as my supervisors. Despite all the books, articles and courses, I believe that without your continuous guidance I would not have completed this journey with the same success. Thank you very much for all the hours you have spent and all your patience in encouraging, inspiring and challenging me to improve my reasoning, my writing and finally to go beyond of what I thought would be possible. Having you as my supervisors, I could build my work on the invaluable combination of unstoppable dedication and drive, paired with enormous experience and broad knowledge. In you I found a very rewarding cooperation and I owe you too much for all that you have done for me in the past three years. Thank you very much Marie and Norman for being such great supervisors and mentors.

Special thanks also to my PhD committee for all your support, valuable comments and suggestions. Thank you, Dr. Nick Collett, Dr. Ning Gao, Dr. Maria-Teresa Marchica, Dr. Roberto Mura, and Dr. Konstantinos Stathopoulos. My thanks also go to the faculty who taught me the first year training courses, Prof. Martyn Andrews, Dr. Ralf Becker, Prof. Michael Brennan, Dr. Leonard Gill, Prof. Julie Froud, Prof. Stuart Hyde, Dr. Asad Kausar, Dr. Arif Khurshed, Prof. Richard Taffler, Dr. Alex Taylor, Prof. Ser-Huang, Poon, Prof. Richard Stapleton, and Prof. Martin Walker. A special thank you also to Mark Greenwood and Xia Hong who supported me on so many data queries during these years. Finally, I would like to thank Lynne Barlow-Cheetham, Madonna Fyne, Anusarin Lowe, and Daniel Wheatcroft for all the support from the administrative side.

A big thank you also to Sandra Dow, Alain Praet, Konstantinos Stathopoulos, Patrick Verwijmeren, an anonymous journal reviewer, and participants at the 2012 BAFA annual conference, the FMA 2012 European conference, and the 10th Day at Ghent university for their valuable comments and suggestions for turning the second chapter of this thesis into my first publication.

Special thanks go to the ESRC for awarding me a tuition fee bursary for the final two years of my studies.

8 Chapter 1 Introduction

1.1. Overview of empirical studies One of the most important challenges for any company is to have sufficient funds to finance all positive (NPV) projects. If companies cannot finance these projects from internal funds, they need to turn to the capital markets and raise the necessary capital externally. When doing so, firms need to answer two questions: (1) How much capital do we need to raise?, and (2) What is the optimal security choice between equity, convertible and straight debt? The first question depends not only on the immediate financing needs of a company, but also on expectations about future earnings and on opportunities created by favorable market conditions to increase cash holdings. The optimal choice between the different security types depends on direct and indirect financing costs relating to security types. Direct costs refer to the fees charged by an underwriter for placing and selling the security offering in financial markets. Indirect costs arise from security offering announcement returns and underpricing. The magnitude of direct and indirect costs is directly related to asymmetric information in the case of equity following the of Myers and Majluf (1984) and to the trade-off between debt-related tax advantages and and agency costs (Modigliani and Miller, 1963; Kraus and Litzenberger, 1973). Both depend on firm-specific characteristics, including firms’ corporate governance, as better corporate governance ameliorates both asymmetric information between management and shareholders and debt-related agency costs. For the second question, firms also need to decide on the optimal flotation method. My thesis presents three essays that consider (i) the optimal security choice between equity, convertible debt and straight debt, controlling for firms’ corporate governance; (ii) stockholders expectations and the signaling content of security offering proceeds, and (iii) direct issuance costs to a security offering. In particular, I contribute to the literature by examining the following issues: “Does corporate governance influence convertible debt issuance?”; “The signaling content of security offerings proceeds”; and “The costs of raising capital: New evidence.” The first essay investigates the link between corporate governance and convertible debt issuance. In recent decades convertible debt has become a major financing source for companies around the world, but despite a large body of empirical literature, firms’ motives for issuing convertible debt remain unclear (Billingsley and Smith, 1996; Lewis et al., 1999; 9 Graham and Harvey, 2001; Dutordoir and Van de Gucht, 2009; Dong et al., 2012). Our goal is to shed light on firms’ motives to issue convertible debt by trying to establish a link to firms’ corporate governance quality. The literature on convertible debt issuance predicts that convertibles can mitigate agency costs (Green, 1984; Mayers, 1998; Isagawa, 2000) and adverse selection costs resulting from asymmetric information about firm value or risk (Brennan and Kraus, 1987; Brennan and Schwartz, 1988; Stein, 1992). The corporate governance literature, in turn, documents that governance mechanisms can reduce agency and adverse selection costs (Bhojraj and Sengupta, 2003; Anderson et al., 2004; Masulis et al., 2007; Becker-Blease and Irani, 2008). As both convertible debt and corporate governance quality can reduce adverse selection costs, we test whether the relation between convertible debt issuance and corporate governance quality is substitutive or complementary. In addition, we examine whether convertible debt issuance can insulate entrenched managers from market discipline. This directly follows from Isagawa (2002) who argues that, like straight debt, convertible debt reduces the probability of a hostile , but unlike straight debt, managers can also avoid bankruptcy by forcing conversion of the bonds into equity by calling them. We examine these questions using a pan-Western European dataset of 176 convertible issues, 350 straight debt issues, and 141 seasoned equity issues made between 2000 and 2010. The European sample allows us to consider both internal (company-specific) and external (country-specific) corporate governance features (Doidge et al., 2007). Specifically, we hand collect data for seven internal and four external corporate governance characteristics and analyze firms’ security choices with multinomial logistic regressions. Our results show that firms with weaker internal and external corporate governance quality are more likely to issue convertible debt than straight debt or seasoned equity. Among corporate governance quality proxies, the impact of the presence of large shareholders and country-specific proxies on firms’ likelihood of issuing convertible debt is particularly strong. As the finding that firms with weaker corporate governance are more likely to issue convertible bonds than straight debt or equity is consistent with both a substitutive and complementary relationship, we distinguish between these possibilities by analyzing the impact of corporate governance on returns around convertible bond announcements. Confirming a substitutive relationship, results show that investors’ react more favorably to convertible bond announcements of firms with weaker corporate governance. Finally, our results indicate that corporate governance has a statistically and economically significant impact on security choice and should be considered when examining firms’ security choice.

10 This essay contributes to the security choice literature by providing new insights into firms’ motivations to issue convertible debt instead of straight debt or equity. We are the first to test Isagawa’s (2002) rationale for convertible debt issuance and show that firms with weaker corporate governance are more likely to issue convertible bonds than straight debt or equity. The latter is consistent with theories arguing that firms issue convertible debt to reduce agency and adverse selection costs associated with other security classes. Moreover, this essay contributes to the corporate governance literature by showing that convertible debt issuance is part of a bundle of corporate governance features that serve to protect shareholder wealth (Ward et al., 2009). It also adds further evidence to the corporate governance literature arguing that firms use corporate governance mechanisms as substitutes (e.g., Westphal and Zajac, 1994; Rediker and Seth, 1995; Agrawal and Knoeber, 1996; Singh and Davidson, 2003; Rutherford et al., 2007) rather than complements (Danielson and Karpoff, 1998; Cremers and Nair, 2005; Schepker and Oh, 2013). The second essay examines the determinants and signaling content of security offering proceeds. It starts from the premise that firms raising external capital have to decide not only the optimal security choice but also how much finance they need to raise. The literature mainly focuses on the security choice decision (e.g. Marsh, 1982; Bayless and Chaplinsky, 1991; Jung et al., 1996; Lewis et al., 1999) and shows that stockholders use firms’ security choice as a signal regarding firm overvaluation following the adverse selection model of Myers and Majluf (1984). In particular, these studies find that equity issues provide the most negative signal, whereas straight bond issues carry no signaling content. Similar to the security choice, the choice of how much capital to raise may reflect managerial private information regarding the firms’ current or future prospects, and as such serve as a signal to the market. The potential signaling content of issue size has, so far, been largely ignored by the literature. Our paper intends to fill this gap by developing and testing three hypotheses on the stock price impact of offerings proceeds. The earnings shortfall hypothesis follows the theoretical findings of Miller and Rock (1985) and predicts a negative market reaction, as larger than expected offerings signal a shortfall in earnings. The overvaluation hypothesis also predicts a negative impact of issue size on security offering announcement returns, as Krasker (1986) shows that issue size directly relates to overvaluation. Finally, the growth opportunities hypothesis predicts a positive impact, as it presumes that larger issue sizes might serve as a signal of profitable growth opportunities (Ambarish et al., 1987). We examine the determinants of issue size and test these three hypotheses on a sample of US corporate seasoned equity, convertible and straight debt offerings between 1999 and

11 2011. We use a research design that is a direct extension of the Heckman (1979) two-step procedure and allows for the selection variable to be continuous (Garen, 1984). In a first step we separately estimate issue size for all three security types using a wide range of variables proxying for funding needs and financing costs. In the second stage we analyze both the stock price impact of predicted issue size and of unexpected issue size. Garen’s (1984) methodology also allows us to test whether the impact of managerial private information on stock returns varies with issue size. We find that issue size is positively related to funding needs for each security type and negatively related to debt- and equity-related financing costs for more debt-and equity(-linked) securities, respectively. We find a positive impact of predicted issue size on announcement returns for both equity and convertible debt and show that larger offerings are more likely to be used to finance growth than to accumulate cash. Confirming the expected earnings shortfall and overvaluation hypotheses, we find a negative impact of unexpected issue size on security offering announcement returns for both equity and convertible debt. Further tests on firms’ future earnings and use of proceeds show that residual issue size is linked to overvaluation rather than an expected shortfall in earnings. For large debt issues, we show that unexpected issue size is linked to growth opportunities. This essay extends the literature on securities issuance in several ways. We show that stockholders use not only the type of security announced, but also the offering size, as a signal regarding the value of firms’ assets in place and growth opportunities. Additionally, we present the factors that determine issue size for all three security classes, namely equity, convertible debt, and straight debt. The few studies examining these determinants focus on a few selective determinants for individual security types (Chaplinsky and Hansen, 1993 and Lee and Masulis, 2009). Finally, our results shed light on mixed previous findings on the impact of issue size on security offering announcement returns (e.g., Eckbo, 1986; Hess and Bhagat, 1986; Chaplinsky and Hansen, 1993; Bayless and Chaplinsky, 1996; Bethel and Krigman, 2004) by controlling for the endogeneity of issue size. The third essay examines the magnitude and determinants of security offering direct issuance costs. Direct issuance costs are negotiated prior to the offering and represent the direct compensation to the syndicate (underwriter spreads) and other direct expenses (registration fee and printing, legal and auditing costs). Lee et al. (1996) report average total direct issuance costs of 7.1% of offering proceeds for seasoned equity offerings, 3.8% for convertible bonds and 2.2% for straight bonds issued between 1990 and 1994 by US corporations. These numbers show that direct issuance costs are a substantial cost to issuing firms. The goal of this essay is to update these numbers for all three security types for

12 a recent sample of security offerings between 1999 and 2011, also taking into account different flotation mechanisms, namely non-shelf, shelf and 144a. We also examine whether there are economies of scale with respect to issue size for the various security types. Another important question we address is what factors determine direct issuance costs for security offering firms. To answer this question the existing literature studies security offering underwriter spreads, but focuses on individual security types and a limited set of explanatory variables (e.g., Altinkilic and Hansen, 2000; Livingston and Zhou, 2002; Ang and Zhang, 2006; Autore et al., 2008). A goal of this essay is to examine underwriter spread determinants across different security classes (equity, convertible, and straight debt) and distribution mechanisms (non-shelf, shelf, and 144a). In considering this issue, it is important to recognize that firms are not randomly assigned to specific security offerings, but rather self-select into security types. Underwriter spreads may thus not be comparable across security types, as they depend on the characteristics of the firm choosing a particular offering. We address this self-selection using a switching regression model (Li and Prabhala, 2007), which allows us to control for unobservable factors influencing firms’ security choice and lets us obtain unbiased estimates of the impact of underwriter spread determinants on fees. This also allows us to calculate counterfactual underwriter spreads to determine the fee that an issuer would have paid if it had issued another security type instead. We are thus able to establish whether, when making the security choice decision, firms try to minimize underwriter spreads in a similar way as they choose the offering with the least negative announcement returns (Dutordoir and Hodrick, 2012). Our results show that direct issuance costs have decreased for all three security classes compared to the findings of Lee at al. (1996) for a sample from the beginning of the 1990s. Consistent with Lee et al., we show that direct issuance costs are larger for equity than for convertible bonds, which are again larger than for straight bonds. For the different distribution mechanisms we find that direct issuance costs are larger for non-shelf than for shelf offerings, and of immediate magnitude for 144a offerings. With respect to economies of scale, on a univariate basis, we only find evidence of scale economies for both types of equity offerings and shelf convertible debt. Moreover, we find no evidence of diseconomies of scale for any security type. The latter changes in a multivariate setting controlling for both marginal and fixed costs following Altinkilic and Hansen (2000). In particular, we find evidence for economies of scale, i.e. decrease in fixed costs, for non-shelf equity and shelf debt, and evidence for diseconomies of scale, i.e. increase in marginal costs, for shelf and 144a straight bonds. Examining the determinants of underwriter spreads, our results show

13 that fees increase with proxies capturing underwriters’ due diligence, pricing, and marketing efforts. Moreover, we find that fee determinants differ across security and distribution mechanisms. For example, underwriting fees of convertible bond issues are significantly higher for offerings underwritten by reputable underwriters, which is consistent with Fang (2005) who shows that underwriters charge a premium for risky security offerings carrying a higher amount of asymmetric information. We also find an increase in underwriter spreads for shelf offerings made during the Global Financial Crisis. Underwriters charge an additional risk premium for this security type due to reputational concerns during this time, as they have little time to adequately certify an “off-the-shelf” issue (Denis, 1991 and Sherman, 1999). As far as self-selection is concerned, we show that underwriter spreads are truncated by firms’ self-selecting into a particular security type and are therefore not readily comparable across security classes and flotation mechanisms. In particular, we find a positive selection effect for non-shelf equity and a negative selection effect for both shelf and 144a straight bonds. This results in higher underwriter spreads for non-shelf equity, and lower underwriter spreads for shelf and 144a straight bonds than if security choices were random. An analysis of counterfactual underwriter spreads by comparing actual fees with unconditional average underwriting fees that would pertain with random assignment of firms to security types confirms this pattern. This essay contributes to the literature in several ways. First, we present the magnitude and scale effects of direct issuance costs of security offerings for an updated sample of US corporate security offerings between 1999 and 2011 for three security classes, namely equity, convertible debt and straight debt, and for three distribution mechanisms, namely non-shelf, shelf and 144a. We examine the determinants of underwriter spreads and contribute to a growing literature in corporate finance controlling for self-selection. In particular, we show that underwriter spreads are truncated by firms’ self-selecting into particular security types and present counterfactual underwriter spreads.

1.2. Thesis structure The thesis structure follows the format accepted by the Manchester Accounting and Finance Group, Manchester Business School. It allows chapters to be incorporated into a format suitable for submission and publication in peer-reviewed academic journals. Therefore, this thesis is structured around three essays containing original research in chapters 2, 3, and 4. The chapters are self-contained, i.e., each chapter has a separate literature review, answers

14 unique and original questions, and employs a distinct analysis with different datasets. The equations, footnotes, tables, and figures are independent and are numbered from the beginning of each chapter. Page numbers, titles, and subtitles have a sequential order throughout the thesis. The thesis continues as follows. Chapter 2 investigates whether corporate governance influences convertible bond issuance. Chapter 3 examines the signaling content of security offering proceeds. Chapter 4 provides new evidence on the direct costs of raising capital. Chapter 5 concludes. In chapters 2, 3 and 4, I use the third person (we, our) rather than the first person (I, my), as these chapters are in the form of published or working papers co-authored with my supervisors.

15 References

Altinkilic, O. and Hansen, R.S., 2000. Are there economies of scale in underwriting fees? Evidence of rising external financing costs. Review of Financial Studies 13, 191–218. Ambarish, R., Kose, J. and William, J., 1987. Efficient signalling with dividends and investments. Journal of Finance 42, 321–343. Anderson, R.C., Mansi, S.A. and Reeb, D.M., 2004. Board characteristics, accounting report integrity, and the cost of debt. Journal of Accounting and Economics 37, 315–342. Ang, J.S. and Zhang, S., 2006. Underwriting relationships: Information production costs, underwriting fees, and first mover advantage. Review of Quantitative Finance and Accounting 27, 205–229. Agrawal, A. and Knoeber, C.R., 1996. Firm performance and mechanisms to control agency problems between managers and shareholders. Journal of Financial and Quantitative Analysis 31, 377–397. Autore, D.M., Kumar, R. and Shome, D.K., 2008. The revival of shelf-registered corporate equity offerings. Journal of Corporate Finance 14, 32–50. Bayless, M. and Chaplinsky, S., 1991. Expectation of security type and the information content of debt and equity offers. Journal of Financial Intermediation 1, 195–214. Bayless, M. and Chaplinsky, S., 1996. Is there a window of opportunity for seasoned equity issuance? Journal of Finance 51, 253–278. Becker-Blease, J.R. and Irani, A.J., 2008. Do corporate governance attributes affect adverse selection costs? Evidence from seasoned equity offerings. Review of Quantitative Finance and Acounting 30, 281–296. Bethel, J.E. and Krigman, L., 2004. Unallocated shelf registrations. Why doesn’t everybody use it? Equity issuance, managerial choice, and market reaction. Working Paper. Bhojraj, S. and Sengupta P., 2003. Effect of corporate governance on bond ratings and yields: The role of institutional investors and outside directors. Journal of Business 76, 455–475. Billingsley, R.S. and Smith, D.M., 1996. Why do firms issue convertible debt? Financial Management 25, 93–99. Chaplinsky, S. and Hansen, R.S., 1993. Partial Anticipation, the flow of information and the economic impact of corporate debt sales. Review of Financial Studies 6, 709–732. Cremers, K.J.M. and Nair, V.B., 2005. Governance mechanisms and equity prices. Journal of Finance 60, 2859–2894.

16 Danielson, M.G. and Karpoff, J.M., 2006. Do pills poison operating performance? Journal of Corporate Finance 12, 536–559. Denis, D.J., 1991. Shelf registration and the market for . Journal of Business 64, 189–212. Doidge, C., Karolyi, G.A. and Stulz, R.M., 2007. Why do countries matter so much for corporate governance? Journal of Financial Economics 86, 1–39. Dong, M., Dutordoir, M. and Veld, C., 2012. Why do firms issue convertible bonds? Evidence from the field. Working Paper. Dutordoir, M. and Hodrick, L.S., 2012. Self-selection and stock returns around corporate security offering announcements. Working Paper. Dutordoir, M. and Van de Gucht, L., 2009. Why do Western European firms issue convertibles instead of straight debt or equity? European Financial Management 15, 563– 583. Green, R.C., 1984. Investment incentives, debt and warrants. Journal of Financial Economics 13, 115–136. Eckbo, B.E., 1986. Valuation effects of corporate debt offerings. Journal of Financial Economics 15, 119–151. Fang, L.H., 2005. Investment bank reputation and the price and quality of underwriting services. Journal of Finance 60, 2729–2761. Garen, J., 1984. A selectivity bias approach with a continuous choice variable. Econometrica 52, 1199–1218. Graham, J.R. and Harvey, C.R., 2001. The theory and practice of corporate finance: Evidence from the field. Journal of Financial Economics 60, 187–243. Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47, 153–161. Hess, A.C. and Bhagat, S., 1986. Size effect of seasoned stock issues: Empirical evidence. Journal of Business 59, 567–584. Isagawa, N., 2000. Convertible debt: An effective instrument to control managerial opportunism. Review of Financial Economics 9, 15–26. Isagawa, N., 2002. Callable convertible debt under managerial entrenchment. Journal of Corporate Finance 8, 255–270. Jung, K., Kim, Y. and Stulz, R., 1996. Timing, investment opportunities, managerial discretion and the security issue decision. Journal of Financial Economics 42, 159–185. Krasker, W.S., 1986. Stock price movements in response to stock issues under asymmetric information. Journal of Finance 41, 93–106. 17 Kraus, A. and Litzenberger, R.H., 1973. A state-preference model of optimal financial . Journal of Finance 33, 911–922. Lee, I., Lochhead, S., Ritter, J. and Zhao, Q., 1996. The costs of raising capital. Journal of Financial Research 19, 59–74. Lee, G. and Masulis, R.W., 2009. Seasoned equity offering: Quality of accounting information and expected flotation costs. Journal of Financial Economics 92, 443–469. Lewis, C.M., Rogalski, R.J. and Seward, J.K., 1999. Is convertible debt a substitute for straight debt or for common equity? Financial Management 28, 5–27. Li, K. and Prabhala, N.R., 2007. Self-selection models in corporate Finance, in Eckbo, B.E., (ed.): Handbook of corporate finance: Empirical corporate finance, Vol. 1, Handbook of finance series, North-Holland, Elsevier. Livingston, M. and Zhou, L., 2002. The impact of rule 144a debt offerings upon bond yields and underwriter fees. Financial Management 31, 5–27. Marsh, P., 1982. The choice between equity and debt: an empirical study. Journal of Finance 37, 121–144. Masulis, R.W., Wang, C. and Xie, F., 2007. Corporate governance and acquirer returns. Journal of Finance 62, 1851–1889. Mayers, D., 1998. Why firms issue convertible bonds: The matching of financial and real investment options. Journal of Financial Economics 47, 83–102. Miller, M.H. and Rock K., 1985. Dividend policy under asymmetric information. Journal of Finance 40, 1031–1051. Modigliani, F. and Miller, M.H., 1963. Corporate income taxes and the : A correction. American Economic Review 53, 433–443. Myers, S.C. and Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13, 187– 221. Rediker, K.J. and Seth, A., 1995. Boards of directors and substitution effects of alternative governance mechanisms. Strategic Management Journal 16, 85–99. Rutherford, M.A., Buchholtz, A.K. and Brown, J.A., 2007. Examining the relationship between monitoring and incentives in corporate governance. Journal of Management Studies 44, 414–430. Schepker, D.J. and Oh, W.Y., 2012. Complementary or substitutive effects? Corporate governance mechanisms and poison pill repeal. Forthcoming Journal of Management.

18 Sherman, A.E., 1999. Underwriter certification and the effect of shelf registration on due diligence. Financial Management 28, 5–19. Singh, M. and Davidson, W.N.III., 2003. Agency costs, ownership structure and corporate governance mechanisms. Journal of Banking and Finance 27, 793–816. Stein, J.C., 1992. Convertible bonds as backdoor equity financing. Journal of Financial Economics 32, 3–21. Ward, A.J., Brown, J.A. and Rodriguez, D., 2009. Governance bundles, firm performance, and the substitutability and complementarity of governance mechanisms. Corporate Governance 17, 646–660. Westphal, J.D. and Zajac, E.J., 1994. Substance and symbolism in CEOs’ long-term incentive plans. Administrative Science Quarterly 39, 367–390.

19 Chapter 2 Does corporate governance influence convertible bond issuance?

Abstract We examine the influence of corporate governance quality on firms’ choice between convertible debt, straight debt, and equity using a Western European sample of security offerings made between 2000 and 2010. We find that weaker firm-specific and country- specific corporate governance quality increases firms’ likelihood of issuing convertible debt instead of straight debt and common equity. We also find that stockholder reactions to convertible debt announcements are more favorable for firms with weaker corporate governance. Our results suggest that corporate governance quality is a significant security choice determinant, with firms using convertible debt as a substitute for high quality governance mechanisms.

20 2.1. Introduction In recent decades convertible debt has become a major financing source for companies around the world. However, despite a large body of empirical literature (Billingsley and Smith, 1996; Lewis et al., 1999; Graham and Harvey, 2001; Dutordoir and Van de Gucht, 2009; Dong et al., 2012), firms’ motives for issuing convertible debt remain unclear. Our goal is to examine the impact of firms’ corporate governance quality on their likelihood to issue convertible debt instead of straight debt or seasoned equity. The literature on convertible debt issuance motives predicts that convertibles can mitigate agency costs (Green, 1984; Mayers, 1998; Isagawa, 2000) and adverse selection costs resulting from asymmetric information about firm value or risk (Brennan and Kraus, 1987; Brennan and Schwartz, 1988; Stein, 1992). The corporate governance literature, in turn, documents that governance mechanisms can reduce agency and adverse selection costs (Bhojraj and Sengupta, 2003; Anderson et al., 2004; Masulis et al., 2007; Becker-Blease and Irani, 2008). We combine these two strands of literature to develop predictions on the relation between corporate governance quality and firms’ likelihood to issue convertible debt. The Substitution hypothesis predicts that, since convertibles and high quality governance mechanisms are both able to reduce agency and adverse selection costs, firms with lower quality governance in place are more likely to issue convertible bonds instead of straight bonds or equity. In contrast, the Complementarity hypothesis assumes that firms with high quality governance are more likely to adopt financing strategies that further improve shareholder value. Given convertibles’ potential to reduce agency and adverse selection costs, this yields the prediction that well governed firms are more inclined to issue convertibles instead of standard non- hybrid financing instruments. While the Substitution and Complementarity hypotheses rely on the assumption that managers act in shareholders’ interests, the Entrenchment hypothesis predicts that entrenched managers use convertibles to further insulate themselves from market discipline. We derive this hypothesis from Isagawa’s (2002) rationale for convertible bond issuance. Isagawa (2002) argues that, like straight debt, convertible debt reduces the probability of a hostile takeover. However, unlike with straight debt, managers can avoid bankruptcy by forcing conversion of the bonds into equity by calling them. Convertibles may therefore help entrenched managers preserve their control benefits, even if this is not in shareholders’ interests. Since managerial entrenchment is likely to be higher in firms with weaker corporate governance (Berger et al., 1997; Bebchuk et al., 2009), the Entrenchment hypothesis predicts a higher likelihood of convertible bond issuance by firms with weaker corporate governance.

21 Ultimately, therefore, the impact of corporate governance on a company’s likelihood to choose convertible debt over standard financing instruments is an empirical question. To examine this question, we use a pan-Western European dataset of 176 convertible issues, 350 straight debt issues, and 141 seasoned equity issues made between 2000 and 2010. The European convertible debt market has experienced dramatic growth in recent decades (Bancel and Mittoo, 2004; Dutordoir and van de Gucht, 2009) and the diversity of European regulatory environments creates a variety of corporate governance systems (Shleifer and Vishny, 1997; Aggarwal et al., 2009), enabling us to consider both internal (company- specific) and external (country-specific) corporate governance features (Doidge et al., 2007). We hand collect data for seven internal and four external corporate governance characteristics. We analyze firms’ security choices with multinomial logistic regressions including corporate governance measures and a range of firm-specific and macroeconomic control variables. Our focus on incremental security issues allows us to conduct the analysis with independent variables measured prior to security offering announcement dates, which has the advantage of mitigating endogeneity problems inherent to many corporate governance studies. Our main results are as follows. Companies with weaker internal and external corporate governance quality are more likely to issue convertible debt than straight debt or seasoned equity. Among corporate governance quality proxies, the impact of the presence of large shareholders is particularly strong. Companies with large shareholders are significantly less likely to issue convertible debt than straight debt and seasoned equity. We also find a significant negative impact of a number of country-specific proxies for corporate governance quality on firms’ likelihood of issuing convertible bonds. The finding that firms with weaker corporate governance are more likely to issue convertible bonds instead of straight debt or equity is consistent with both the Substitution and Entrenchment hypotheses. To discriminate between these two hypotheses, we analyze the impact of corporate governance on stock returns around convertible bond announcements. Consistent with the Substitution hypothesis, we find that stockholders perceive convertibles as more valuable for firms with weaker corporate governance. Overall, our results indicate that corporate governance characteristics have a statistically and economically significant impact on firms’ security choices, and security choice models should therefore incorporate these characteristics. Our paper contributes to the literature on securities issuance by providing new insights into the so far unresolved question of why firms choose convertible bonds instead of straight

22 bonds or equity. 1 Our key finding that firms with lower quality corporate governance mechanisms in place are more likely to issue convertibles instead of straight bonds or equity is consistent with theories predicting a role for convertibles in reducing the agency and adverse selection costs associated with non-hybrid financing instruments. In addition, to our knowledge, we are the first to empirically test Isagawa’s (2002) rationale for convertible debt issuance. Our findings also complement the study of Lee at al. (2009) who examine the relationship of external corporate governance and firms’ ownership structure to convertible bond design, but do not consider security choice. Our paper also contributes to the corporate governance literature. While early studies examine individual corporate governance mechanisms in isolation, a more recent stream of articles documents that firms tend to use governance mechanisms as substitutes (e.g., Westphal and Zajac, 1994; Rediker and Seth, 1995; Agrawal and Knoeber, 1996; Singh and Davidson, 2003; Rutherford et al., 2007) or as complements (Danielson and Karpoff, 1998; Cremers and Nair, 2005; Schepker and Oh, 2013). A common feature of these studies is that they focus on traditional corporate governance measures such as board and ownership structure. Our findings suggest that researchers should consider convertible bond issuance as part of a bundle of corporate governance measures that serve to protect shareholder interests (Ward et al., 2009). The remainder of the paper continues as follows. The next section reviews the relevant literature and develops the hypotheses. Section 3 describes the dataset and discusses the research methodology. Section 4 reports and discusses the empirical results. Section 5 concludes the paper.

2.2. Literature review and hypothesis development Our hypotheses draw from three strands of literature: studies of the relation between convertible bond issuance and firms’ financing costs, studies of the relation between corporate governance and firms’ financing costs, and studies of the interdependency between corporate governance mechanisms. In this section, we first briefly review these relevant studies, and then formulate our testable predictions.

2.2.1. Convertible bond issuance as a tool to reduce agency and adverse selection costs Theories of convertible bond issuance broadly subdivide into two groups. The first,

1 Dutordoir et al. (2012) provide an extensive overview of empirical evidence on firms’ motives to issue convertibles. Their overall conclusion is that this evidence is mixed and inconclusive.

23 largest group considers convertible debt as a solution to agency conflicts. Jensen (1986), Stulz (1990), and Hart and Moore (1995) argue that straight bonds can mitigate managerial overinvestment by reducing free cash flows and imposing the threat of bankruptcy. Consistent with these rationales, Berger et al. (1997), Morellec (2004), and Harvey et al. (2004) obtain empirical evidence suggesting that straight debt acts to reduce management’s empire building tendencies. However, as Jensen et al. (1992) and Isagawa (2000) point out, straight debt reduces managerial overinvestment at the expense of creating new agency problems between bondholders and shareholders. This is where convertible debt comes in. Green (1984) shows that convertible debt can mitigate levered firms’ shareholder incentives to engage in overly risky projects (the asset substitution problem of Jensen and Meckling, 1976). The underlying intuition is that shareholders the profits from high risk projects with convertible bondholders, which reduces their incentives to invest in such projects in the first place. Mayers (1998) argues that convertible debt is more suitable than straight debt for financing real investment options. Convertible debt creates less potential for shareholders to engage in harmful overinvestment than do long-term bonds, since the firm can commit to redeem the convertible debt if future investment options have no value. Moreover, the firm saves on the issuing costs of sequential short-term debt offerings because conversion of the convertible bonds creates equity that the firm can use to finance profitable future investment options. Isagawa (2000) presents a formal model showing that convertible debt can tackle both overinvestment and underinvestment incentives (i.e., the debt overhang problem of Myers, 1977). The superiority of convertible debt over straight debt lies in the conversion design, whereby conversion occurs when outstanding debt causes underinvestment but does not occur when the presence of straight debt prevents managers from overinvesting. A second group of theories predict that convertibles can alleviate adverse selection costs resulting from asymmetric information. Brennan and Kraus (1987) and Brennan and Schwartz (1988) argue that, because changes in a firm’s risk have opposite effects on the value of a convertible’s bond and option components, convertible bond value is relatively insensitive to firm risk. As such, convertible debt is a suitable alternative to straight debt when investors have difficulty assessing the risk of a company’s current and future assets, complicating the determination of an accurate bond yield. Stein (1992) develops a model in which firms use convertibles to mitigate the adverse selection costs resulting from equity financing described by Myers and Majluf (1984), while at the same time avoiding financial distress costs associated with straight bond financing. Existing quantitative and qualitative evidence on the validity of these different convertible bond issuance rationales is mixed and inconclusive

24 (Dutordoir et al., 2012). Isagawa (2002) relaxes the above models’ implicit assumption that convertible bond issuers act in shareholders’ interests. His model is based on Zwiebel (1996), who argues that an entrenched manager may prefer straight debt over equity in order to avoid a hostile takeover and the associated loss in control benefits. Straight debt, however, may result in the loss of the manager’s control benefits when the firm runs into financial distress. Isagawa (2002) shows that an entrenched manager can avoid both a hostile takeover and bankruptcy by issuing callable convertible debt. As such, callable convertible debt issuance decreases the value of the firm, by allowing managers to make conversion-forcing bond calls when liquidation of the firm would be optimal for shareholders. To our knowledge, the literature has not yet empirically tested the Isagawa rationale.

2.2.2. Corporate governance as a tool to reduce agency and adverse selection costs Corporate governance deals with the ways in which suppliers of finance to corporations assure themselves of a return on their investment (Shleifer and Vishny, 1997). Many studies suggest that corporate governance mechanisms induce managers to take investment decisions that are in line with shareholders’ interests, thereby reducing agency problems.2 For example, Masulis et al. (2007) find that firms with higher quality governance mechanisms are less likely to engage in value-decreasing acquisitions, and Lin and Chang (2012) document that well governed firms are more likely to engage in successful product introductions. A limited number of articles examine the relation between corporate governance quality and adverse selection costs resulting from information asymmetry. Leland and Pyle (1977) show that in a separating equilibrium initial owners signal favorable private information about firms’ quality by retaining a larger fraction of equity. As a result debt holders are more willing to lend larger amounts of debt. Bhojraj and Sengupta (2003) hypothesize that higher quality governance mechanisms reduce information asymmetries between firms and their lenders, leading to lower costs of debt financing. In line with this prediction, they find that firms with greater institutional ownership and stronger outside control of the board enjoy lower bond yields and higher ratings on their new bond issues. Anderson et al. (2004) also find that firms’ cost of debt is inversely related to proxies for corporate governance quality.

2 We acknowledge that besides the below mentioned agency conflicts between managers, shareholders, and bondholders, there are also conflicts between largest and minority shareholders prevalent in European firms, potentially affecting financial decisions (e.g., Shleifer and Vishny, 1997; Claessen et al., 2002; La Porta et al., 2002; Faccio and Lang, 2002; and Faccio et al., 2011). We do not consider these conflicts in the course of this study, as there is no theoretical rationale for convertible bond issuance to mitigate these particular agency conflicts.

25 Becker-Blease and Irani (2008) hypothesize that high quality corporate governance mitigates the adverse selection problem described by Myers and Majluf (1984), since shareholders are less worried about managerial opportunism in such firms. Consistent with this prediction, they find that proxies for corporate governance quality attenuate shareholders’ negative reaction to seasoned equity offering announcements.

2.2.3. The interdependency between governance mechanisms A third stream of literature relevant for our study focuses on the interdependency between corporate governance mechanisms. Early research on the determinants and effects of various governance mechanisms typically assumes that these mechanisms operate independently (Danielson and Karpoff, 1998). However, as argued by Gompers et al. (2003) and Chen et al. (2007), among others, separate investigation of individual governance attributes ignores the possibility that these attributes serve as substitutes or complements. The substitution perspective holds that control mechanisms offer alternative ways to incentivize managers and any one mechanism can substitute for another, which is especially relevant when implementing governance measures is costly to the firm (Agrawal and Knoeber, 1996). The literature provides ample empirical evidence for this perspective. For example, Westphal and Zajac (1994) find that the use of long-term CEO incentive plans is negatively related to monitoring processes in place. Further studies documenting substitution effects among governance mechanisms include Rediker and Seth (1995), Agrawal and Knoeber (1996), Singh and Davidson (2003), and Rutherford et al. (2007). Several other studies discuss substitutive effects between corporate governance and firms’ financial structure. Novaes and Zingales (1995) and Berger et al. (1997) show that firms with weak managerial incentives have lower leverage. Extending this finding, Datta et al. (2005) show an inverse relationship between manager–shareholder alignment through stock ownership and the amount of short term debt in firms’ . With respect to external corporate governance, Demirguc-Kunt and Maksimovic (1999) find that stronger creditor rights associate with higher firm leverage. Finally, linking external corporate governance to security design Lee et al. (2009) show that firms issue more equity-like convertibles in countries with strong shareholder rights. Conversely, firms in countries with strong creditor protection tend to issue more debt-like convertible bonds. This relationship is less (more) pronounced in countries with higher managerial agency costs due to a stronger separation of cash flow and control rights for equity (debt). Conversely, the complementarity perspective holds that any one governance mechanism

26 may be insufficient to reduce firms’ agency and adverse selection costs. Principals will seek to implement as many governance mechanisms as possible, inducing synergistic effects, in order to reduce the potential for agent opportunism (Schepker and Oh, 2013). Empirical evidence on this perspective is scarce compared with that for the substitution perspective. Danielson and Karpoff (1998) find that certain governance provisions tend to appear together because management views their effects as complementary. Cremers and Nair (2005) find that internal and external governance mechanisms are complements in being associated with long- term abnormal returns. Schepker and Oh (2013) document complementarity of governance mechanisms in the context of poison pill repeals.

2.2.4. Hypotheses Together, the above streams of literature suggest the following three testable hypotheses. The Substitution hypothesis predicts that, since convertible bonds and corporate governance mechanisms are alternative ways for firms to achieve lower agency and adverse selection costs, firms use convertible bonds as substitutes for high quality corporate governance. If this hypothesis holds, corporate governance quality measures should have a negative impact on firms’ propensity to issue convertibles instead of straight debt or equity. Moreover, if the stock market recognizes the potential of convertible debt to mitigate the higher agency and adverse selection costs associated with poor corporate governance, stock returns around convertible debt announcements should be more favorable for firms with weaker corporate governance.3 The substitutive effects of convertibles and traditional governance mechanisms such as board and ownership structure might result from the documented static nature of these internal governance mechanisms (Gompers et al., 2003; Bebchuk et al., 2009). It may be difficult for firms to change these mechanisms at short notice. Moreover, external (country- wide) governance mechanisms are outside the firms’ control. By contrast, firms can structure and issue convertible bonds very quickly (often overnight), and they may therefore provide a flexible and fast way for firms to reduce their agency and adverse selection costs. The complementarity viewpoint on governance mechanisms inspires our second hypothesis. This viewpoint implies that firms that already have high quality governance mechanisms in place are more inclined to look for further means of reducing agency and adverse selection costs, and issue convertibles as part of a pattern of strong corporate

3 The general finding in the literature is that convertible debt announcements are associated with negative abnormal stock returns that are intermediate in magnitude between those associated with seasoned equity offerings and straight debt offerings (e.g., Dann and Mikkelson, 1984; Mikkelson and Partch, 1986; Lewis et al., 1999). This result is consistent with the adverse selection model of Myers and Majluf (1984).

27 governance. If this hypothesis holds, corporate governance quality measures should have a positive impact on firms’ propensity to issue convertibles instead of straight debt or equity. Unlike the two other hypotheses, the Complementarity hypothesis does not yield an explicit prediction on the impact of corporate governance quality on the stock price reaction to convertible bond announcements. A third hypothesis follows Isagawa’s (2002) argument that entrenched managers in companies with weak corporate governance issue convertibles to further secure their positions. If this Entrenchment hypothesis holds, corporate governance quality should have a negative impact on firms’ propensity to issue convertibles instead of straight bonds. Moreover, if the market realizes the potential of convertibles to further entrench managers, stock price reactions to convertible debt announcements should be more negative for firms with weaker corporate governance in place. Table 1 summarizes the three hypotheses by showing the sign of the expected effect of corporate governance quality on firms’ propensity to issue convertible debt and on stock returns around convertible bond announcements. It is important to note that our hypotheses regarding short-run announcement returns assume that stockholders are aware of management’s motives for issuing convertibles (to protect shareholder interests versus to preserve management’s control benefits) at the announcement of these offerings. Moreover, we assume stock markets are informationally efficient (Fama, 1970). That is, stockholders unbiasedly assess the implications of the announced offering and of the associated managerial motives for future cash flows, and the firm’s stock price instantaneously reflects this information. Other event studies of the impact of corporate governance measures on stock returns rely on similar assumptions (e.g., Sundaramurthy et al., 1997; Masulis et al., 2007; Lin and Chang, 2012). We do not examine long-term returns following convertible bond issuance due to the methodological problems associated with such an analysis.4

2.3. Data 2.3.1. Security issues We test the hypotheses on a sample of security offerings by firms domiciled in 13 Western European countries, namely Austria, Belgium, Denmark, Finland, France, Germany, Italy, the Netherlands, Norway, Spain, Sweden, Switzerland, and the United Kingdom (UK), between January 2000 and September 2010. The geographical scope of our study follows

4 Kothari and Warner (1997) discuss the econometric problems associated with long-horizon event studies.

28 Dutordoir and Van de Gucht (2007), who identify these countries as the main convertible debt markets in Western Europe. We download all convertible debt, straight debt, and seasoned equity issues from Thomson ONE Banker. In line with the literature, the sample excludes financial and utility companies and private issues. We aggregate issues offered in several tranches. Applying these criteria gives 593 convertible debt issues, 654 straight debt issues, and 635 seasoned equity issues. From this sample, we retain all issues that meet the following criteria: (1) company accounts data are available for the fiscal year-end before the issue date; (2) stock price data are available for the year preceding the issue date; and (3) corporate governance data are available for the fiscal year-end before the issue date. Applying these criteria yields a final sample of 176 convertible debt issued by 129 different firms, 350 straight debt issued by 154 different firms, and 141 seasoned equity issues issued by 124 different firms. Table 2 gives the number of convertible debt, straight debt, and equity issues by year and country. Table 2, panel A shows that the popularity of convertible debt issues varies over time. After eight years of roughly constant convertible debt issues, with an average of around 14 per year, the number drops to four in 2008, the start of the Global Financial Crisis. After 2008 the convertible bond market rebounds with 42 issues in 2009 and 17 in the first nine months of 2010. French firms issue over 40% of the convertible bonds. Prior studies also document that French issuers dominate the European convertible bond market (Ammann et al., 2003; Bancel and Mittoo, 2004; Dutordoir and Van de Gucht, 2007).5 Besides Germany (22 issues), the UK (18 issues), and the Netherlands (15 issues), the remaining countries each have less than ten convertible bond issues. Panel B shows that the number of straight debt issues is considerably more uniform over the sample period. Issuance is highest in the first two years. After this, the number of issues fluctuates around 30, but drops substantially in 2008. France also accounts for the largest number of straight debt offerings, comprising 30% of the issues. As with convertible bonds, the next biggest offering nations are Germany (13%), the UK (10%), and the Netherlands (8%), with the remaining countries each contributing less than 10% of the sample. Panel C shows the distribution of seasoned equity issues. There are no equity issues in the first four years of the sample and no more than one offering per year between 2007 and 2009. These years coincide with equity market downturns caused by the stock market crash at the beginning of the decade and by the Global Financial Crisis. Equity issuance is highest in 2005

5 Robustness checks reported below show that the dominance of French convertibles does not drive our results.

29 and 2006.6 The UK, arguably the most developed equity market in Europe, accounts for almost 40% of all seasoned equity issues. The next biggest issuing countries are Germany (19%) and France (8%). Other countries each contribute less than 7% of overall equity issues.

2.3.2. Proxies for corporate governance quality Since corporate governance mechanisms within a firm may work simultaneously and influence each other (Rediker and Seth, 1995), we measure corporate governance quality using a wide range of proxy variables. To measure firm-specific (internal) corporate governance quality, we use the proxy variables below. The first three capture the quality of monitoring of the firm’s management, while the last four capture the degree of alignment between managers’ and shareholders’ interests. Table 3 provides precise definitions of all the variables in the study and their data source. Ownership concentration: Blockholders are assumed to actively monitor management and thereby enhance corporate governance quality because they can more easily bear the costs of collecting information on management’s behavior than can smaller shareholders (Stiglitz, 1985). Unreported analyses of the ownership characteristics of our security samples indicate that for convertible debt and equity, the average equity ownership percentage falls below 5% for the fourth largest shareholder. For straight debt, this happens for the third largest shareholder. 7 An equity ownership of 5% is widely considered to be the threshold for considering a shareholder as “large” (i.e., a blockholder) (e.g., Agrawal and Knoeber, 1996; Cremers and Nair, 2005; Cronqvist and Fahlenbrach, 2009). Our first ownership concentration measure, labeled ownership concentration (1), therefore measures the percentage of shares held by the company’s three largest shareholders. Several other studies use ownership measures focusing on companies’ three largest shareholders, e.g., Mehran (1992), La Porta et al. (1998), and Rossi and Volpin (2004).8 We construct an alternative ownership concentration measure (ownership concentration (2)) equal to the percentage of shares held by the company’s largest shareholder. Rediker and Seth (1995) and Voulgaris et al. (2010), among others, use an equivalent measure. Outside directors: Outside directors are members of the board who are not and never

6 Robustness checks reported below show that the clustering of offerings in 2005 and 2006 does not drive our results. 7 Results of untabulated analyses reported throughout the paper are available from the corresponding author. 8 Mehran (1992) measures in his study on executive compensation ownership concentration as the percentage of equity held by the three largest shareholders in one of the following groups, CEOs, top executives, other directors. La Porta et al. (1998) and Rossi and Volpin (2004) construct a country level ownership concentration measure which is the average equity stake owned by the three largest shareholders in the ten largest nonfinancial domestic firms.

30 have been company employees. We assume they are better monitors of management, since their human capital is not tied to the firm (Rosenstein and Wyatt, 1990; Agrawal and Knoeber, 1996). We use the percentage of independent board members as our measure for this variable. Board size: Board size is the total number of directors on the company’s board. Several studies show that larger boards are inefficient monitors of management due to free-rider problems, resulting in lower firm performance (Yermack, 1996; Conyon and Peck, 1998; Core et al., 1999). CEO tenure: CEO tenure is the number of years the CEO has held that position in the firm. Berger et al. (1997) argue that CEOs with longer tenure in a firm may be more entrenched, leading to lower corporate governance quality. Moreover, Murphy (1986) finds that CEO tenure is negatively correlated with managerial pay–performance sensitivity. The reason is that there is greater information asymmetry about CEO ability at the beginning of a contract, leading companies to rely more on performance-based compensation.9 CEO age: Gibbons and Murphy (1992) and Palia (2001) find that CEOs closer to retirement receive a significantly larger portion of stock options and performance-based compensation in their salary contracts. Career concerns incentivize the performance of younger managers, reducing the need for such contracts. This means that CEO age can proxy for performance-based compensation. Managers’ interests coincide more closely with those of outside shareholders the larger their equity stakes (Jensen and Meckling, 1976). Everything else equal, therefore, an older CEO should translate into a higher manager–shareholder alignment and better corporate governance quality. A growing literature, however, suggests that older CEOs tend to be more risk averse, which may translate into larger alignment with shareholder interests (e.g., Prendergast and Stole, 1996; Bertrand and Mullainathan, 2003; Li et al., 2011; Yim, 2013). The direction of the relation of CEO age with corporate governance quality is therefore unclear. Founder CEO dummy: Founding CEOs typically own substantially larger equity stakes in their companies than non-founding CEOs, resulting in greater alignment with shareholder interests (Adams et al., 2010).10 Outside CEO dummy: CEOs hired from outside the firm have less time to accumulate large stockholdings in the company, making their interests less aligned with those of shareholders.

9 We do not have data on executive compensation packages for European companies. 10 We do not have direct data on managerial stock ownership for European companies.

31 We measure all of these firm-specific governance characteristics at the fiscal year-end before the convertible bond’s issue date. We collect blockholder information from Thomson ONE Banker and Orbis. The other six internal characteristics are mostly hand collected from BoardEx. As BoardEx coverage is not sufficient for all companies and years, we supplement BoardEx data with data from company reports, capital market filings, and newspaper articles. The legal system, culture, and institutional framework of the country of domicile of security issuers may also affect corporate governance quality (Cremers and Nair, 2005; Doidge et al., 2007; Aggarwal et al., 2009). As external measures of corporate governance, we consider shareholder and creditor rights using the anti-director rights and creditor rights indices of La Porta et al. (1998). 11 Both indices capture the legal framework for good corporate governance with a higher score on the index indicating better external governance. We follow Korkeamaki (2005) and use dummy variables taking a value of one for index scores of four or higher for the anti-directors rights index, and three or higher for the creditor rights index. Hostile are another channel of external monitoring. If a company performs poorly due to managerial opportunism or inefficiency, it is more likely to become a hostile takeover target with its managers being either replaced or better monitored (Jensen, 1986). One condition for takeovers to be a successful control mechanism is a liquid capital market. We therefore include proxies for a country’s stock market and credit market development, in line with Korkeamaki (2005). Stock market development is domestic stock market capitalization deflated by GDP. Credit market development is total private domestic claims deflated by GDP. Both variables are measured at the fiscal year-end before the issue and obtained from EuroMonitor. Table 4 reports descriptive statistics for the corporate governance characteristics of the convertible debt, straight debt, and seasoned equity samples. Panel D tests pairwise differences in means between the convertible debt sample and the other two security samples.

11 We consider both the shareholder and creditor rights index, as several studies point out the importance of considering the interaction between shareholder-manager-creditor conflicts rather than solely focusing on shareholder-manager conflicts. Chava et al. (2010) show that managerial entrenchment increases the likelihood of using bond covenants, as empire building reduces firm profitability and increases default risk. Bondholders thus benefit from greater manager-shareholder alignment. Moreover, Johnson et al. (2000) and Djankov et al. (2008) document that strong shareholder rights prevent managers from expropriating assets from all stakeholders, including bondholders. In the same spirit, Miller and Reisel (2012) show that bondholders do not demand more covenant protection when shareholder rights are strong, in countries outside the U.S. On the other hand, stronger creditor rights allow for a larger debt capacity resolving managerial agency problems (Jensen, 1986; Hart and Moore, 1995), which benefits shareholders. So, overall, these studies show that an increase in manager-creditor alignment does not aggravate manager-shareholder conflicts and vice versa. In conclusion, good corporate governance can align shareholders, creditors and managers towards the long-term goal of firm value maximization.

32 The two ownership concentration measures show that, on average, European security issuers have concentrated ownership structures, with average shareholdings of the largest shareholder (as captured by ownership concentration (2)) of the order of 20%, and average shareholdings of the three largest shareholders (as captured by ownership concentration (1)) of the order of 30%. But convertible bond issuers, on average, have significantly less concentrated ownership structures than straight debt or equity issuers. Assuming that blockholders are more effective at monitoring management, this result is consistent with companies with weaker corporate governance relying more on convertible debt. We also find that convertible debt issuers have CEOs with significantly longer tenure than straight debt and equity issuers. Furthermore, convertible debt issuers have a significantly higher percentage of outside CEOs than straight debt issuers. Finally, a significantly lower percentage of convertible debt issuers than equity issuers have founder CEOs. On the whole, these findings support the Substitution and Entrenchment hypotheses, predicting a negative relation between corporate governance quality and convertible bond issuance. However, we also obtain some support for the Complementarity hypothesis, as convertible issuers have a higher percentage of outside directors than straight debt and equity issuers. The finding of significantly older CEOs for convertible than for equity issuers may also be in line with this hypothesis. The finding that convertible issuers have significantly smaller boards than straight debt issuers, but significantly larger boards than seasoned equity issuers, provides mixed evidence. For the external channels of corporate governance, convertible issuers, on average, come from countries with lower shareholder and creditor rights protection and less developed capital markets. Most of the univariate findings are thus consistent with the Substitution and Entrenchment hypotheses, predicting a negative impact of corporate governance quality on firms’ likelihood to issue convertible debt.

2.3.3. Firm-specific control variables Besides corporate governance characteristics, we control for firm-specific characteristics affecting the choice between convertibles, straight debt, and equity. In line with Dutordoir and Van de Gucht (2007), we group the firm-specific characteristics into proxies for equity-related adverse selection costs, proxies for debt-related financing costs, and proxies for general financing costs. Table 3 provides a detailed description of the measurement and data sources for these variables. All control variables are measured prior to the security issue date.

33 2.3.3.1. Proxies for equity-related financing costs According to the adverse selection framework of Myers and Majluf (1984), the announcement of equity-like financing may signal that the firm is overvalued, leading to a negative stock price reaction to the offering announcement. We expect a negative impact of proxies for the magnitude of this equity-related adverse selection effect on firms’ likelihood of choosing more equity-like securities. Lucas and McDonald (1990) argue that the equity- related adverse selection problem is likely to be smaller for firms with a large stock price run- up before the issue, since stockholders may interpret the run-up as a signal of good investment projects. However, the pre-offering stock price run-up may also proxy for firm overvaluation and, as such, be associated with higher equity-related financing costs. As Myers and Majluf (1984) argue, equity-related adverse selection costs may be higher for firms with more financial slack available. Such firms could have used internal funds instead and are therefore more likely to be perceived as overvalued. Following Krasker (1986), equity-related adverse selection costs should be higher for larger issues. As equity-related adverse selection cost proxies, we therefore include the pre-offering stock price run-up, financial slack over total assets, and the ratio of offering proceeds to the market value of common equity.

2.3.3.2. Proxies for debt-related financing costs Our analysis also includes a number of debt-related financing cost proxies suggested by capital structure theories. We include the ratio of income taxes to total assets as an inverse debt-related cost measure. This ratio captures the extent to which firms can exploit the tax deductibility of debt interest payments (Modigliani and Miller, 1963). We include return on assets (ROA), as high profitability before the issue makes it easier for a company to pay interest on debt securities (Lewis et al., 1999). We expect a negative impact of these inverse debt-related financing cost proxies on firms’ likelihood to issue more equity-like securities. To measure firms’ financial distress costs, we include short- and long-term debt to total assets and stock return volatility. We include short-term debt in addition to long-term debt since, due to its maturity, high short-term debt may be a better indicator of financial distress than high long-term debt (Diamond, 1991). Leverage and stock return volatility can also proxy for asset substitution costs (Green, 1984) and stock return volatility can capture risk uncertainty (Brennan and Schwartz, 1988). We predict a positive impact of these debt-related financing cost proxies on firms’ likelihood to choose relatively more equity-like securities.

34 2.3.3.3. Proxies for general financing costs In addition to specific equity- and debt-related financing cost measures, we include a number of widely used control variables that can capture a range of financing costs. Since these variables can proxy for both debt-related and equity-related financing costs, we do not have clear predictions on their impact on firms’ propensities to choose more equity-like securities. In particular, we control for a company’s total assets, market-to-book-ratio, and sales growth. Total assets may proxy for the magnitude of asymmetric information and financial distress costs (Lewis et al., 1999). The market-to-book ratio may proxy for the availability of profitable growth opportunities, resulting in lower external financing costs. On the other hand, high growth firms tend to suffer from higher asymmetric information related to their value and risk and from debt-related underinvestment problems (Myers, 1977), increasing their external financing costs. A similar ambiguous interpretation holds for sales growth, an alternative growth measure in our analysis.

2.3.4. Macroeconomic control variables As argued by Choe et al. (1993) and Bayless and Chaplinsky (1996), financing costs vary not only at the firm level but also at a macroeconomic level. We therefore include several widely used macroeconomic financing cost proxies in our analysis (e.g., Lewis et al., 1999; Dutordoir and Van de Gucht, 2009; Frank and Goyal, 2009). We expect a positive (negative) impact of macroeconomic debt-related (equity-related) financing cost proxies on firms’ propensity to issue more equity-like securities. As an inverse proxy for the economy-wide level of equity-related financing costs, we include the stock market run-up. Stock market volatility and the five-year German Treasury bond yield act as proxies for the economy-wide level of debt-related financing costs. A six- month European leading indicator acts as an inverse proxy for external financing costs in general, as it measures economy-wide growth opportunities. Table 3 provides a more detailed description of the measurement and data sources for these macroeconomic control variables. Stock market run-up and volatility are measured over a (−200, −20) window before issuance and the Treasury bond yield and leading indicator are calculated over the quarter prior to the issue month. Table 5 reports descriptive statistics for the firm-specific and macroeconomic control variables for the three security samples. Panel D gives t-statistics for pairwise differences in means between the convertible sample and the other two security samples. For the equity-related financing cost proxies, convertible debt issuers have a significantly

35 larger stock price run-up than straight debt issuers, and a significantly smaller (larger) ratio of financial slack to total assets than straight debt (equity) issuers. These findings are in line with our predictions. However, we also find a significantly larger offering proceeds ratio for convertible debt issuers than for straight debt issuers, which is not in line with our expectations. For the debt-related financing cost measures, convertible debt issuers have significantly smaller income tax to total assets than straight debt issuers, significantly smaller (higher) ROAs than straight debt (equity) issuers, and significantly larger (smaller) stock return volatility than straight debt (equity) issuers. These results are in line with our predictions. But we also find that convertible bond issuers have significantly higher short- and long-term leverage than equity issuers, which is unexpected. For the general financing cost measures, convertible bond issuers have significantly smaller (larger) total assets than straight debt (equity) issuers, significantly lower market-to-book ratios than equity issuers, and significantly higher sales growth than straight debt issuers. As expected, compared to equity, convertible debt issues are less likely following stock market run-ups and more likely when Treasury bond yields are high. Convertible bond issues are associated with higher values of stock market volatility and the 6-month leading indicator than straight debt and equity issues. Not in line with theory is a lower likelihood of convertible bond issues compared to straight debt when Treasury bond yields are high. In conclusion, the control variable descriptive statistics are largely in line with our predictions and confirm that convertible bond issuers tend to have high costs of both straight debt and equity financing. To check for multicollinearity problems, we analyze pairwise Pearson correlations between the corporate governance characteristics and between the corporate governance characteristics and the control variables. The results of this untabulated analysis indicate that correlations do not exceed 0.42 and are below 0.3 for the large majority of variables.

2.4. Empirical results In a first step of our security choice analysis, we test whether the independence of irrelevant alternatives (IIA) assumption holds among firms’ choices between convertible debt, straight debt, and seasoned equity. The IIA property means that the log odds ratio of any two alternatives in the security choice menu does not depend on the availability of the third. If this assumption holds, we can estimate the choice between the three financing options with a multinomial logit model (Train, 2009). If it does not hold, we have to resort to a nested model specification. A Hausman test indicates that the IIA assumption is not violated in our dataset,

36 leading us to opt for a multinomial logit analysis.12 The output of the multinomial logit model consists of two pairwise regressions: one that models firms’ likelihood to choose straight debt over convertibles (set as the base outcome) and one that models firms’ likelihood to choose seasoned equity over convertibles.13 Tables 6 and 7 report the results of these two pairwise regressions. For ease of interpretation, we reverse the coefficient signs in the regressions so that the coefficients represent firms’ likelihood to choose convertibles (the base outcome) instead of straight debt (Table 6) or seasoned equity (Table 7). We stress that these pairwise regression results are the outcome of a multinomial security choice model that simultaneously incorporates all three security types (667 offerings in total), since firms are likely to consider the three financing options simultaneously in reality. To further determine the validity of our hypotheses, we examine the effects of detailed ownership characteristics on security choices (Table 8) and the effects of corporate governance on convertible debt announcement returns (Table 9). The remainder of this section discusses the results of these analyses in more detail. We also outline a number of robustness tests.

2.4.1. The choice between convertible and straight debt Table 6 reports the results of the multinomial logit analysis of the determinants of firms’ choice between convertible debt and straight debt. We take the natural logarithm of ownership concentration (1), board size, CEO tenure, and CEO age.14 Regression (1) reports results for the security choice model using only control variables on the right-hand side. The results are largely consistent with our predictions and with the univariate results. In particular, convertible debt issuers have a significantly larger stock price run-up, long-term debt ratio, stock return volatility, sales growth and stock market volatility, and significantly smaller issue proceeds and total assets than straight debt issuers. Regression (2) extends regression (1) by adding the internal and external corporate governance characteristics. We find a significant negative impact of ownership concentration (1) on the likelihood of issuing convertible debt over straight debt. This result supports the

12 Erel et al. (2012) use a similar approach in their security choice analysis. 13 We use robust standard errors in all regression analyses. Following the security choice literature, we do not use standard errors clustered at firm level as it is common in studies on using panel data (Petersen, 2009). Security issuance is rare for firms, so that most firms only appear once in the sample. Confirming this argument, our results remain unchanged using standard errors clustered at firm level. 14 We use this transformation in all subsequent regression models.

37 Substitution and Entrenchment hypotheses. 15 Also in line with these hypotheses is a significant positive impact of the outside CEO dummy variable on the likelihood to issue convertible debt. Findings on the impact of external corporate governance characteristics also support these hypotheses. In particular, the creditor rights dummy variable and stock market development both have a significant negative impact on firms’ likelihood to issue convertible debt. Credit market development has a weakly significant positive impact, in line with the Complementarity hypothesis. A χ 2 -test of the joint significance of the corporate governance variables gives a value of 75.54, significant at less than 1%. The remaining regressions provide a number of robustness tests. Regression (3) substitutes ownership concentration (2) (capturing the shareholding of the largest shareholder) for ownership concentration (1) (capturing the shareholding of the three largest shareholders). Regression (4) includes a dummy variable taking the value one for offerings made during 2005 and 2006 to control for the clustering of security issues in these years. The findings remain the same for both regressions. Regression (5) adds a dummy variable taking the value one for French security issues, thereby controlling for the possibility that the dominance of French convertible debt issues drives the results. The regression shows a significant positive impact of this dummy variable, whilst the findings on the impact of corporate governance characteristics are largely consistent with those in previous regressions. 16 The only difference is that the creditor rights variable is now insignificant. The impact of corporate governance characteristics on firms’ trade-off between convertible debt and straight debt is economically significant. An untabulated analysis of the marginal effects of the significant corporate governance determinants (evaluated at the variable means) shows that a 1% increase in the ownership concentration (2) measure reduces the likelihood of a convertible debt issue by 0.4%. The presence of an outside CEO, in turn, increases the likelihood of a convertible debt issue by 12%. Finally, for companies from countries with strong creditor protection or a more developed stock market, the likelihood of a convertible debt issue falls by 14% and 31%, respectively. Regression (6) tests the robustness of our results to using internal and external corporate

15 Several studies find evidence that the impact of ownership concentration on firms’ corporate governance quality becomes negative when ownership concentration is very high (e.g., Morck et al., 1988; Claessens et al., 2002). In unreported robustness tests, we include squared values of the ownership concentration measures to test for a concave impact. We find that the squared ownership concentration measures are never significant and that other results are robust to their inclusion. 16 The significant positive impact of the French dummy variable on firms’ likelihood to issue convertible debt instead of straight debt is consistent with the popularity of convertible debt in France documented by several previous studies (Ammann et al., 2003; Bancel and Mittoo, 2004; Dutordoir and Van de Gucht, 2007) as well as by business press articles (Wright, 2000; de Teran, 2001).

38 governance indices based on the seven internal and four external corporate governance variables. We construct these indices following the methodology of Maskara and Mullineaux (2011) in developing their information asymmetry index. We determine the quintile of each continuous corporate governance measure, calculate the quintile average across both internal and external continuous variables, and add the values of dummy variables to the quintile averages. We find a significant negative impact for both indices, corroborating our evidence for the Substitution and Entrenchment hypotheses.

2.4.2. The choice between convertible debt and seasoned equity In this section we examine the impact of corporate governance quality on firms’ propensity to issue convertible debt instead of equity. The Substitution (Complementarity) hypotheses predict a negative (positive) impact of corporate governance quality proxies on firms’ propensity to issue convertible debt instead of equity. The Entrenchment hypothesis does not yield a prediction on the choice between convertible debt and equity, as it perceives convertible debt as an alternative to straight debt. Table 7 reports the results. Regression (1) includes only control variables on the right hand side. The findings are mostly consistent with our predictions. We find a significant negative impact of income taxes to total assets and the leading indicator, and a significant positive impact of stock market run- up, Treasury bond yields, and stock market volatility on firms’ likelihood to issue convertibles instead of equity. We also find a significant positive impact of total assets. Regression (2), which includes the different governance quality attributes, shows that firms characterized by weaker corporate governance, as proxied by lower values of ownership concentration (1), shareholder rights, and creditor rights, have a significantly higher likelihood of issuing convertible debt instead of equity. Regression (3) shows results using ownership concentration (2) instead of ownership concentration (1). The findings remain the same. Regression (4) controls for issues clustered in 2005 and 2006, while regression (5) controls for French issues. In both cases the results remain mostly unchanged. Most important, the significant negative impact of ownership concentration and shareholder rights remains intact. The only difference is that the creditor rights variable loses its significant negative impact. The marginal effects of the significant corporate governance variables evaluated at the variable means (not tabulated) show that a 1% increase in ownership concentration (2) reduces the likelihood of a convertible debt issue by 0.5%, and that coming from a country with strong creditor protection reduces this likelihood by 33%.

39 Regression (6) examines the security choice using the two corporate governance indices instead of the separate corporate governance variables. A significant negative impact of the external governance quality index on the propensity to issue convertible bonds confirms the previous findings. Overall, these results suggest that corporate governance quality has an economically significant, negative impact on firms’ likelihood to issue convertibles instead of seasoned equity.

2.4.3. Detailed analysis of blockholder categories In this subsection, we explore whether the negative impact of ownership concentration on firms’ likelihood to issue convertible debt instead of non-hybrid securities is robust across ownership categories. Cronqvist and Fahlenbrach (2009) provide evidence that blockholders are heterogeneous in terms of beliefs, skills, and preferences. Most relevant for our research are differences in the strength of monitoring activity across blockholder categories. We expect blockholder categories with stronger monitoring incentives to have a stronger positive impact on firms’ corporate governance quality, and therefore a stronger impact on firms’ choice between convertible debt and other security types. Using Orbis and Thomson ONE Banker, we classify each blockholder into one of the following ten categories: banks, financial advisory firms, foundations, governments, hedge funds/ firms, industrials, insurance companies, management, mutual funds, and private (mostly family) owners. Among the ten categories, we expect foundations and private owners to have the strongest monitoring incentives due to their limited wealth diversification. Following Thomsen et al. (2006) who study corporate ownership in Europe, we create dummy variables taking the value one if a particular blockholder category owns at least 10% of the company’s shares at the fiscal year-end before the offering announcement date. Using these dummy variables, we re-estimate the multinomial logit model to test whether the negative impact of blockholders is robust across different blockholder types. Table 8 reports the results. For the propensity to issue convertible debt instead of straight debt, the results show that foundations, governments, and private owners have a significant negative impact on the likelihood of issuing convertible debt. The results on foundations and private owners are consistent with these blockholder types having strong incentives to monitor firms. All other categories, besides banks (significantly positive), have an insignificant impact on this choice. The impact of all other corporate governance characteristics and control variables is similar to

40 the results obtained with the aggregate ownership concentration measures. For the propensity to issue convertible debt instead of equity, financial advisory firms, governments, and private owners have a significant negative impact, while the coefficients on the other blockholder categories are insignificant. The significant negative effect of private owners is in line with their monitoring incentives resulting in a strong link with corporate governance quality. Finally, the significant negative impact of governments suggests that governments use their large equity stakes for stronger monitoring in line with the public interest to limit managerial discretion (Boycko et al., 1993; Wolfram, 1998). Moreover, it does not confirm the argument that governments simply serve special interest groups such as trade unions which will ultimately help them to win elections (Lopez-de-Silanes et al., 1997).

2.4.4. Stock returns around convertible debt announcements The security choice analysis suggests that companies with weaker corporate governance are more likely to issue convertible bonds. This result could stem either from convertible debt being used to reduce financing costs (the Substitution hypothesis) or from convertible debt being issued by entrenched managers (the Entrenchment hypothesis). To disentangle these two explanations, this section examines announcement returns of convertible debt issues. If the Substitution hypothesis holds, companies with weak corporate governance use convertible debt to reduce external financing costs. If stockholders acknowledge this motivation, we should observe a more favorable stock price reaction to convertible debt announcements by issuers with weaker governance quality, since convertibles are more useful for these firms. Conversely, if entrenched managers issue convertible debt to further insulate themselves from market forces and stockholders are aware of this motivation, we should observe a less favorable reaction to convertible debt announcements by weak governance issuers. To examine which prediction holds, we regress cumulative abnormal returns (CARs) calculated using a single factor market model on corporate governance characteristics and control variables. We estimate the market model over trading days −300 to −46 relative to the announcement date. Following Lease et al. (1991) and Abhyankar and Dunning (1999) we calculate CARs over a (0, 1) window to allow for announcements occurring on day 0 after the close of trade. We use the same corporate governance and control variables as in the logistic regressions following Jung et al. (1996) and Lewis et al. (1999), who argue that a theory of corporate security choice should explain both the choice itself and the stock market reaction to the security choice announcement. Table 9 shows the results of these regressions. Regression (1) uses the corporate governance and control variables as independent

41 variables. The results show a significant negative impact of the founder CEO dummy variable and of creditor rights, which are both in line with the Substitution hypothesis. The remaining corporate governance proxies are insignificant. To further disentangle the Substitution and Entrenchment hypotheses, regression (2) includes a dummy variable taking the value one if the convertible bonds are callable. If the Entrenchment hypothesis holds, callable convertibles should yield more negative stockholder reactions since conversion-forcing bond calls might allow entrenched managers to avoid bankruptcy even if liquidation would be optimal for the firm (Isagawa, 2002). As we are now considering convertible bond design, we also include the convertibles delta in the regression for completeness. Delta is the sensitivity of the convertible bond value with respect to the underlying stock value at the announcement date. A higher delta indicates a more equity-like convertible bond. Delta is calculated as  lnS X   r  T 2 2    eTT N d  e N   (1) 1     T  where δ is the continuously compounded dividend yield for the fiscal year end preceding the announcement date, T is the initial convertible debt maturity (in years), S is the price of the underlying stock measured seven days before the announcement date, X is the conversion price, r is the continuously compounded yield on a five-year German Treasury bond (measured on the announcement date), and σ is the annual stock return volatility. The regression results show insignificant impacts of the call dummy and delta on announcement returns. The founder CEO dummy and creditor rights variables keep their significant negative impacts. We thus conclude that the significant results in the announcement returns analysis hint in the direction of the Substitution rather than the Entrenchment hypothesis. However, only a few corporate governance and control variables have a significant impact on convertible bond announcement returns. This low explanatory power is probably attributable to the high noise to signal ratio of daily abnormal stock returns. Other studies in the literature also obtain poor explanatory power for regressions explaining stock price reactions to security offering announcements (e.g., Lewis et al., 1999; Dutordoir and Hodrick, 2012). Moreover, the announcement returns may be affected by arbitrage- related convertible bond short selling of hedge funds around the announcement date resulting in significantly negative announcement effects of an average magnitude of −3 to −4% (de Jong et al., 2011). Brown et al. (2012) show that firm with high costs of seasoned equity financing actively sell their convertible debt issues to hedge funds, who combine the long convertible position with a short position in the firms stock. To maximize these hedging

42 profits, hedge funds prefer, for example, the underlying stock to have a larger stock return volatility. Hedge funds thus distribute equity exposure to a large number of investors via their short positions and firms can reduce their financing costs compared to issuing seasoned equity.

2.5. Conclusion We examine the impact of corporate governance quality on firms’ choice between convertible debt, straight debt, and equity. We formulate three hypotheses on the potential impact of corporate governance quality on convertible bond issuance and announcement returns, and test these hypotheses on a sample of Western European security offerings made between 2000 and 2010. Our main finding is that companies with weaker corporate governance are significantly more likely to issue convertible debt than straight debt or seasoned equity. Our results on internal corporate governance mechanisms indicate a significant negative impact of corporate blockholders on the convertible bond choice. This result mainly holds for blockholders with strong monitoring incentives. The strong impact of blockholders on corporate security choices is in line with findings of other studies on corporate governance in Europe. For example, Drobetz et al. (2004) argue that US firms, which traditionally have highly dispersed ownership, rely mainly on the legal protection of minority investors, the monitoring role of boards, and the market for corporate control to reduce agency conflicts. By contrast, German firms rely heavily on the monitoring role of blockholders. We also find a significant negative impact of several country-specific proxies for corporate governance quality on firms’ propensity to issue convertible debt. The results are robust to measuring corporate governance quality through composite indices rather than with individual proxies. Our analysis suggests that future models of corporate security choice should control for both internal and external (country-specific) corporate governance quality proxies. To further disentangle whether the security choice is consistent with shareholder value maximization (the Substitution hypothesis) or (the Entrenchment hypothesis), we analyze stock returns around convertible bond announcements. In line with the former hypothesis, we find that convertible debt announcement returns are negatively influenced by some measures of corporate governance quality. Our results thus paint a favorable image of firms’ motives for using convertible debt. That is, firms seem to use convertibles to achieve lower agency and adverse selection costs, rather than as an entrenchment mechanism. A potential reason for the lack of evidence for Isagawa’s (2002)

43 entrenchment rationale in our data is that we focus on a Western European setting. It would be interesting to verify the validity of this rationale for a sample of security issues in emerging markets, where managers and families routinely employ pyramid ownership structures to give themselves control rights that far exceed their proportional cash flows (Harvey et al., 2004). A priori, we would expect to obtain more evidence for an entrenchment rationale on security issuance in such a context.

44 References

Abhyankar, A. and Dunning, A., 1999. Wealth effects of convertible bond and convertible preference share issues: An empirical analysis of the UK market. Journal of Banking and Finance 23, 1043–1065. Adams, R.B., Hermalin, B.E. and Weisbach, M.S., 2010. The role of boards of directors in corporate governance: A conceptual framework and survey. Journal of Economic Literature 48, 58–107. Aggarwal, R., Erel, I., Stulz, R. and Williamson, R., 2009. Differences in governance practices between U.S. and foreign firms: Measurement, causes, and consequences. Review of Financial Studies 22, 3171–3209. Agrawal, A. and Knoeber, C.R., 1996. Firm performance and mechanisms to control agency problems between managers and shareholders. Journal of Financial and Quantitative Analysis 31, 377–397. Ammann, M., Kind, A. and Wilde, C., 2003. Are convertible bonds underpriced? An analysis of the French market. Journal of Banking and Finance 27, 635–653. Anderson, R.C., Mansi, S.A. and Reeb, D.M., 2004. Board characteristics, accounting report integrity, and the cost of debt. Journal of Accounting and Economics 37, 315–342. Bancel, F. and Mittoo, U.R., 2004. Why do European firms issue convertible debt? European Financial Management 10, 339–373. Bayless, M. and Chaplinsky, S., 1996. Is there a window of opportunity for seasoned equity issuance. Journal of Finance 51, 253–278. Bebchuk, L., Cohen, A. and Ferrel, A., 2009. What matters in corporate governance? Review of Financial Studies 22, 783–827. Becker-Blease, J.R. and Irani, A.J., 2008. Do corporate governance attributes affect adverse selection costs? Evidence from seasoned equity offerings. Review of Quantitative Finance and Accounting 30, 281–296. Berger, P., Ofek, E. and Yermack, D., 1997. Managerial entrenchment and capital structure decisions. Journal of Finance 52, 1411–1438. Bertrand, M. and Mullainathan, S., 2003. Enjoying the quiet life? Corporate governance and managerial preferences. Journal of Political Economy 111, 1043–1075. Bhojraj, S. and Sengupta P., 2003. Effect of corporate governance on bond ratings and yields: The role of institutional investors and outside directors. Journal of Business 76, 455–475. Billingsley, R.S. and Smith, D.M., 1996. Why do firms issue convertible debt? Financial Management 25, 93–99. 45 Boycko, M., Shleifer, A. and Vishny, R.W., 1993. Privatizing Russia. Brookings papers on economic activity 2, 139–192. Brennan, M.J. and Kraus, A., 1987. Efficient financing under asymmetric information. Journal of Finance 25, 1225–1243. Brennan, M.J. and Schwartz, E.S., 1988. The case for convertibles. Journal of Applied Corporate Finance 1, 55–64. Brown. S.J., Grundy, B.D., Lewis, C.M. and Verwijmeren, P., 2012. Convertibles and hedge funds as distributors of equity exposure. Review of Financial Studies 25, 3077–2112. Chava, S., Praveen, K. and Warga, A., 2010. Managerial agency and bond covenants. Review of Financial Studies 23, 1120–1148. Chen, A., Kao, L., Tsao, M. and Wu, C., 2007. Building a corporate governance index from the perspectives of ownership and leadership for firms in Taiwan. Corporate Governance 15, 251–261. Choe, H., Masulis, R.W. and Nanda, V., 1993. offerings across the business cycle: Theory and evidence. Journal of Empirical Finance 1, 3–31. Claessens, S., Djankov, S., Fan, J.P.H. and Lang, L.H.P., 2002. Disentangling the entrenchment and incentive effects of large shareholdings. Journal of Finance 57, 2741– 2771. Conyon, M.J. and Peck, S.I., 1998. Board size and corporate performance: Evidence from European countries. European Journal of Finance 4, 291–304. Core, J.E., Holthausen, R.W. and Larcker, D.F., 1999. Corporate governance, chief executive officer compensation, and firm performance. Journal of Financial Economics 51, 371– 406. Cremers, K.J.M. and Nair, V.B., 2005. Governance mechanisms and equity prices. Journal of Finance 60, 2859–2894. Cronqvist, H. and Fahlenbrach, R., 2009. Large shareholders and corporate policies. Review of Financial Studies 22, 3941–3976. Danielson, M.G. and Karpoff, J.M., 2006. Do pills poison operating performance? Journal of Corporate Finance 12, 536–559. Dann, L.Y. and Mikkelson, W.H., 1984. Convertible debt issuance, capital structure change and financing-related information: Some new evidence. Journal of Financial Economics 13, 157–186. Datta, S., Iskandar-Datta, M. and Raman, K., 2005. Managerial stock ownership and the maturity structure of corporate debt. Journal of Finance 60, 2333–2350.

46 Demirguc-Kunt, A. and Maksimovic, V., 1999. Institutions, financial markets, and firms’ choice of debt maturity. Journal of Financial Economics 54, 295–336. De Jong, A., Dutordoir, M. and Verwijmeren, P., 2011. Why do convertible issuers simultaneously repurchase stock? An arbitrage-based explanation. Journal of Financial Economics 100, 113–129. de Teran, N., 2001. French flair and sophistication dominate European market. The Financial News, 12/02/2001. Diamond, D.W., 1991. Debt maturity structure and liquidity risk. Quarterly Journal of Economics 106, 709–737. Djankov, S., La Porta, R., Lopes-de-Silanes, F. and Shleifer, A., 2008. The law and economics of self-dealing. Journal of Financial Economics 88, 430–465. Doidge, C., Karolyi, G.A. and Stulz, R.M., 2007. Why do countries matter so much for corporate governance? Journal of Financial Economics 86, 1–39. Drobetz, W., Schillhofer, A. and Zimmermann, H., 2004. Corporate governance and expected stock returns: Evidence from Germany. European Financial Management 10, 267–293. Dong, M., Dutordoir, M. and Veld, C., 2012. Why do firms issue convertible bonds? Evidence from the field. Working Paper. Dutordoir, M. and Hodrick, L.S., 2012. Self-selection and stock returns around corporate security offering announcements. Working Paper. Dutordoir, M., Lewis, C.M., Seward, J.K. and Veld, C., 2012. What we do and do not know about convertible bond financing. Working Paper. Dutordoir, M. and Van de Gucht, L., 2007. Are there windows of opportunity for convertible debt issuance? Evidence for Western Europe. Journal of Banking and Finance 31, 2828– 2846. Dutordoir, M. and Van de Gucht, L., 2009. Why do Western European firms issue convertibles instead of straight debt or equity? European Financial Management 15, 563– 583. Erel, I., Julio, B., Kim, W. and Weisbach, M.S., 2012. Macroeconomic conditions and capital raising. Review of Financial Studies 25, 341–376. Faccio, M. and Lang, L.H.P., 2002. The ultimate ownership of Western European corporations. Journal of Financial Economics 65, 365–395. Faccio, M., Marchica, M-T. and Mura, R. 2011. Large shareholder diversification and corporate risk-taking. Review of Financial Studies 24, 3601–3641.

47 Fama, E.F., 1970. Efficient capital markets: A review of theory and empirical work. Journal of Finance 25, 383–417. Frank, M.Z. and Goyal, V.K., 2009. Capital structure decisions: Which factors are reliably important? Financial Management 38, 1–37. Gibbons, R. and Murphy, K.J., 1992. Optimal incentive contracts in the presence of career concerns. Theory and evidence. Journal of Political Economy 88, 468–505. Gompers, P., Ishii, J. and Metrick, A., 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118, 107–156. Graham, J.R. and Harvey, C.R., 2001. The theory and practice of corporate finance: Evidence from the field. Journal of Financial Economics 60, 187–243. Green, R.C., 1984. Investment incentives, debt and warrants. Journal of Financial Economics 13, 115–136. Hart, O. and Moore, J., 1995. Debt and : An analysis of the role of hard claims in constraining management. American Economic Review 85, 567–585. Harvey, C.R., Lins, K.V. and Roper, A.H., 2004. The effect of capital structure when expected agency costs are extreme. Journal of Financial Economics 74, 3–30. Isagawa, N., 2000. Convertible debt: An effective instrument to control managerial opportunism. Review of Financial Economics 9, 15–26. Isagawa, N., 2002. Callable convertible debt under managerial entrenchment. Journal of Corporate Finance 8, 255–270. Jensen, M.C., 1986. Agency costs of free cash flows, corporate finance and takeovers. American Economic Review 76, 323–329. Jensen, G.R., Solberg, D.P. and Zorn, T.S., 1992. Simultaneous determination of insider ownership, debt, and dividend policies. Journal of Financial and Quantitative Analysis 27, 247–261. Jensen, M.C. and Meckling W.H., 1976. Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305–360. Johnson, S., Boone, P., Breach, A. and Friedman, E., 2000. Corporate governance and the Asian financial crisis. Journal of Financial Economics 58, 141–186. Jung, K., Kim, Y-C. and Stulz, R.M., 1996. Timing, investment opportunities, managerial discretion, and the security issue decision. Journal of Financial Economics 42, 159–186. Korkeamaki, T.P., 2005. Effects of law on corporate financing practices – international evidence from convertible bond issues. Journal of Corporate Finance 11, 809–831.

48 Kothari, S.P. and Warner, J.B., 1997. Measuring long-horizon security price performance. Journal of Financial Economics 43, 301–339. Krasker, W.S., 1986. Stock price movements in response to stock issues under asymmetric information. Journal of Finance 41, 93–105. La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R.W., 1998. Law and finance. Journal of Political Economy 106, 1113–1155. La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R.W., 2002. Investor protection and corporate valuation. Journal of Finance 58, 1147–1169. Lee, C-F., Lee, K-W. and Yeo, G. H-H., 2009. Investor protection and convertible debt design. Journal of Banking and Finance 33, 985–995. Lease, R.C., Masulis, R.W. and Page, J.R., 1991. An investigation of market microstructure impacts on event study returns. Journal of Finance 46, 1523–1536. Leland, H.E. and Pyle, D.H., 1977. Informational asymmetries, financial structure, and financial intermediation. Journal of Finance 32, 371–387. Lewis, C.M., Rogalski, R.J. and Seward, J.K., 1999. Is convertible debt a substitute for straight debt or for common equity? Financial Management 28, 5–27. Li, X., Low, A. and Makhija, A.K., 2011. Career concerns and the busy life of the young CEO. Working paper. Lin, W.C. and Chang, S.C., 2012. Corporate governance and the stock market reaction to new product announcements. Review of Quantitative Finance and Accounting 39, 273–291. Lucas, D.J. and McDonald, R.L., 1990. Equity issues and stock price dynamics. Journal of Finance 45, 1019–1043. Maskara, P.K. and Mullineaux, D.J., 2011. Information asymmetry and self-selection in bank loan announcement studies. Journal of Financial Economics 101, 684–694. Masulis, R.W., Wang, C. and Xie, F., 2007. Corporate governance and acquirer returns. Journal of Finance 62, 1851–1889. Mayers, D., 1998. Why firms issue convertible bonds: The matching of financial and real investment options. Journal of Financial Economics 47, 83–102. Mehran, H., 1992. Executive incentive plans, corporate control, and capital structure. Journal of Financial and Quantitative Analysis 27, 539–560. Mikkelson, W.H. and Partch, M.M., 1986. Valuation effects of security offerings and the issuance process. Journal of Financial Economics 15, 31–60.

49 Miller, D.P. and Reisel, N., 2012. Do country-level investor protections affect security-level contract design? Evidence from foreign bond covenants. Review of Financial Studies 25, 408–438. Modigliani, F. and Miller, M.H., 1963. Corporate income taxes and the cost of capital: A correction. American Economic Review 53, 433–443. Morck, E., Shleifer, A. and Vishny, R.W., 1988. Management ownership and market valuation: An empirical analysis. Journal of Financial Economics 20, 293–315. Morellec, E., 2004. Can managerial discretion explain observed leverage ratios? Review of Financial Studies 17, 257–294. Murphy, K.J., 1986. Incentives, learning and compensation. A theoretical and empirical investigation of managerial labor contracts. Rand Journal of Economics 17, 59–76. Myers, S.C., 1977. Determinants of corporate borrowing. Journal of Financial Economics 5, 147–175. Myers, S.C. and Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13, 187–221. Novaes, W. and Zingales, L., 1995. Capital structure choice when managers are in control: Entrenchment versus efficiency. Working Paper. Palia, D., 2001. The endogeneity of managerial compensation in firm valuation: A solution. Review of Financial Studies 14, 735–764. Rediker, K.J. and Seth, A., 1995. Boards of directors and substitution effects of alternative governance mechanisms. Strategic Management Journal 16, 85–99. Petersen, M.A., 2009. Estimating standard errors in finance panel datasets: Comparing approaches. Review of Financial Studies 22, 435–480. Prendergast, C. and Stole, L.A., 1996. Impetuous youngsters and jaded old-timers: Acquiring a reputation for learning. Journal of Political Economy 104, 1105–1134. Rosenstein, S. and Wyatt, J.G., 1990. Outside directors, board independence and shareholder wealth. Journal of Financial Economics 26, 175–191. Rossi, S. and Volpin, P.F., 2004. Cross-country determinants of . Journal of Financial Economics 74, 277–304. Rutherford, M.A., Buchholtz, A.K. and Brown, J.A., 2007. Examining the relationship between monitoring and incentives in corporate governance. Journal of Management Studies 44, 414–430.

50 Schepker, D.J. and Oh, W.Y., 2013. Complementary or substitutive effects? Corporate governance mechanisms and poison pill repeal. Forthcoming Journal of Management. Shleifer, A. and Vishny, R.W., 1997. A survey of corporate governance. Journal of Finance 52, 737–783. Singh, M. and Davidson, W.N.III., 2003. Agency costs, ownership structure and corporate governance mechanisms. Journal of Banking and Finance 27, 793–816. Stein, J.C., 1992. Convertible bonds as backdoor equity financing. Journal of Financial Economics 32, 3–21. Stiglitz, J.E., 1985. Credit markets and the control of capital. Journal of Money, Credit and Banking 17, 133–152. Stulz, R.M., 1990. Managerial discretion and optimal financing policies. Journal of Financial Economics 26, 3–27. Sundaramurthy, C., Mahoney, J.M. and Mahoney, J.T., 1997. Board structure, antitakeover provisions and stockholder wealth. Strategic Management Journal 18, 231–245. Thomsen, S., Pedersen, T. and Kvist, H.K., 2006. Blockholder ownership: Effects on firm value in market and control based governance systems. Journal of Corporate Finance 12, 246–269. Train, K., 2009. Discrete choice models with simulation, 2nd Ed. Cambridge University Press. Voulgaris, G., Stahopoulos, K. and Walker M., 2010. Compensation consultants and CEO pay. Corporate Governance 18, 511–526. Ward, A.J., Brown, J.A. and Rodriguez, D., 2009. Governance bundles, firm performance, and the substitutability and complementarity of governance mechanisms. Corporate Governance 17, 646–660. Westphal, J.D. and Zajac, E.J., 1994. Substance and symbolism in CEOs’ long-term incentive plans. Administrative Science Quarterly 39, 367–390. Wolfram, C.D., 1998. Increases in executive pay following privatization. Journal of Economics and Management Strategy 7, 327–361. Wright, W., 2000. Gallic excellence in convertibles. The Financial News, 10/07/2000. Yermack, D., 1996. Higher market valuations for firms with a small board of directors. Journal of Financial Economics 40, 185–211. Yim, S., 2013. The acquisitiveness of youth: CEO age and acquisition behavior. Journal of Financial Economics 108, 250–273. Zwiebel, J., 1996. Dynamic capital structure under managerial entrenchment. American Economic Review 86, 1197–1215.

51 Table 1 Predicted impact of corporate governance quality on the choice to issue convertibles and on stock returns around convertible debt announcements

This table reports the expected impact of corporate governance quality on the choice to issue convertible debt compared with both debt and equity, and on convertible debt announcement returns. The Substitution hypothesis predicts a substitution effect between convertible debt and corporate governance quality. The Complementarity hypothesis predicts a complementary relationship between convertible debt and corporate governance quality. The Entrenchment hypothesis follows Isagawa (2002) and predicts that, compared to straight debt, convertible debt is used by firms with entrenched managers (weaker corporate governance) to secure their positions. CD is convertible debt, SD straight debt, and SE seasoned equity. Rationale Impact of corporate governance quality on CD vs. SD CD vs. SE Announcement returns Substitution hypothesis − − − Complementarity hypothesis + + No prediction Entrenchment hypothesis − No prediction +

52 Table 2 Descriptive statistics for security issues

The table reports the number of issues by year and by country, along with percentages, for samples of convertible debt (panel A), straight debt (panel B), and seasoned equity (panel C) issues offered by industrial companies from 13 European countries between January 2000 and September 2010. Data on all issues are from Thomson ONE Banker. Issue year Number of issues Percentage Country Number of issues Percentage Panel A: Convertible debt issues 2000 7 3.95 Austria 9 5.11 2001 14 7.91 Belgium 3 1.70 2002 14 7.91 Denmark 0 0.00 2003 19 10.73 Finland 3 1.70 2004 15 8.47 France 73 41.48 2005 14 7.91 Germany 22 12.50 2006 9 5.08 Italy 6 3.39 2007 21 12.43 Netherlands 15 8.52 2008 4 2.26 Norway 5 2.84 2009 42 23.73 Spain 8 4.52 2010 17 9.60 Sweden 5 2.84 Total 176 100.00 Switzerland 9 5.11 United Kingdom 18 10.22 No. of firms 129 73.30 Total 176 100.00 Panel B: Straight debt issues 2000 41 11.71 Austria 7 2.00 2001 54 15.43 Belgium 8 2.29 2002 37 10.57 Denmark 3 0.86 2003 31 8.86 Finland 7 2.00 2004 34 9.71 France 101 28.86 2005 34 9.71 Germany 43 12.29 2006 19 5.43 Italy 16 4.57 2007 24 6.86 Netherlands 27 7.71 2008 11 3.14 Norway 4 1.14 2009 36 10.29 Spain 2 0.57 2010 29 8.29 Sweden 15 4.29 Total 350 100.00 Switzerland 41 11.71 United Kingdom 76 21.71 No. of firms 154 44.00 Total 350 100.00 Panel C: Equity issues 2000 0 0.00 Austria 2 1.42 2001 0 0.00 Belgium 1 0.71 2002 0 0.00 Denmark 3 2.13 2003 0 0.00 Finland 8 5.67 2004 7 4.96 France 12 8.51 2005 78 55.32 Germany 27 19.15 2006 44 31.21 Italy 3 2.13 2007 1 0.71 Netherlands 8 5.67 2008 0 0.00 Norway 6 4.26 2009 1 0.71 Spain 1 0.71 2010 10 7.09 Sweden 9 6.38 Total 141 100.00 Switzerland 5 3.55 United Kingdom 56 39.72 No. of firms 124 87.94 Total 141 100.00

53 Table 3 Variable descriptions

This table defines the corporate governance (panel A) and control variables (panel B) and their data sources. Within each category, variables are in alphabetical order. All variables are measured at the end of the fiscal year before the security offering, unless noted otherwise. Variable Definition Source Panel A: Corporate governance variables Board size Total number of directors on the BoardEx, company reports company's board

CEOage Ageofthecompany’sCEO BoardEx,company reports, capital market filings, newspaper articles

CEO tenure Number of years the company’s BoardEx, company CEO has been in office reports, capital market filings, newspaper articles

Credit market development Sum of all private domestic Euromonitor claims divided by the GDP of the issuing firm's country of domicile

Creditor rights Dummy variable taking the value La Porta et al. (1998) one for countries with a creditor rights index of three or above

External governance quality index Quintile average across all External corporate external continuous corporate governance variables governance variables plus the value of external dummy variables

Founder CEO Dummy variable taking the value BoardEx, company one if the CEO is the founder of reports, newspaper articles the company

Internal governance quality index Quintile average across all Internal corporate internal continuous corporate governance variables governance variables plus the value of internal dummy variables

Outside CEO Dummy variable taking the value BoardEx, company one if the CEO was hired from reports, capital market outside the company and was filings, and newspaper never an employee of the articles company

Outside directors Percentage of independent board BoardEx, company reports members (directors who are not and have never been employed by the company) relative to the total number of board members

Ownership concentration (1) Percentage of shares held by the Orbis, Thomson ONE company's three largest shareholders Banker

54 Table 3 (continued)

Ownership concentration (2) Percentage of shares held by the Orbis, Thomson ONE company’s largest shareholder Banker

Shareholder rights Dummy variable taking the value La Porta et al. (1998) one for countries with an anti- director rights index score of four or above

Stock market development Stock market capitalization Euromonitor divided by the GDP of the issuing firm's country of domicile Panel B: Control variables Delta Sensitivity of the bond with respect Thomson ONE Banker, to the underlying stock value at the Datastream announcement date

Financial slack/total assets Net operating cash flow minus Worldscope cash dividends minus capital expenditures over the book value of total assets

Income tax/total assets Income taxes paid over the book Worldscope value of total assets

Leading indicator Six-month European leading Datastream indicator calculated as the log growth rate over the quarter preceding the issue month

Long-term debt/total assets Book value of long-term debt over Worldscope the book value of total assets

Market-to-book ratio Market value of equity over the Worldscope book value of equity

Return on assets (%) EBIT over the value of total assets Worldscope

Sales growth (%) Growth in sales in the fiscal year Worldscope before the security issue

Short-term debt /total assets Book value of short-term debt Worldscope over the book value of total assets

Stockmarketrun-up ReturnontheMSCIEuropean Datastream equity market index over the window (−200, −20)

Stock price run-up Cumulative daily stock return Datastream over the window 76 to 2 trading days before the security issue

Stock return volatility Annualized stock return volatility Datastream based on daily stock returns measured over the fiscal year before the security issue

55 Table 3 (continued)

Stock market volatility Return volatility of the MSCI Datastream European equity market index over the window (−200, −20)

Treasury bond yield Average yield on 5-year German Datastream Treasury bonds, over the quarter preceding the issue month

Total assets (log) Logarithm of the book value of Worldscope total assets (millions USD)

Total proceeds/market value Total proceeds of the security Thomson ONE Banker, issue over market value of equity Datastream

56 Table 4 Descriptive statistics for corporate governance characteristics

This table reports descriptive statistics for the corporate governance characteristics of companies in the convertible debt (panel A), straight debt (panel B), and equity (panel C) samples. Panel D reports t-statistics for pairwise differences in means of the corporate governance characteristics between convertible debt (CD) and straight debt and convertible debt and seasoned equity issuing firms. Table 3 gives the definition and source of all variables. + (–) next to a variable name indicates it is a proxy (inverse proxy) for corporate governance quality. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variable Mean Median Std. Dev. Min. Max. Panel A: Convertible debt issues Ownershipconcentration(1)(%)(+) 33.47 29.44 20.54 0.99 88.83 Ownershipconcentration(2)(%)(+) 19.92 14.51 16.67 0.57 81.49 Outside directors (+) 0.32 0.33 0.23 0.00 0.80 Board size (−) 12.91 12.00 5.45 3.00 29.00 CEO tenure (years) (−) 5.91 4.05 6.31 0.10 35.80 CEO age (+/−) 53.26 53.00 7.55 32.00 73.00 Outside CEO (−) 0.35 0.00 0.48 0.00 1.00 FounderCEO(+) 0.09 0.00 0.28 0.00 1.00 Shareholder rights (+) 0.18 0.00 0.38 0.00 1.00 Creditor rights (+) 0.28 0.00 0.45 0.00 1.00 Stockmarketdevelopment(+) 0.84 0.74 0.51 0.13 3.10 Creditmarketdevelopment(+) 1.18 1.08 0.36 0.75 2.14 Panel B: Straight debt issues Ownershipconcentration(1)(%)(+) 33.27 30.50 21.93 1.61 93.00 Ownershipconcentration(2)(%)(+) 22.54 17.00 19.19 0.64 90.15 Outside directors (+) 0.30 0.31 0.23 0.00 0.90 Board size (−) 15.06 14.00 5.43 3.00 30.00 CEO tenure (years) (−) 5.20 3.70 4.82 0.10 32.00 CEO age (+/−) 53.46 54.00 7.21 30.00 83.00 Outside CEO (−) 0.24 0.00 0.42 0.00 1.00 FounderCEO(+) 0.08 0.00 0.27 0.00 1.00 Shareholder rights (+) 0.23 0.00 0.42 0.00 1.00 Creditor rights (+) 0.36 0.00 0.48 0.00 1.00 Stockmarketdevelopment(+) 1.13 1.01 0.66 0.13 3.17 Creditmarketdevelopment(+) 1.18 1.15 0.35 0.40 2.23 Panel C: Equity issues Ownershipconcentration(1)(%)(+) 37.39 33.75 18.83 0.69 88.94 Ownershipconcentration(2)(%)(+) 22.37 18.43 16.45 0.27 72.82 Outside directors (+) 0.30 0.33 0.23 0.00 0.80 Board size (−) 10.59 9.00 5.38 3.00 28.00 CEO tenure (years) (−) 5.09 4.00 4.28 0.10 20.60 CEO age (+/−) 50.66 51.00 8.16 33.00 68.00 Outside CEO (−) 0.32 0.00 0.47 0.00 1.00 FounderCEO(+) 0.15 0.00 0.36 0.00 1.00 Shareholder rights (+) 0.45 0.00 0.50 0.00 1.00 Creditor rights (+) 0.60 1.00 0.49 0.00 1.00 Stockmarketdevelopment(+) 1.00 1.04 0.47 0.23 2.52 Creditmarketdevelopment(+) 1.27 1.48 0.32 2.23 0.67

57 Table 4 (continued)

Panel D: Pairwise differences in means Governance characteristic CD vs. Straight Debt CD vs. Equity Ownership concentration (1) (%) (+) 0.20 −3.92** Ownership concentration (2) (%) (+) −2.62* −2.45* Outside directors (+) 0.02 0.02 Board size (−) −2.14*** 2.31*** CEO tenure (years) (−) 0.71* 0.82* CEO age (+/−) −0.20 2.60*** Outside CEO (−) 0.12*** 0.03 Founder CEO (+) 0.01 −0.06** Shareholder rights (+) −0.06* −0.27*** Creditor rights (+) −0.08** −0.32*** Stock market development (+) −0.29*** −0.16*** Credit market development (+) 0.01 −0.09**

58 Table 5 Descriptive statistics for control variables

This table reports descriptive statistics for the control variables of all companies in the convertible debt sample (panel A), straight debt sample (panel B), and equity sample (panel C). Panel D reports t-statistics for pairwise differences in means of the control variables between convertible debt (CD) and straight debt and convertible debt and seasoned equity issuing firms. Table 3 gives the definition and source of all variables. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Characteristic Mean Median Std. Dev. Min Max Panel A: Convertible debt issues Stock price run-up 0.10 0.05 0.35 −0.46 2.81 Financial slack/total assets −0.02 −0.01 0.10 −0.56 0.35 Totalproceeds/marketvalue 0.21 0.13 0.39 0.0003 3.83 Income tax/total assets 0.01 0.01 0.03 −0.31 0.11 Return on assets (%) 2.32 3.49 10.39 −54.44 31.84 Short-term debt/total assets 0.09 0.07 0.08 0.00 0.40 Long-term debt/total assets 0.24 0.22 0.15 0.00 0.71 Stock return volatility 32.13 30.79 11.44 4.70 63.28 Sales growth 0.23 0.06 1.36 −0.91 17.65 Total assets (millions) 14848 4541 27465 33.00 141521 Market-to-book ratio 2.86 1.71 3.87 0.21 31.73 Stock market run-up 0.02 0.09 0.24 −0.76 0.41 Stock market volatility 0.02 0.01 0.01 0.01 0.03 Treasury bond yield 3.33 3.25 0.88 1.54 5.12 Leading indicator 0.01 0.01 0.01 −0.02 0.02 Panel B: Straight debt issues Stock price run-up 0.03 0.03 0.14 −0.47 0.66 Financial slack/total assets 0.001 0.004 0.06 −0.52 0.17 Totalproceeds/marketvalue 0.11 0.05 0.21 0.0001 1.92 Income tax/total assets 0.02 0.02 0.02 −0.05 0.10 Return on assets (%) 5.66 5.24 6.07 −54.44 38.61 Short-term debt/total assets 0.09 0.08 0.06 0.00 0.37 Long-term debt/total assets 0.23 0.21 0.13 0.00 0.64 Stock return volatility 25.33 23.79 8.33 5.86 61.77 Sales growth 0.11 0.08 0.24 −0.63 1.43 Total assets (millions) 27233 14835 38460 290.00 244143 Market-to-book ratio 3.26 2.14 3.57 0.22 43.03 Stock market run-up 0.02 0.07 0.20 −0.78 0.56 Stock market volatility 0.01 0.01 0.01 0.01 0.03 Treasury bond yield 3.64 3.62 0.94 1.54 5.14 Leading indicator 0.01 0.01 0.01 −0.02 0.02

59 Table 5 (continued)

Panel C: Equity issues Stock price run-up 0.08 0.06 0.18 −0.39 1.18 Financial slack/total assets −0.06 0.01 0.27 −2.10 0.39 Totalproceeds/marketvalue 0.23 0.08 0.63 0.0004 7.06 Income tax/total assets 0.01 0.01 0.03 −0.11 0.15 Return on assets (%) −1.89 4.20 23.35 −168.96 32.11 Short-term debt/total assets 0.06 0.04 0.07 0.00 0.37 Long-term debt/total assets 0.17 0.13 0.17 0.00 0.65 Stock return volatility 36.68 32.89 14.84 8.10 79.07 Sales growth 0.17 0.06 0.79 −0.98 8.41 Total assets (millions) 9985 802 26270 1.00 191159 Market-to-book ratio 4.51 2.24 7.24 0.36 62.34 Stock market run-up 0.11 0.14 0.09 −0.45 0.20 Stock market volatility 0.01 0.01 0.00 0.01 0.03 Treasury bond yield 3.01 3.01 0.45 1.54 4.30 Leading indicator 0.01 0.01 0.00 0.00 0.01 Panel D: Pairwise differences in means Control Variables CD vs. Straight Debt CD vs. Equity Stock price run-up 0.06*** 0.02 Financial slack/total assets −0.02*** 0.04** Total proceeds/market value 0.10*** −0.01 Income tax/total assets −0.01*** 0.00 Return on assets (%) −3.32*** 4.21** Short-term debt/total assets 0.00 0.03*** Long-term debt/total assets 0.02 0.07*** Stock return volatility 6.84*** −4.54*** Sales growth 0.11* 0.06 Total assets (log) −1.15*** 1.83*** Market-to-book ratio −0.44 −1.65*** Stock market run-up −0.00 −0.09*** Stock market volatility 0.00*** 0.01*** Treasury bond yield −0.31*** 0.32*** Leading indicator 0.00** 0.00*

60 Table 6 Multinomial logistic regression analysis of the choice between convertible and straight bonds

This table reports coefficients and p-values of multinomial logit regressions for the choice between convertible debt and straight debt. These pairwise regression results are the outcome of a multinomial security choice model that simultaneously incorporates the choice between convertible debt, straight debt, and equity. Regression (1) includes all the control variables as explanatory variables, regression (2) adds all internal and external corporate governance characteristics, regression (3) substitutes ownership concentration (1) with ownership concentration (2), regression (4) adds a dummy variable to control for year clustering in 2005/2006, regression (5) adds a dummy variable to control for French issues, and regression (6) uses internal and external corporate governance quality indices. Table 3 gives the definition and source of all variables. The McFadden R 2 and observation numbers refer to the entire multinomial model. χ2 is the test statistic for the joint significance of all 11 corporate governance variables. + (–) next to a variable name indicates it is a proxy (inverse proxy) for corporate governance quality. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) (3) (4) (5) (6) Intercept 0.64 2.56 3.13 1.94 0.68 −1.24 (0.57) (0.46) (0.37) (0.58) (0.85) (0.35) Internal governance quality index (+) −0.29** (0.05) External governance quality index (+) −0.29*** (0.00) Ownership concentration (1) (%) (+) −0.02*** −0.02*** −0.02*** (0.01) (0.01) (0.01) Ownership concentration (2) (%) (+) −0.03*** (0.00) Outside directors (%) (+) 0.05 −0.14 −0.10 0.07 (0.92) (0.80) (0.86) (0.90) Board size (log) (−) −0.02 −0.12 −0.06 0.10 (0.95) (0.73) (0.86) (0.78) CEO experience (log) (−) −0.06 −0.05 −0.07 −0.09 (0.51) (0.63) (0.50) (0.35) CEO age (log) (+/−) −0.01 −0.14 −0.12 0.24 (0.99) (0.87) (0.89) (0.78) Outside CEO (−) 0.49* 0.49* 0.51** 0.51** (0.06) (0.06) (0.05) (0.05)

61 Table 6 (continued)

Founder CEO (+) −0.64 −0.71 −0.76 −0.75 (0.18) (0.12) (0.13) (0.11 Shareholder rights (+) −0.46 −0.47 −0.36 −0.51 (0.24) (0.25) (0.36) (0.21) Creditor rights (+) −0.78** −0.90*** −0.83** -0.33 (0.02) (0.01) (0.01) (0.35) Stock market development (+) −1.33*** −1.43*** −1.43*** −1.24*** (0.00) (0.00) (0.00) (0.00) Creditmarketdevelopment(+) 0.85* 1.04** 0.94** 1.36*** (0.07) (0.03) (0.04) (0.01) Stockpricerun-up 0.94** 1.05** 1.16** 0.99** 0.94** 1.14** (0.03) (0.02) (0.03) (0.03) (0.04) (0.02) Financial slack/total assets −0.12 −0.33 −0.22 −0.20 −0.13 −0.52 (0.93) (0.81) (0.87) (0.89) (0.93) (0.68) Total proceeds/market value −0.61* −0.76** −0.63** −0.80** −0.67** −0.80** (0.10) (0.02) (0.05) (0.01) (0.04) (0.02) Income tax/total assets −6.70 −8.39 −7.79 −11.40* −8.94 −6.94 (0.21) (0.13) (0.16) (0.08) (0.12) (0.14) Return on assets −0.04 −0.03 −0.03 −0.04 −0.02 −0.03 (0.18) (0.27) (0.28) (0.22) (0.47) (0.25) Short-term debt/total assets 1.01 0.39 0.76 0.23 0.61 0.04 (0.51) (0.84) (0.70) (0.90) (0.76) (0.98) Long-term debt/total assets 1.78** 1.60* 1.57* 1.57* 2.01** 1.88** (0.02) (0.06) (0.07) (0.07) (0.02) (0.01) Stock return volatility 0.05*** 0.06*** 0.06*** 0.06*** 0.06*** 0.06*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

62 Table 6 (continued)

Total assets (log) −0.47*** −0.63*** −0.62*** −0.66*** −0.71*** −0.56*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Market-to-book ratio −0.01 −0.03 −0.04 −0.03 -0.04 −0.02 (0.59) (0.29) (0.20) (0.36) (0.19) (0.48) Salesgrowth 0.63* 0.97*** 0.96** 0.87** 1.03*** 0.85** (0.06) (0.01) (0.01) (0.02) (0.01) (0.02) Stockmarketrun-up 0.92 0.97*** 0.61 1.17 0.78 0.94 (0.31) (0.01) (0.53) (0.25) (0.43) (0.33) Stock market volatility 66.22** 44.88 39.64 88.82** 42.93 66.01** (0.02) (0.17) (0.23) (0.02) (0.20) (0.03) Treasurybondyield 0.08 0.43** 0.46** 0.68*** 0.41** 0.11 (0.58) (0.03) (0.02) (0.00) (0.04) (0.50) Leading indicator −15.57 −9.59 −7.92 −16.98 −11.91 −14.80 (0.58) (0.75) (0.80) (0.58) (0.70) (0.61) 2005/2006dummy 1.02*** (0.01) Francedummy 1.05*** (0.00)

McFadden R2 35.39 40.58 41.39 43.64 41.84 37.21  2 75.54*** 82.63*** 80.04*** 57.34*** 24.08*** N 667 667 667 667 667 667

63 Table 7 Multinomial logistic regression analysis of the choice between convertible debt and seasoned equity

This table reports coefficients and p-values of multinomial logit regressions of the choice between convertible debt and seasoned equity. These pairwise regression results are the outcome of a multinomial security choice model that simultaneously incorporates the choice between convertible debt, straight debt, and equity. Regression (1) includes all the control variables as explanatory variables, regression (2) adds all internal and external corporate governance characteristics, regression (3) substitutes ownership concentration (1) with ownership concentration (2), regression (4) adds a dummy variable to control for year clustering in 2005/2006, regression (5) adds a dummy variable to control for French issues, and regression (6) uses internal and external corporate governance quality indices. Table 3 gives the definition and source of all variables. The McFadden R 2 and observation numbers refer to the entire multinomial model. χ2 is the test statistic for the joint significance of all 11 corporate governance variables. + (–) next to a variable name indicates it is a proxy (inverse proxy) for corporate governance quality. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) (3) (4) (5) (6) Intercept −13.55*** −16.55*** −16.77*** −11.02** −20.42*** −14.75*** (0.00) (0.00) (0.00) (0.05) (0.00) (0.00) Internal governance quality index (+) −0.11 (0.58) External governance quality index (+) −0.30*** (0.00) Ownership concentration (1) (%) (+) −0.02** −0.02** −0.02* (0.05) (0.04) (0.06) Ownership concentration (2) (%) (+) −0.02** (0.04) Outside directors (%) (+) −0.09 −0.09 0.28 0.18 (0.90) (0.90) (0.72) (0.81) Board size (log) (−) 0.63 0.53 0.55 1.02** (0.19) (0.27) (0.24) (0.04) CEO experience (log) (−) 0.06 0.09 0.10 0.03 (0.67) (0.56) (0.55) (0.82) CEO age (log) (+/−) 0.59 0.56 0.70 1.09 (0.61) (0.63) (0.55) (0.37) Outside CEO (−) −0.07 −0.10 −0.25 −0.08 (0.86) (0.79) (0.53) (0.84)

64 Table 7 (continued)

Founder CEO (+) −0.01 −0.05 −0.23 −0.31 (0.99) (0.92) (0.66) (0.56) Shareholder rights (+) −1.14** −1.12** −1.23** −1.18** (0.04) (0.04) (0.05) (0.03) Creditor rights (+) −0.99** −1.02** −0.76 −0.40 (0.04) (0.03) (0.14) (0.43) Stock market development (+) −0.17 −0.22 0.06 −0.09 (0.69) (0.63) (0.90) (0.85) Creditmarketdevelopment(+) 0.48 0.57 0.30 1.33 (0.52) (0.46) (0.70) (0.10) Stock price run-up −0.55 −0.45 −0.51 −0.11 −0.29 −0.41 (0.64) (0.75) (0.72) (0.92) (0.84) (0.74) Financial slack/total assets −1.01 −1.73 −1.75 −1.87 −1.60 −1.52 (0.43) (0.20) (0.21) (0.15) (0.24) (0.22) Total proceeds/market value 0.19 0.13 0.17 −0.02 0.20 −0.04 (0.67) (0.79) (0.73) (0.96) (0.68) (0.93) Income tax/total assets −14.67** −15.64*** −15.15*** −15.21* −14.79*** −14.53** (0.04) (0.01) (0.01) (0.06) (0.01) (0.03) Returnonassets 0.01 0.02 0.02 0.03 0.03 0.02 (0.38) (0.24) (0.26) (0.14) (0.17) (0.30) Short-term debt/total assets 3.08 0.94 0.84 0.76 1.48 2.12 (0.21) (0.71) (0.75) (0.77) (0.58) (0.39) Long-term debt/total assets 1.65 2.71** 2.64** 2.41** 3.51*** 1.68 (0.12) (0.02) (0.03) (0.05) (0.00) (0.12) Stock return volatility 0.01 0.02 0.02 0.01 0.02 0.01 (0.30) (0.15) (0.14) (0.38) (0.16) (0.42)

65 Table 7 (continued)

Total assets (log) 0.28*** 0.10 0.12 0.10 −0.05 0.20** (0.00) (0.41) (0.29) (0.39) (0.70) (0.04) Market-to-book ratio −0.04 −0.06 −0.06* −0.05 −0.07* −0.04 (0.30) (0.11) (0.10) (0.22) (0.06) (0.26) Sales growth 0.06 −0.01 −0.00 −0.03 −0.04 0.07 (0.47) (0.93) (0.98) (0.78) (0.71) (0.45) Stockmarketrun-up 5.88*** 6.71*** 6.72*** 6.64*** 7.06*** 5.97*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Stock market volatility 532.14*** 573.73*** 573.99*** 302.62*** 578.05*** 534.63*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Treasurybondyield 1.72*** 1.94*** 1.97*** 1.41*** 1.95 1.78*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Leading indicator −101.42** −91.75* −92.05* −87.63* −102.98* −104.88** (0.05) (0.09) (0.09) (0.09) (0.06) (0.04) 2005/2006 dummy −2.38*** (0.00) Francedummy 1.87 (0.00)

McFadden R2 35.39 40.58 41.39 43.64 41.84 37.21  2 75.54*** 82.63*** 80.04*** 57.34*** 24.08*** N 667 667 667 667 667 667

66 Table 8 Logistic regression analysis of the security choice considering detailed ownership categories

This table reports coefficients and p-values of multinomial logit regressions for the security choice between convertible debt (CD) and straight debt and between convertible debt and equity. Both regressions include dummy variables taking the value one if a company has a blockholder owning more than 10% of the firm’s equity belonging to one of the following ten categories: banks, financial advisory firms, foundations, governments, hedge funds/private equity firms, industrials, insurance companies, management, mutual funds, and private (mostly family) owners. Table 3 gives the definition and source of all variables. χ2 is the test statistic for the joint significance of the 10 corporate governance variables and the 10 ownership categories. + (–) next to a variable name indicates it is a proxy (inverse proxy) for corporate governance quality. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables Straight Debt Equity Intercept 2.15 −16.32*** (0.55) (0.00) Banks 0.82* 0.33 (0.07) (0.59) Financial advisors −0.40 −0.96* (0.25) (0.06) Foundations −1.74** 1.07 (0.03) (0.28) Government −1.25*** −1.41* (0.01) (0.05) Hedge funds/Private equity firms −0.46 0.09 (0.46) (0.90) Industrials 0.11 −0.12 (0.70) (0.78) Insurance companies −0.34 −1.46 (0.68) (0.18) Management 0.55 1.67 (0.49) (0.27) Mutual funds −0.31 −0.77 (0.47) (0.32) Private owners −1.05*** −1.36** (0.01) (0.02) Outside directors (%) (+) 0.21 −0.17 (0.70) (0.83) Board size (log) (−) −0.03 0.76 (0.94) (0.13) CEO tenure (log) (−) −0.11 0.02 (0.27) (0.87) CEO age (log) (+/−) −0.12 0.24 (0.89) (0.84) Outside CEO (−) −0.61 −0.02 (0.25) (0.98) FounderCEO(+) 0.49* 0.04 (0.06) (0.92) Shareholder rights (+) −0.66 −1.33** (0.10) (0.02) Creditor rights (+) −0.58* −0.87* (0.09) (0.08) Stock market development (+) −1.15*** 0.08 (0.00) (0.87)

67 Table 8 (continued)

Creditmarketdevelopment(+) 0.74 0.54 (0.15) (0.53) Stock price run-up 0.99* −0.56 (0.07) (0.73) Financial slack/total assets −0.94 −1.99 (0.53) (0.20) Total proceeds/market value −0.85*** 0.07 (0.01) (0.89) Income tax/total assets −11.56* −20.88** (0.08) (0.01) Return on assets −0.02 0.03 (0.40) (0.27) Short-term debt/total assets −0.37 1.08 (0.86) (0.71) Long-term debt/total assets 1.69* 3.57*** (0.07) (0.01) Stock return volatility 0.05*** 0.02 (0.00) (0.19) Total assets (log) −0.60*** 0.11 (0.00) (0.39) Market-to-book ratio −0.02 −0.04 (0.47) (0.26) Sales growth 1.03*** 0.06 (0.01) (0.77) Stockmarketrun-up 0.79 7.25*** (0.45) (0.00) Stock market volatility 55.06 588.82*** (0.11) (0.00) Treasury bond yield 0.46** 1.95 (0.03) (0.00) Leading indicator −8.09 −108.38** (0.79) (0.05)

McFadden R2 42.36 42.36  2 98.05*** 98.05*** N 667 667

68 Table 9 Regression analysis of stock returns around convertible debt announcements

This table reports coefficients and p-values of linear regressions of announcement returns of convertible debt issues. The dependent variable is the CAR (Cumulative Abnormal Return) calculated using a one factor market model over the window (0, 1). Company and market returns are from Datastream. Regression (1) includes all internal and external corporate governance characteristics as well as control variables. Regression (2) includes a dummy variable for callable convertibles and convertible bond delta. Table 3 gives the definition and source of all variables. + (–) next to a variable name indicates it is a proxy (inverse proxy) for corporate governance quality. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) Intercept 12.47 17.36 (0.35) (0.25) Ownership concentration (1) (+) −0.02 −0.03 (0.31) (0.23) Outside directors (%) (+) −0.55 −1.56 (0.65) (0.27) Board size (log) (−) 0.16 −0.14 (0.77) (0.83) CEO experience (log) (−) 0.23 0.25 (0.19) (0.18) CEO age (log) (+/−) −2.19 −2.36 (0.42) (0.40) Outside CEO (−) 0.57 0.41 (0.37) (0.43) Founder CEO (+) −1.83** −2.06** (0.03) (0.02) Shareholder rights (+) 1.85 2.27 (0.12) (0.12) Creditor rights (+) −1.55* −2.08** (0.10) (0.05) Stock market development (+) −0.64 −0.67 (0.32) (0.21) Creditmarketdevelopment(+) 0.42 0.57 (0.63) (0.54) Stock price run-up 0.48 1.30 (0.55) (0.21) Financial slack/total assets 2.57 5.06 (0.43) (0.27) Total proceeds/market value 0.39 −0.65 (0.43) (0.72) Income tax/total assets 6.32 5.64 (0.13) (0.65) Return on assets −0.01 0.00 (0.61) (0.97) Short-term debt/total assets 2.98 1.23 (0.34) (0.65) Long-term debt/total assets 1.86 1.16 (0.30) (0.56) Stock return volatility −0.00 −0.06 (0.66) (0.14)

69 Table 9 (continued)

Total assets (log) −0.00 −0.31 (0.43) (0.28) Market-to-book ratio −0.21 −0.06 (0.16) (0.16) Sales growth −0.06 −1.93*** (0.29) (0.01) Stock market run-up −2.27 −2.85 (0.29) (0.21) Stock market volatility −127.01 −160.78* (0.12) (0.08) Treasury bond yield −0.22 −0.35 (0.45) (0.39) Leading indicator 112.91* 121.10 (0.08) (0.12) Call Dummy −0.27 (0.67) Delta 2.01 (0.23)

R 2 17.10 23.76 Adjusted R 2 2.34 7.59 N 173 161

70 Chapter 3 The signaling content of security offering proceeds

Abstract A number of theories predict that the magnitude of security offering proceeds provides a signal to the market regarding the issuers’ expected earnings, overvaluation, or growth opportunities. We test the validity of these theories using a sample of U.S. seasoned equity, convertible debt, and straight debt offerings between 1999 and 2011. Using a continuous- variable selectivity correction, we find that stockholders use predicted issue sizes of equity and convertible offerings as signals of growth opportunities, whilst larger than predicted issue sizes signal issuer overvaluation. For straight debt issues, we find that larger than expected issue sizes signal growth opportunities. Our findings suggest that event studies on the announcement effects of securities issues should control for the endogeneity of issue size.

71 3.1. Introduction When companies need to raise financing externally, they face two decisions: what security type to issue, and how much financing to raise. Numerous studies have shown that stockholders use companies’ security choice as a signal regarding firm overvaluation. In line with Myers and Majluf’s (1984) signaling model, equity is commonly found to provide the most negative signal, whilst straight bonds carry virtually no signaling content.1 However, the potential signaling content of the size of the security offering has been largely ignored in the literature so far. Our paper intends to fill this gap by developing and testing three hypotheses on the stock price impact of offering proceeds derived from the theories of Miller and Rock (1985), Krasker (1986), and Ambarish et al. (1987).2 The earnings shortfall hypothesis predicts that larger than expected security offering sizes signal an expected reduction in earnings, and should therefore negatively affect stockholder reactions to all security offerings (Miller and Rock, 1985). The overvaluation hypothesis argues that the market might use offering sizes as a signal of issuer overvaluation (Krasker, 1986). It therefore leads to the prediction of a negative impact of issue size on security offering announcement returns, the magnitude of which should be larger for more equity(-linked) securities. The growth opportunities hypothesis, based on Ambarish et al. (1987), in turn, argues that security offerings can provide a positive signal to the market regarding the value of firms’ growth opportunities. It leads to the prediction of a positive impact of issue size on security offering announcement returns, the magnitude of which should again be larger for more equity(-linked) offerings. To examine the determinants of issue size and to test the three hypotheses on the stock market reaction to issue size, we use a sample of U.S. corporate security offerings including 1,160 seasoned equity, 716 convertible debt, and 1,951 straight debt issues made between 1999 and 2011. We control for the endogeneity of issue size using an extension of Heckman’s (1979) two-step procedure developed by Garen (1984) for a continuous selection variable, in our case issue size. This allows us to separate the stock price impact of predicted issue size from that of unexpected issue size.3 Moreover, unlike two-stage least squares instrumental

1 E.g., Marsh (1982); MacKie-Mason (1990); Bayless and Chaplinsky (1991); Jung et al. (1996); Lewis et al. (1999); Hovakimian et al. (2001). 2 We use the terms offering proceeds and issue size interchangeably throughout the paper. 3 Acharya (1988) argues that abnormal stock returns should only be affected by unexpected information in an efficient market. However, security offering announcement returns may still be influenced by expected issue sizes to the extent that the market could not fully anticipate the security offering, which seems a reasonable assumption. That is, the issuers’ stock price prior to the security offering announcement only reflects the expected issue size multiplied by the market’s assessment of the probability of the security offering, and any remaining influence of issue size on announcement returns reflects the resolution of the market’s uncertainty about the offering taking place. 72 variables procedures, the Garen methodology enables us to test for a different impact of managerial private information for different levels of issue size. We conduct this analysis separately for each security type. We control for a large set of variables capturing firm- specific and macroeconomic financing costs. Our first-stage results on the determinants of offering proceeds indicate that, whilst issue size is partly dictated by funding needs, it also seems partly under the discretion of the issuing firm. We find that offering sizes of equity(-linked) securities are negatively affected by equity-related adverse selection costs proxies, while proceeds of straight debt issues are negatively affected by debt-related financing costs proxies. Our main results regarding the signaling content of offering proceeds are as follows. For seasoned equity and convertible debt, we find a positive impact of predicted issue size on announcement returns. Further tests on firms’ use of proceeds show that larger equity(-linked) offerings are more likely to be used to finance growth, rather than to accumulate cash. Our results are therefore consistent with the growth opportunities hypothesis predicting that stockholders use large issue sizes as a signal of profitable growth options. However, we also find a negative impact of unpredicted issue size on equity and convertible announcement returns. This finding is consistent both with the expected earnings shortfall and overvaluation hypotheses. However, for the earnings shortfall hypothesis, we would also expect to find a negative impact of residual offering sizes for straight debt, which we do not find. We show in additional tests that unexpected equity and convertible debt issue sizes are not linked to a shortfall in earnings in the years after issuance, but rather to an increase in cash holdings, supporting the overvaluation hypothesis. The picture emerging from our results on equity- linked securities offering is, therefore, that the market reacts positively to larger offerings to the extent that they cover firms’ actual funding needs. Issue sizes beyond those funding needs (as captured by the residual issue size variable in our models) are viewed as a negative signal. For straight debt, our results show no impact of predicted issue size, and a positive impact of unexpected issue size for large offerings, in line with the growth opportunities hypothesis. Further confirming this hypothesis, additional tests show that for straight debt unexpected issue size is linked to investment in growth opportunities. Overall, our findings indicate that stockholders use not only the type of security announced, but also the offering size, as a signal regarding the value of firms’ assets in place and growth opportunities. Our results highlight the importance of controlling for the endogeneity of issue size in event studies of security offering announcement effects. Previous studies focusing on the signaling content of security offering announcements tend to use

73 offering proceeds as a control variable, not taking into account its nature as a choice variable resulting in potential endogeneity issues, and obtain mixed results. Most studies show an insignificant impact of issue size on security offering announcement returns for equity (e.g., Jung et al. 1996; Autore et al., 2008), for convertible debt (e.g., Eckbo, 1986; Lewis et al., 2003), and for straight debt (e.g., Eckbo, 1986, Jung et al., 1996). Some studies find a significant negative impact of issue size on the announcement returns, again for all three security types (e.g., Bayless and Chaplinsky, 1996 for equity, Dutordoir and Van de Gucht, 2007 for convertible debt and Chaplinsky and Hansen, 1993 for straight debt,), whereas other studies show a significant positive impact (e.g., Hess and Bhagat, 1986 for equity and Hansen and Crutchley, 1990 for all three security types). Another goal of this study is thus to examine whether we can obtain more consistent results for all threes security types by taking into account the endogenous nature of security offering proceeds. Our study also extends the securities issuance literature by providing evidence on the determinants of security offering proceeds for different security types. The few existing empirical studies examining those determinants focus on a few selective determinants for individual security types (Chaplinsky and Hansen, 1993 and Lee and Masulis, 2009). The remainder of the paper continues as follows. The next section gives an overview of the literature and develops our research questions. Section 3 describes the dataset. Section 4 describes the methodology and control variables. Section 5 reports and discusses the empirical findings. Section 6 concludes the paper.

3.2. Research hypotheses We derive our testable predictions from three theoretical models on asymmetric information and securities issuance. Miller and Rock (1985) start from an asymmetric information framework and argue that firms’ dividend (financing) policy reveals unobserved earnings. In this model, they treat security issuance as a negative dividend. As such, larger securities offering sizes signal larger expected earnings shortfalls. The earnings shortfall hypothesis following from this model therefore predicts a negative impact of issue size on security offering announcement returns, irrespective of the security type. Krasker (1986), extending the Myers and Majluf (1984) signaling framework, develops an asymmetric information model in which equity offering proceeds provide a signal regarding the value of the issuing firms’ assets in place. He shows that firms with overpriced stock have greater incentives to choose larger offerings, raising equity-related adverse selection costs. His model leads to the overvaluation hypothesis predicting a negative impact of issue size on security

74 offering announcement returns. The negative impact should be stronger for more equity(- linked) security types. Ambarish et al. (1987) assume that prior to an offering there is asymmetric information either about firms’ assets in place or about firms’ growth opportunities. Asymmetric information about firms’ assets in place yields a negative seasoned equity offering announcement effect, similar to the predictions of Myers and Majluf (1984). However, when asymmetric information about growth opportunities dominates, seasoned equity issues can convey good news about growth opportunities, as raising capital to invest in high Net Present Value projects adds value to the firm, with larger issue sizes indicating more profitable growth options. The growth opportunities hypothesis derived from this model therefore predicts a positive impact of issue size on security offering announcement returns, the magnitude of which should be larger for equity(-linked) securities offerings. To our knowledge, these hypotheses have not yet been systematically examined in the literature. Empirical studies on security offering announcement returns do tend to include issue size, but mainly as a control variable without underlying theoretical motivation. Table 1 gives an overview of these studies. Some studies find a negative impact of issue size on security offering announcement returns (e.g., Chaplinsky and Hansen, 1993 for straight debt, and Bayless and Chaplinsky, 1996 and Demiralp et al., 2011 for equity), others find a positive impact (e.g., Hess and Bhagat, 1986 and Bethel and Krigman, 2004 for equity, and Hansen and Crutchley, 1990 for all three security classes), while most find an insignificant impact (e.g., Eckbo, 1986 for straight debt, Lewis et al., 2003 for convertible debt, and Autore et al., 2008 for equity). Our goal is to obtain more conclusive evidence by taking into account the endogeneity of issue size. This allows us to formally test the hypotheses of Miller and Rock (1985), Krasker, (1986), and Ambarish et al. (1987).

3.3. Data on security issues We conduct our study on a sample of seasoned equity, convertible debt, and straight debt offerings by US firms between January 1999 and December 2011. We download all security offerings from the SDC Global New Issues Database. In line with the literature, the sample excludes financial and utility companies and non-144a private placements. We also exclude mortgage and asset-backed bonds and secondary equity offerings. Finally, we manually remove all pass-through securities, rights offerings, unit offerings, convertible

75 and exchangeable bonds.4 We aggregate multiple tranches of the same bond offering by the same firm on the same day into one offering. Applying these criteria, leads to an initial sample of 3,364 seasoned equity, 1,426 convertible debt, and 12,699 straight debt offerings. From this sample, we retain all issues that meet the following criteria: (1) company accounts data are available from the Compustat Fundamentals Annual database; (2) stock price data are available from the Center for Research in Security Prices (CRSP); and (3) deal specific data are available from SDC. This leads to a final sample of 1,160 equity, 716 convertible debt and 1,951 straight debt issues. From these issues our sample contains 762, 497 and 702 unique firm observations for seasoned equity, convertible debt and straight debt respectively. Table 2 gives the number and percentages of equity, convertible debt, and straight debt issues per year. It also shows the average issue size per year scaled by market value of which we take the natural logarithm in the later analyses. The table shows that the frequency of security issues varies substantially over time for all three security classes, consistent with prior studies (e.g., Erel et al., 2012; Dutordoir and Hodrick, 2012). The number of seasoned equity issues moves cyclically with peaks in 2004 and between 2009 and 2011, whereas average issuance volume is more uniform. Conversely, issuance numbers are more uniform for convertible debt, whereas there are larger fluctuations for average issuance volumes. Finally, straight debt offerings peak in 2003 and 2009. In other years the number of offerings and average issuance volumes fluctuate substantially.

3.4. Research methodology 3.4.1. Controlling for the endogeneity of issue size Our goal is to measure the impact of issue size on security offering announcement returns, controlling for its potential endogeneity. We measure abnormal stock returns following a standard event study methodology (Brown and Warner, 1985), estimating the market model over the window (−300, −46) using the CRSP equally-weighted market index to proxy for the market return. We calculate cumulative abnormal returns (CARs) over the window (0, 1). We use this announcement return window to account for announcements occurring after the close of trade on day 0 (Lease et al., 1991 and Abhyankar and Dunning, 1999). For all security issues, we identify the announcement date as the filing date from SDC and for issues for which SDC gives no filing date as the issue date. We find average (median) announcement returns of −1.91% (−1.40%) for equity, −4.23% (−3.96%) for convertible debt

4 We keep shelf offerings in our sample, as this registration type has experienced substantial growth in recent years to even outnumber traditional registration types (Autore et al., 2008). Untabulated results show that our empirical results are robust to including a dummy to control for shelf registrations. 76 and −0.05% (−0.04) for straight debt offerings.5 Announcement returns are thus significantly smaller for convertible debt compared to both equity and straight debt offerings and significantly smaller for equity than for straight debt offerings. The highly negative announcement period return for convertible debt is probably a result of arbitrage-related short selling (de Jong et al., 2011). We estimate the impact of issue size on CARs using a regression model, = + + + (1)

ݖ݁ and௜ dependsݑ௜ on firm- and issue-specific݅ݏreturnݑ݁ ݏݏܫcumulative௜ ߙ௜ ߚଵ abnormalܺ௜ ߚଶ ܴܣܥwhere is the characteristics,ܥܣܴ the logarithm of total offering proceeds including over-allotment options, 6 , andܺ unobservable factors, . However, this model may be biased if correlates ݖ݁ factors, , affecting issueݑ size. ݑ݅ݏݑ݁ unobservableݏݏwithܫ To avoid such bias, weߝ control for endogeneity by including the first stage residuals from the issue size model in the second stage announcement return regressions. As issue size is continuous, we do not have a binary selection variable as demanded by the classical Heckman (1979) methodology. However, as pointed out by Li and Prabhala (2007) on page 46, in cases where we are able to observe the magnitude of the selection variable, i.e. having a continuous selection variable, we can observe the error term directly and do not have to take its expectation as we have to do in the 1/0 selection case. We can thus use the error term as the selection correction instead of including the inverse Mills ratio. Including the error term itself also offers one extra degree of freedom to work against collinearity issues arising from the selection control function or inverse Mills ratio being roughly linear in parts of its domain and thus potentially having very little variation with respect to the other variables in the second stage equation. Including the observed error term, however, introduces an independent (not conditional) source of variation in the selection correction term. Exclusion restrictions are thus not strictly necessary in this case. In the first stage, we model issue size by the following equation, estimated separately for seasoned equity, convertible and straight debt, = + + (2) the݅ݖ݁௜ logarithmߙ ߚ ofܺ௜ totalߝ௜ offering proceeds including over-allotment options ݏݑ݁isݏݏܫ where ݖ݁݅ݏݑ݁ݏݏܫ 5 t-statistics of –10.04 for equity, –16.78 for convertible debt, and –0.58 for straight debt indicate that CARs for equity and convertible debt are significantly different from zero. A non-parametric Rank test confirms this finding. 6 We use robust standard errors in all regression analyses. Following the security choice literature, we do not use standard errors clustered at firm level as it is common in studies on using panel data (Petersen, 2009). Security issuance is rare for firms, so that most firms only appear once in the sample. Confirming this argument, our results remain unchanged using standard errors clustered at firm level. 77 (in case of equity and convertible bond issues), and depends on firm- and issue-specific characteristics, , and unobservable factors, .

In the secondܺ stage, we include the fittedߝ residuals as a selectivity correction term in the announcement return equations. By adding this projectedߝෝప residual as a control function (or “auxiliary” regressor) in equation (1), it “absorbs” the correlation between the error term and the regressor. The resulting residual is now a well-behaved disturbance that is

,ݖ with݁ all other regressors in the CARݑෝప equation, including . Formally݅ݏݑ݁ݏݏܫuncorrelated ݖ݁ we again݅ݏݑ݁ whichݏݏܫ,the announcement return equation augmented by the first stage residuals estimate separately for the three security issue types is, = + + + + (3) thatݖ݁௜ thisߚଷ controlߝෝప ݑ௜ function approach and 2SLS ݅ݏݑ݁ݏshow ݏܫWooldridge௜ ߙ௜ ߚଵܺ (2009)௜ ߚଶ ܴܣܥImbens and instrumental variables analysis statistically lead to similar results.7 However, the advantage of the control function approach compared to a standard 2SLS approach is that the latter does not allow us to separate the effect of the residual issue size, captured by from that of predicted issue size, captured by . Separating out the effect of residual,ߝෝప unexpected issue size ݖ݁ argues that abnormal stock returns to corporate݅(1988) ݏݑ݁ݏݏܫis useful, since Acharya announcements should be only affected by unexpected news. It only takes simple algebra to see that, after controlling for residual issue size, the actual variable captures the

ݖ݁݅ݏݑ݁ݏݏܫ :effect of predicted offering proceeds, . That is (ଓݖ݁ + ) + + (4ݏ෣ݑ݁ݏ)ݏܫ + + = ଓݖ݁ప ߝෝప ߚଷߝෝప ݑ௜ݏݑ෣݁ݏݏܫ ௜ ߙ௜ ߚଵܺ௜ ߚଶܴܣܥ ,Or = + + ( ) + ( + ) + (5)

ଓ thatݖ݁ప it allowsߚଶ usߚଷ toߝෝప testݑ whether௜ the impact ofݏݑ෣݁ isݏݏܫ ௜ ߙ of௜ thisߚଵܺ methodology௜ ߚଶܴܣܥAnother advantage the self-selection correction term (i.e., the residual issue size) varies with different levels of the continuous endogenous variable (i.e., issue size) (Garen, 1984).8 In other words, this allows us to test whether the impact of managerial private information regarding issue size on security offering announcement returns varies with issue size. Formally, the announcement returns equation augmented by the interaction between the endogenous variable ( ) and the first stage residuals is: ௜ ݖ݁݅ݏݑ݁ݏݏܫ = + + + + × + (6)

ݖ݁௜ ߝෝప ݑ௜݅ݏݑ݁ݏݏܫݖ݁௜ ߚଷߝෝప ߚସ݅ݏݑ݁ݏݏܫ௜ ߙ௜ ߚଵܺ௜ ߚଶܴܣܥ

7 Consistent with theory we obtain similar results when using 2SLS in the later latter? analysis. 8 Garen (1984) shows that both 2SLS and 3SLS are inconsistent in this case, as the error term × does not converge to zero, and instrumental variables do not correct for an error term that has non-zero ݖ݁௜ ߝ௜Ƹ݅ݏݑ݁ݏݏܫ .expectation 78 3.4.2. Firm- and issue-specific characteristics To analyze the impact of issue size on security offering announcement returns, we first determine what factors drive the size of the offering. We use a large set of potential determinants of issue size. As these variables are publically available, stockholders should base their expectation on these characteristics. We group the variables in terms of funding needs, debt-related financing costs, equity-related financing costs, and variables that can proxy for both debt- and equity-related financing costs. Table 3 gives an overview and description of all the explanatory variables. According to the pecking order theory of Myers and Majluf (1984), firms first try to fund projects from retained earnings. If insufficient retained earnings are available, firms turn to the capital markets to raise debt and equity to fill the funding deficit. We measure the funding deficit following the methodology of Frank and Goyal (2003). They calculate the funding deficit as the sum of cash dividends, net investments and change in net working capital less internal cash flow. As the need for external capital is a positive function of the funding deficit, we predict a positive impact of funding deficit on issue size for all three securities. As funding deficit thus correlates with issue size, it fulfills the relevance condition to be included as an exclusion restriction in the first step issue size equation.9 The second exclusion condition is that it does not belong in the model being estimated, in our case the announcement return regressions. This is also fulfilled, as it is a logical consequence of firms having a funding deficit to raise external capital, which does not yield a stockholder reaction as such. Stockholders will only react to whether it is the optimal security choice and to whether issue size meets their expectations. Firms’ funding deficit thus only impacts announcement returns through issue size.10 We predict a negative impact of debt-related financing costs on offering size of debt(- linked) security types, as an increased cost of capital negatively impacts the amount of positive NPV investments and thus funding needs. Specifically, we control for firms’ profitability, income taxes, financial leverage, stock return volatility, and credit rating and for Treasury Bond yields. We include the ratio of earnings before interest and taxes (EBIT) to total assets, as higher earnings indicate a greater ability to pay interest on debt securities (Lewis et al., 1999). Income tax scaled by total assets is an inverse proxy for debt-related financing costs, as it captures the potential tax-shield resulting from the tax-deductibility of

9 Strictly speaking we do not need to include an exclusion restriction in the first stage issue size regression following the argument of Li and Prabhala (2007, p.46) stated in section 3.4.1. 10 When including funding deficit in the announcement return regressions for equity, convertible debt and straight debt, it does not have a significant regression coefficient. 79 interest payments (Modigliani and Miller, 1963). Leverage, the ratio of long-term debt to total assets, controls for financial distress costs and costs of underinvestment and asset substitution. Debt-related financing costs increase with stock return volatility, proxying for an increase in asymmetric information. We include an index of firms’ credit rating,11 as this is directly associated with the conditions under which the firm raises debt capital. Finally, Treasury Bond yields proxy for economy wide debt financing costs with higher yields making debt issuance more expensive. The adverse selection framework of Myers and Majluf (1984) motivates the equity- financing costs, which we expect to have a negative impact on issue size for equity(-linked) security types. They suggest that an equity offering may indicate the firm is overvalued, leading to a negative stock price reaction on the announcement date. To capture adverse selection, we include stock price run-up and firms’ trading volume. Stock-price run-up is an inverse proxy for equity-related financing costs. Lucas and McDonald (1990) argue that firms that have a greater stock price run-up before equity issuance experience less severe adverse selection problems, as a high stock price run-up signals high quality future investment projects. However, a high stock-price run-up may also be associated with equity overvaluation and proxy for high equity-related financing costs. We can thus not make a clear prediction for the impact. Trading volume measures stock liquidity. Butler et al. (2005) show that higher liquidity reduces adverse selection and thus SEO flotation costs. There are also factors that influence financing costs for both debt- and equity(-linked) security types, for which we also predict a negative impact on security offering size. We include total assets, market-to-book ratio, a leading indicator, and dummies taking the value one if a reputable underwriter brings the offering to the market and if the issue is made during hot market periods. Total assets proxy for asymmetric information and financial distress costs, both of which fall as firm size increases (MacKie-Mason, 1990). The market-to-book ratio may proxy for the availability of future growth opportunities and hence for lower external financing costs. But it can also proxy for opportunistic equity issues associated with higher financing costs, if it indicates equity overvaluation. We include the leading indicator for the US economy over the quarter preceding the issue month. As a measure of economy-wide growth opportunities, it proxies for firms’ financing needs, as companies invest in good states of the economy and divest in bad states. Security issues brought to the capital market by

11 Following de Jong et al. (2012) we apply a numerical credit rating transformation. We assign a value of one to a Standard & Poors (S&P) AAA rating and add one for each subsequent lower rating. If the S&P rating is unavailable, we use the Moody’s rating instead. For unrated firms, we assign the value of the lowest rating, as unrated firms should have the highest risk.

80 reputable underwriters have, on average, higher issue proceeds, better offer prices and more favorable announcement returns (e.g., Carter and Manaster, 1990; Fang, 2005). Issues underwritten by reputable investment banks should thus yield lower financing costs. Finally, several studies find that hot market conditions offer windows of opportunity for security issuance, as external financing costs are low (e.g., Bayless and Chaplinsky, 1996; Dutordoir and Van de Gucht, 2007). Table 4 reports descriptive statistics for the control variables for the three security samples. Panel B gives t-statistics for pairwise differences in means between the three securities. For our funding deficit proxy, we find larger deficits for equity issuing firms than for convertible and straight debt issuers. This is in line with the pecking order theory of Myers and Majluf (1984), which predicts that firms turn to equity issuance as a last resort to obtain funding. For debt-related financing costs, equity issuers have lower EBIT and pay lower income taxes than convertible bond issuers, which in turn have lower earnings and taxes than straight debt issuers. As both EBIT and income taxes proxy inversely for debt-related financing costs, these findings are consistent with our prediction. Not in line with our prediction is that both equity and convertible debt issuers have lower leverage than straight debt issuers. A possible explanation is that equity issuers are smaller than convertible debt and straight debt issuers and cannot take on higher leverage. Consistent with the theory that stock return volatility proxies for debt-related asymmetric information, we find larger stock return volatilities for equity than for convertible debt issuers and larger stock return volatilities for convertible debt than for straight debt issuers. Equity issuers have lower credit ratings than convertible debt issuers, which in turn have lower credit ratings than straight debt issuers. Finally, proxying for external debt financing costs, debt(-linked) securities are issued during times of higher Treasury bond yields, which is not consistent with our prediction. For equity-related financing costs, equity issuers have a larger stock price run-up than convertible issuers, which in turn have a larger stock price run-up than straight debt issuers. Both findings accord with predictions. Not in line with prediction is the smaller equity trading volume after convertible debt, for which the trading volume is again smaller than for straight debt issues. This may be a result of equity issuers being smaller, as indicated by total assets.12 For the variables controlling for both equity- and debt-related financing costs, companies issuing equity are smaller in terms of total assets than companies issuing convertible debt,

12 In line with this explanation is that equity issues have significantly smaller average total assets ($1,685m) than convertible bond ($3,338m) and straight debt issuers ($15,560m). 81 which in turn are smaller than companies issuing straight debt. Consistent with the notion that equity issues may be opportunistic, the market-to-book ratio is higher for equity issuers than for convertible and straight bond issuers. Results for the leading indicator show that equity(- linked) securities are issued more often than debt(-linked) securities during times of economic expansion. Choosing a reputable investment bank to underwrite the security issues happens less frequently for equity issues than for convertible debt issues, which in turn are less often underwritten by a reputable investment bank than are straight debt issues. Finally, for the hot markets dummy the results do not show a clear pattern. Equity issues occur less often during hot market periods than convertible debt issues, but both occur more often during hot market periods than straight debt issues.

3.5. Empirical results This section presents the results on the determinants of issue size and on the impact of issue size on security offering announcement returns controlling for endogeneity.

3.5.1. First stage results on security offering size We model issue size separately for each security using determinants proxying for funding needs and debt-related, equity-related, and general financing costs. This analysis serves as first step selection equation to control for the endogeneity of issue size in the second step announcement return regressions. Table 5 reports the results. Regression (1) reports the regression results for equity issue size. The results are largely consistent with our predictions. We find a positive impact of funding deficit on equity issue size. For the debt-related financing costs proxies, we find a negative impact of EBIT, which is expected as larger earnings reduce the need to raise external capital. Inconsistent with our predictions are the positive impact of income tax and the negative impact of leverage. As higher stock return volatility proxies for greater asymmetric information, its negative impact is in line with our expectations, as is the positive impact of lower credit ratings. The reason for the latter is that companies with lower credit ratings find it more difficult to raise debt capital and turn to equity financing instead. As higher Treasury Bond yields make debt financing more expensive, the positive impact on equity issue size is in line with expectations. For equity-related financing costs, we find, as predicted, a positive impact of stock price run- up and trading volume on issue size. The positive impact of stock price run-up is a first indication of the overvaluation hypothesis, as it indicates that firms make larger issues when stock prices are high or potentially overvalued. For general financing costs, larger firms, in

82 terms of total assets, obtain larger proceeds, as do issues underwritten by a reputable underwriter. The positive impact of the market-to-book ratio may indicate that equity capital is raised out of opportunism or to fund firm growth. Finally, in times of economic expansion firms raise more equity capital, as shown by the positive impact of the leading indicator. Column two of table 5 presents the results on the determinants of convertible debt issue size. As for equity, we find a positive impact of firms’ funding deficit and EBIT. Income tax is significantly positive, consistent with our predictions, as issuance of a debt(-linked) security and thus future debt interest payments reduce firms’ tax burdens. Higher asymmetric information, proxied by stock return volatility, has a negative impact on convertible debt issue size, whereas higher Treasury Bond yields have a positive impact. For equity-related financing costs we find, as expected, a positive impact of trading volume controlling for stock liquidity. Results on general financing costs show a positive impact of total assets, market-to- book ratio, the leading indicator and a dummy taking the value one if the issue is underwritten by a reputable underwriter. All of these results are in line with expectations. Results on debt issue size in column (3) show, as for the other two security types, a positive impact of funding deficit. For the debt-related financing costs, we find a positive impact of income tax confirming the argument of the tax advantage of debt. Not in line with our predictions is the positive impact of the credit rating index, as we expect firms with lower credit ratings to obtain smaller proceeds. The negative impact of Treasury Bond yields confirms the notion that times with higher Treasury Bond yields increase external debt financing costs. Conversely, the positive impact of trading volume on debt issue size is somewhat unexpected, but may be due to trading volume proxying for firm size. For the general financing cost variables, we find a positive impact of firm size and a negative impact of the market-to-book ratio. The latter hints of higher market-to-book ratios capturing firm overvaluation more than growth opportunities. The negative impact of the leading indicator is unexpected, as times of economic expansion should yield higher issue proceeds, as there is a larger demand for funding. However, stock market levels positively correlate with economic expansions, potentially opening a window of opportunity for equity issuance. Finally, in line with expectations, we find a positive impact of reputable underwriters and hot debt markets on debt issue size. In conclusion, we find support for our hypothesis that a larger funding deficit leads to larger security offering proceeds. Moreover, results mostly support the hypotheses regarding a negative impact of debt-related, equity-related, and general financing costs on issue size. Finally, funding deficit, which serves as an exclusion restriction for our second stage analysis

83 is statistically significant for each security type.

3.5.2. Endogeneity of issue size and security offering announcement returns Tables 6 to 8 present the results of regressions of seasoned equity, convertible, and straight debt offering announcement returns (CARs) on issue sizes and control variables.13 Regressions in column (1) include a selectivity corrections term, reflecting the fitted residual issue size from the first-stage regressions. In column (2) we augment the regressions by including the interaction between the selectivity correction term and issue size. This allows us to test whether the impact of managerial private information regarding issue size on announcement returns varies with offering size. Column (3) controls for the endogeneity of the security choice decision. We follow Dutordoir and Hodrick (2012) and add a selectivity correction term obtained from an ordered probit model estimating the security choice between seasoned equity, convertible debt, and straight debt.14 Finally, column (4) displays the results that would be obtained without controlling for endogeneity by adding the self-selection correction term.

3.5.2.1. Seasoned equity Table 6 presents the results of a cross-sectional analysis of seasoned equity announcement returns. As in previous studies of security offering announcement returns (e.g., Altinkilic and Hansen, 2003; Autore et al. 2008; Dutordoir and Hodrick, 2012), the regression has a low R2, which is probably attributable to the high noise to signal ratio in daily abnormal stock returns. Results in column (1) show that actual issue size has a positive impact on equity announcement returns, which is in line with the growth opportunities hypothesis. By contrast, the issue size selectivity correction term has a negative impact on equity announcement returns. Its coefficient captures stockholders’ updated beliefs regarding managerial private information driving offering sizes. The finding of a negative impact may therefore indicate that the market perceives high unexpected equity offering sizes as a signal of overvaluation. However, a negative coefficient on the selectivity correction term may also capture stockholders’ beliefs on managerial private information about a shortfall in earnings. We

13 In addition to these variables, proxies for corporate governance quality may also have a significant impact on security offering announcement returns. 14 Results in the Appendix show a negative impact of debt-related financing costs and a positive impact of equity-related financing costs on the likelihood to issue straight debt over convertible debt and seasoned equity. The regression R2 is 29.71. As an exclusion restriction we use the number of rival 2-digit industry issues for the specific security type in the year before issuance. When including residuals from two distinct models first stage models, i.e. the residuals from both the issue size and security choice model, we follow Fishe et al. (1981) and Carson (2007).

84 disentangle these two interpretations in further analyses. The results on the control variables show a positive impact of EBIT and leverage. The latter confirms our prediction that an equity issue can alleviate financial distress resulting from high leverage. Stock return volatility and the credit rating index proxy for the level of asymmetric information and riskiness of the issuer. Stock markets react less favorably when firms with high levels of asymmetric information come to the market, which is confirmed by the negative impact of these two variables. The negative impact of Treasury Bond yields and total assets is somewhat unexpected. A possible explanation for the latter is that larger firms could have issued debt security types instead. Finally, we find a negative impact of the market-to-book ratio. As this can proxy for firm overvaluation, this result is in line with our expectation and hints of equity issues being made opportunistically. In column (2) we find the interaction term between the selectivity correction term and issue size to be insignificant, suggesting that the negative signal of managerial private information regarding firm overvaluation or a shortfall in earnings does not lead to more negative announcement returns for larger security offerings. Results are unchanged for the other control variables. The results in column (3) remain mostly unchanged. The issue size selectivity correction term keeps its negative impact. The security choice selectivity correction term itself has a significant negative coefficient. As the generalized residuals for equity are by construction negative, this implies that unobservable variables increasing firms’ likelihood to opt for equity financing lead to higher stockholder reactions to seasoned equity announcements. This finding suggests that markets consider the security choice decision of these firms to offer equity as a positive signal. The bad signal of equity offerings is thus only related to managerial private information regarding larger than expected issue size, as it suggests firm overvaluation or a future shortfall in earnings. The results in column (4) differ to some extent from the previous results. EBIT and leverage lose their positive impact, whereas a higher stock price run-up before issuance now has a positive impact. Issue size itself is not statistically significant in this model, which can potentially be explained by the positive impact of expected and the negative impact of unexpected issue size neutralizing each other. This confirms the importance of controlling for the endogeneity of issue size and may be an explanation for previous studies mostly finding an insignificant impact of issue size on equity announcement returns (table 1).

85 3.5.2.2. Convertible debt Table 7 reports the results of the CAR analysis for convertible bond issues. In column (1), actual issue size has a significantly positive coefficient, as it has for equity. The issue size selectivity correction term from the convertible debt issue size regression has a significant negative coefficient. As with equity, this is consistent with both the earnings shortfall and the overvaluation hypothesis. Consistent with predictions, we furthermore find a positive impact of EBIT on convertible debt announcement returns, as higher earnings make it easier to repay debt interest and a negative impact of income taxes. Convertible debt can be an instrument for flexible capital structure adjustment, allowing firms to reduce financial leverage if financial distress costs are high. The positive impact of leverage on convertible debt announcement returns is consistent with this. As for seasoned equity, the negative impact of total assets is unexpected, as larger firms have less asymmetric information, but again stockholders may have expected a straight debt issue for these firms. Confirming the overvaluation argument for equity(-linked) securities, the market-to-book ratio has a negative impact on convertible debt announcement returns. Finally, inconsistent with our predictions, having an issue underwritten by a reputable investment bank has a negative impact on the market reaction. As for equity, results in column (2) show that the interaction effect between the issue size selectivity correction term and issue size is insignificant. However, the selectivity correction term itself remains significant negative. For the other control variables the results remain unchanged. Results in column (3) confirm the negative impact of the issue size selectivity correction term on announcement returns. The security choice selectivity correction term has a negative impact. As the generalized residuals from the ordered probit model are by construction positive for convertible debt, this finding is inconsistent with security issuers self-selecting into the security offering that yields the most favorable market reaction (Dutordoir and Hodrick, 2012). In column (4), we find that actual issue size keeps its positive impact. Income tax, firm size and the market-to-book ratio are insignificant, stock return volatility is significantly negative and trading volume and the leading indicator have a positive impact on announcement returns.

3.5.2.3. Straight debt In table 8 we analyze straight debt announcement returns controlling for the endogeneity

86 of issue size. Column (1) shows an insignificant impact of actual issue size and the selectivity correction term from the debt issue size regressions. We find a negative impact of financial leverage on announcement returns, which is consistent with further debt offerings increasing financial distress costs. The positive impact of stock price run-up may indicate that stockholders react positively to a debt issue in these situations, as it signals firm undervaluation. Again this is consistent with equity(-linked) issues being made opportunistically. Finally, the negative impact of the market-to-book ratio is not in line with expectations. Column (2), however, shows a positive impact of the interaction between the issue size selectivity correction term and issue size, in line with the growth opportunities hypothesis. The selectivity correction term itself remains insignificant. This suggests that for straight debt the magnitude of the signal matters, i.e., the positive stock holder reaction increases with issue size. For the other control variables the results are unchanged. Results in column (3) remain unchanged from column (1). Finally, in column (4) we find a significantly negative coefficient of actual issue size. Leverage and the market-to-book ratio still have a negative impact and stock price run-up has a positive impact.

3.5.3. Post issue earnings and use of proceeds To disentangle the two hypotheses, i.e., earnings shortfall and overvaluation, that may explain the negative impact of the issue size selectivity correction term on equity and convertible debt announcement returns and to provide further evidence for the growth opportunities hypothesis, we test whether firms’ earnings decrease after issuance and examine the use of proceeds from the security offerings. If the negative impact of the issue size selectivity correction term relates to managerial private information regarding a future shortfall in earnings, we should observe a negative relation between the first stage residuals from the issue size regressions and future earnings. We partition observations by the first stage residuals into quartiles and compare the quartile averages of firms’ operating income before depreciation and amortization scaled by total assets for the two years after issuance. Results in table 9 show insignificant differences for the quartile averages for both years after issuance and average earnings for the two years together. Moreover, the overall level of earnings remains constant for both years after issuance. This suggests that the negative impact of the issue size selectivity correction on equity and convertible debt announcement returns is not related to an earnings shortfall. Second, we analyze for what corporate purpose firms use the proceeds from the security

87 offering. We follow the methodology of Kim and Weisbach (2008) and examine balance sheet and cash flows items for four years after issuance to establish whether firms use the capital raised to fund direct investments or whether they use it for other purposes, i.e., to increase cash holdings.15 In the latter case, management raises more funds than needed for investment activities and, as raising capital is expensive, must have opportunistic motives such as firm overvaluation. We measure direct investment as changes in capital expenditures, acquisitions, R&D, inventory and total assets. We calculate changes in cash holdings, inventory and total assets as, ( )/ + 1 (7)

,଴൯ fiscal൧ year end before the offeringݏݐ the݁ݏݏwhere V is the variable being݈݊ൣ൫ measured,ܸ௧ − ܸ଴ 0ݐ݋indicatesݐ݈ܽܽ and t is the number of years after year 0. For capital expenditures, acquisitions and R&D we sum values including the respective year after issuance and calculate / ) + 1] (8) ௧ ଴ issue size or the face value of residualݏݐ݁ݏactual ݏWe regress each of these variables݈݊[(∑௜ୀଵ onܸ௜ theݐ݋ logݐ݈ܽ ofܽ issue size over total assets plus one, the log of other sources over total assets plus one,16 the log of total assets, and year fixed effects. We test whether large security offerings are related to growth-related uses of proceeds, explaining the positive market reaction to actual issue size for equity and convertible debt. If the growth opportunities interpretation holds, we expect a positive impact of actual issue size on variables capturing direct investment for all three security types. Table 10 reports the results. Furthermore, if the overvaluation hypothesis holds and larger than expected equity(- linked) issues are made by firms exploiting overvalued stock, we expect to find a positive relation between residual issue size and changes in cash holdings in the years after issuance for both equity and convertible debt issues. By contrast, given the finding of a positive impact of residual issue sizes for straight bonds, we expect a positive relation between straight debt offering residual sizes and growth-related uses of proceeds. Table 10 reports the results. We find a positive impact of residual issue size on an increase in cash holdings in the two years after issuance for both equity and convertible debt and an insignificant impact for straight debt. For the variables measuring firms’ direct investments in the years after issuance,

15 We do not analyse stated uses of proceeds as these are mostly vague, i.e. 81% of the equity issuing firms, 92% of convertible debt issuing firms and 43% of the straight debt issuing firms state “General Corporate Purposes” as the primary use of proceeds. 16 Other sources are defined as the difference between total sources (funds from operations, sale of property, plant and equipment, long-term debt issuances, and sale of common and preferred stock) and the proceeds raised with the respective issue. 88 namely the sum of capital expenditures, acquisitions and R&D, and the growth in inventory and total assets, we find an insignificant impact of residual issue size for equity with the exception of a positive impact on acquisitions and total assets. The result for total assets may be an automatic effect of further capital adding to total assets. For convertible debt, coefficient signs are more mixed, but do not suggest a significant impact on growth. This picture changes for straight debt, for which residual issue size has a positive impact on direct investment proxies (except for inventory) for almost all four years. Together these results provide further evidence for the overvaluation and positive growth opportunities hypotheses, as they suggest firms use the residual, unexpected portion of equity(-linked) issues to increase cash holdings, whereas they use the residual, unexpected part of debt issues to fund direct investments. The results for actual issue size show a positive impact on increases in direct investments for all years and all three security types. This supports the notion that large offering sizes serve as a signal of growth opportunities for equity and convertible debt. For straight debt issues, however, despite a positive impact of larger issue size on increases in direct investments, we do not find a favorable market reaction in our announcement return regressions. This finding may be a result of the partial anticipation of debt issues from pre- issue disclosures obscuring the economic impact of the issue (Chaplinsky and Hansen, 1993). Moreover, our results show that firms tend to use equity, convertible debt, and straight debt proceeds to finance R&D expenses. This result extends the findings of Brown et al. (2009) that young firms tend to finance R&D solely through equity and internal cash flows. We find that firms seem to use a range of security types to finance R&D. In conclusion, our analysis of the uses of security offering proceeds corroborates the notion that security offering proceeds signal the value of firms’ assets in place for equity and convertible debt and signal firms’ growth opportunities for all three security types. For the latter, we show that larger offering size itself relates to investments in firm growth and thus signals valuable investment opportunities to the markets. This is particularly the case for straight debt as managerial private information signals valuable growth opportunities for this security type.

3.6. Conclusion We examine the determinants of issue size for seasoned equity, convertible debt, and straight debt offerings by U.S. firms between 1999 and 2011 and study the impact of issue size on announcement returns of offerings of these security classes, controlling for endogeneity. In particular, we test the theoretical implications of the models of Miller and

89 Rock (1985), Krasker (1986) and Ambarish et al. (1987). Our main findings are as follows. We find a positive impact of firms’ funding needs on issue size for all security types and a negative impact of debt- and equity-related financing costs on issue size of more debt- and equity(-linked) securities respectively. For the impact of issue size on security offering announcement returns, we find positive impact of predicted issue size on announcement returns for both equity and convertible debt. Further tests on firms’ use of proceeds show that larger equity(-linked) offerings are more likely to be used to finance growth than to accumulate cash. We find a negative impact of unpredicted issue size on announcement returns for equity and convertible debt offerings. This finding supports the overvaluation hypothesis, but also partly the earnings shortfall hypothesis. However, for straight debt we find a positive impact of managerial private information for very large issues, which is consistent with the growth opportunities hypothesis and inconsistent with the earnings shortfall hypothesis. We conduct additional tests to disentangle the hypotheses. Specifically, we analyze firms’ earnings in the years after issuance and firms’ use of proceeds. We show that larger than expected issue size is not related to an expected shortfall in earnings, whereas our analysis of the use of proceeds suggests that firms use larger than expected equity(-linked) issues to build up financial slack. Moreover, we show that debt issuers use larger than expected proceeds to fund firm growth. These results provide further evidence in support of the overvaluation hypothesis for equity and convertible debt and in support of the growth opportunities hypothesis for straight debt. Finally, there are two main implications our study. First, our findings suggest that event studies on the announcement effects of securities issues should control for the endogeneity of issue size. Second, firms’ management should be careful in using their discretion in determining security offering proceeds, as market may interpret issue sizes beyond those expected as a negative signal for certain security types, i.e. equity and convertible debt. A potential limitation of our study is that we do not control for concurrent stock repurchases which may affect issue proceeds. This could potentially affect convertible bond issue size, as De Jong et al. (2011) point out that roughly 24% of the convertible debt issues over the period 2005 to 2007 were combined with stock repurchases.

90 Appendix: Security choice between seasoned equity, convertible debt and straight debt

We estimate the security choice between seasoned equity, convertible debt and straight debt using an ordered probit model. The dependent variable is categorical taking the value one for equity, the value two for convertible debt and the value three for straight debt. Moreover, as theory predicts higher debt-related financing costs for equity-type securities, and higher equity-related adverse selection costs for debt-type securities, we follow the security choice literature and use the same variables that determine issue size as determinants in our first stage security choice model (e.g., Marsh, 1982; Bayless and Chaplinsky, 1991; Lewis et al., 1999). As an exclusion restriction we use the number of rival 2-digit industry issues for the specific security type in the year before issuance. The results below show a negative impact of debt-related financing costs and a positive impact of equity-related financing costs on the likelihood to issue straight debt over convertible debt and seasoned equity. The regression’s R2 is 29.71. Variables (1) Intercept 1 1.06*** (0.00) Intercept 2 1.85*** (0.00) EBIT 0.74*** (0.00) Income tax 1.88** (0.02) Leverage 0.12 (0.25) Stock return volatility −3.99*** (0.00) Credit rating (index) −0.05*** (0.00) TB Yield 0.02 (0.11) Stock price run-up −20.66*** (0.00) Trading volume −0.00*** (0.00) Total assets (log) 0.36*** (0.00) Market-to-book ratio −0.01** (0.03) Leading indicator 1.24 (0.28) Reputable underwriter (dummy) 0.12**

91 (0.02) Hot markets −0.41*** (0.00) Rivalsecurityissuance 0.01** (0.01)

R 2 29.71 N 3827 This table reports coefficients and p-values from ordered probit regressions of the security choice between seasoned equity, convertible debt and straight debt offerings. Table 3 gives the definition and source of all variables. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels.

Following Dutordoir and Hodrick (2012), we obtain generalized residuals from this model to control for firms self-selecting into a particular security type in our second stage announcement return regressions. As we are now including residuals from two distinct first stage models, i.e. the residuals from both the issue size and security choice model, we follow the methodology of Fishe et al. (1981) and Carson (2007). Fishe et al (1981) analyze the impact of women’s choice to obtain college education and to join the labor market on the earnings of women. Both the choice to obtain college education and to join the labor market are measured by first stage probit models, which residuals are then included in the second stage earnings equation. Carson (2007) uses a structural model to measure the impact of ex ante and ex post control on supplier performance. In a first step he estimates both ex ante and ex post control and includes the residuals from both models in his second stage supplier performance model.

92 References

Abhyankar, A. and Dunning, A., 1999. Wealth effects of convertible bond and convertible preference share issues: An empirical analysis of the UK market. Journal of Banking and Finance 23, 1043–1065. Acharya, S., 1988. A generalized econometric model and tests of a signaling hypothesis with two discrete signals. Journal of Finance 43, 413–429. Altinkilic, O. and Hansen, R.S., 2003. Discounting and underpricing in seasoned equity offers. Journal of Financial Economics 69, 285–323. Ambarish, R., Kose, J. and William, J., 1987. Efficient signalling with dividends and investments. Journal of Finance 42, 321–343. Ammann, M., Fehr, M. and Seiz, R., 2006. New evidence on the announcement effect of convertible and exchangeable bonds. Journal of Multinational Financial Management 15, 43–63. Asquith, P. and Mullins, D.W. Jr., 1986. Equity issues and offering dilution. Journal of Financial Economics 15, 61–89. Autore, D.M., Kumar, R. and Shome, D.K., 2008. The revival of shelf-registered corporate equity offerings. Journal of Corporate Finance 14, 32–50. Bayless, M. and Chaplinsky, S., 1991. Expectation of security types and the information content of debt and equity offers. Journal of Financial Intermediation 1, 195–214. Bayless, M. and Chaplinsky, S., 1996. Is there a window of opportunity for seasoned equity issuance? Journal of Finance 51, 253–278. Bethel, J.E. and Krigman, L., 2004. Unallocated shelf registrations. Why doesn’t everybody use it? Equity issuance, managerial choice, and market reaction. Working Paper. Bhagat, S., Marr, M.W. and Thompson, G.R., 1985. The rule 415 experiment: equity markets. Journal of Finance 40, 1385–1401. Brown, J.R., Fazzari, S.M. and Petersen, B.C., 2009. Financing innovation and growth: Cash flow, external equity and the 1990s R&D boom. Journal of Finance 64, 151–185. Brown, S. and Warner, J.B., 1985. Using daily stock returns: the case of event studies. Journal of Financial Economics 14, 3–31. Butler, A.W., Grullon, G. and Weston, J.P., 2005. Stock market liquidity and the cost of issuing equity. Journal of Financial and Quantitative Analysis 40, 331–348. Burlacu, R., 2000. New evidence on the pecking order hypothesis: the case of French convertible bonds. Journal of Multinational Financial Management 10, 439–459.

93 Carson, S.J., 2007. When to give up control of outsourced new product development? Journal of Marketing 71, 49–66. Carter, R.B. and Manaster, S., 1990. Initial public offerings and underwriter reputation. Journal of Finance 45, 1045–1067. Chang, S., Chen, S. and Liu, Y., 2004. Why firms use convertibles: A further test of the sequential-financing hypothesis. Journal of Banking and Finance 28, 1163–1183. Chaplinsky, S. and Hansen, R.S., 1993. Partial anticipation, the flow of information and the economic impact of corporate debt sales. Review of Financial Studies 6, 709–732. De Jong, A., Dutordoir, M. and Verwijmeren, P., 2011. Why do convertible issuers simultaneously repurchase stock? An arbitrage-based explanation. Journal of Financial Economics 100, 113–129. De Jong, A., Dutordoir, M., van Genuchten, N. and Verwijmeren, P., 2012. Convertible arbitrage price pressure and short-sale constraints. Financial Analysts Journal 68, 70–88. Denis, D.J., 1991. Shelf registration and the market for seasoned equity offering. Journal of Business 64, 189–212. Denis, D.J., 1994. Investment opportunities and the market reaction to equity offerings. Journal of Financial and Quantitative Analyis 29, 159–177. Demiralp, I., D’Mello, R., Schlingemann, F.P. and Subramaniam, V., 2011. Are there monitoring benefits to institutional ownership? Evidence from seasoned equity offerings. Journal of Corporate Finance 17, 1340–1359. Dutordoir, M. and Hodrick, L.S., 2012. Self-selection and stock returns around corporate security offerings announcements. Working Paper. Dutordoir, M. and Van de Gucht, L., 2007. Are there windows of opportunity for convertible debt issuance? Evidence for Western Europe. Journal of Banking and Finance 31, 2828– 2848. Eckbo, B.E., 1986. Valuation effects of corporate debt offerings. Journal of Financial Economics 15, 119–151. Erel, I., Julio, B., Kim, W. and Weisbach, M.S., 2012. Macroeconomic conditions and capital raising. Review of Financial Studies 25, 341–376. Fang, L.H., 2005. Investment bank reputation and the price and quality of underwriting services. Journal of Finance 60, 2729–2761. Fishe, R.P.H., Trost, R.P. and Lurie, P.M., 1981. Labor force earnings and college choice of young women: An examination selectivity bias and comparative advantage. Economics of Education Review 2, 169–191.

94 Frank, M.Z. and Goyal, V.K., 2003. Testing the pecking order theory of capital structure. Journal of Financial Economics 67, 217–248. Gajewski, J-F. and Ginglinger, E., 2002. Seasoned equity issues in a closely held market: Evidence from France. European Finance Review 6, 291–319. Garen, J., 1984. The returns to schooling: A selectivity bias approach with a continuous choice variable. Econometrica 52, 1199–1218. Hansen, R.S. and Crutchley, C., 1990. Corporate earnings and financings. An empirical analysis. Journal of Business 63, 347–371. Heckman, J.J., 1979. Sample selection bias as a specification error. Econometrica 47, 153– 161. Hess, A.C. and Bhagat, S., 1986. Size effects of seasoned stock issues: Empirical evidence. Journal of Business 59, 567–584. Hovakimian, A., Opler, T. and Titman, S., 2001. The debt-equity choice. Journal of Financial and Quantitative Analysis 36, 1–24. Imbens, G.W. and Wooldridge, J.M., 2009. Recent developments in the econometrics of program evaluation. Journal of Economic Literature 47, 5–86. Johnson, S.A., 1995. Dividend payout and the valuation effects of bond announcements. Journal of Financial and Quantitative Analysis 30, 407–423. Jung, K., Kim, Y. and Stulz, R.M., 1996. Timing, investment opportunities, managerial discretion and the security issue decision. Journal of Financial Economics 42, 159–185. Kang, J-K. and Stulz, R.M., 1996. How different is Japanese corporate finance? An investigation of the information content of new security issues. Review of Financial Studies 9, 109–139. Kim, Y. and Stulz, R.M., 1992. Is there a global market for convertible bonds? Journal of Business 65, 75–91. Kim, W. and Weisbach, M.S., 2008. Motivations for public equity offers: An international perspective. Journal of Financial Economics 87, 281–307. Krasker, W.S., 1986. Stock price movements in response to stock issues under asymmetric information. Journal of Finance 41, 93–106. Korajczyck, R.A., Lucas, D.J. and McDonald, R.L., 1990. Understanding stock price behavior around the time of equity issues, in Hubbard, R., (ed.): Asymmetric information, corporate finance and investment, University of Chicago Press, Chicago, Illinois. Lease, R.C., Masulis, R.W. and Page, J.R., 1991. An investigation of market microstructure impacts on event study returns. Journal of Finance 46, 1523–1536.

95 Lee, G. and Masulis, R.W., 2009. Seasoned equity offering: Quality of accounting information and expected flotation costs. Journal of Financial Economics 92, 443–469. Lewis, C.M., Rogalski, R.J. and Seward, J.K., 1999. Is convertible debt a substitute for straight debt or for common equity? Financial Management 28, 5–27. Lewis, C.M., Rogalski, R.J. and Seward, J.K., 2003. Industry conditions, growth opportunities and market reactions to convertible debt financing decisions. Journal of Banking and Finance 27, 153–181. Li, K. and Prabhala, N.R., 2007. Self-selection models in corporate Finance, in Eckbo, B.E., (ed.): Handbook of corporate finance: Empirical corporate finance, Vol. 1, Handbook of finance series, North-Holland, Elsevier. Lopez-de-Silanes, F., Shleifer, A. and Vishny, R.W., 1997. Privatization in the United States. Rand Journal of Economics 28. Lucas, D. and McDonald, R., 1990. Equity issues and stock price dynamics. Journal of Finance 45, 1019–1043. Marsh, P., 1982. The choice between equity and debt: an empirical study. Journal of Finance 37, 121–144. McKie-Mason, J., 1990. Equity issues and stock price dynamics. Journal of Finance 45, 1019–1045. Mikkelson, W.H. and Partch, M.M., 1986. Valuation effects of security offerings and the issuance process. Journal of Financial Economics 15, 31–60. Miller, M.H. and Rock K., 1985. Dividend policy under asymmetric information. Journal of Finance 40, 1031–1051. Modigliani, F. and Miller, M.H., 1963. Corporate income taxes and the cost of capital: A correction. American Economic Review 53, 433–443. Myers, S. and Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13, 187– 221. Petersen, M.A., 2009. Estimating standard errors in finance panel datasets: Comparing approaches. Review of Financial Studies 22, 435–480. Shyam-Sunder, L., 1991. The stock price effect of risky versus safe debt. Journal of Financial and Quantitative Analysis 26, 549–558. Slovin, M.B., Sushka, M.E. and Bendeck, Y.M., 1994. Seasoned common stock issuance following an IPO. Journal of Banking and Finance 18, 207–226.

96 Table 1 Summary of previous empirical results on the impact of issue size on seasoned equity, convertible debt, and straight debt announcement returns

This table summarizes empirical results from previous studies on the impact of issue size on equity, convertible debt and straight debt announcement returns. If a study reports mixed results, all findings are reported. Study Security type Market Period Impact of issue size Bhagat et al. (1985) Equity USA 1982–1983 Insignificant Asquith and Mullins (1986) Equity USA 1963–1981 Negative / insignificant Eckbo (1986) Straight and convertible debt USA 1964–1981 Insignificant Hess and Bhagat (1986) Equity USA 1963–1978 Positive / insignificant Mikkelson and Partch (1986) Equity and convertible and straight debt USA 1972–1982 Negative / insignificant Hansen and Crutchley (1990) Equity and convertible and straight debt USA 1975–1982 Positive / insignificant Korajczyk et al. (1990) Equity USA 1978–1983 Negative / insignificant Denis (1991) Equity USA 1982–1986 Insignificant Shyam-Sunder (1991) Straight debt USA 1980–1984 Insignificant Kim and Stulz (1992) Convertible debt USA 1970–1984 Insignificant Chaplinsky and Hansen (1993) Straight debt USA 1974–1984 Negative Denis (1994) Equity USA 1977–1990 Insignificant Slovin et al. (1994) Equity USA 1973–1988 Insignificant Johnson (1995) Straight debt USA 1977–1983 Negative / insignificant Bayless and Chaplinsky (1996) Equity USA 1968–1990 Negative Jung at al. (1996) Equity and straight debt USA 1977–1984 Insignificant Kang and Stulz (1996) Equity and convertible and straight debt Japan 1985–1991 Insignificant Abhyanker and Dunning (1999) Convertible debt UK 1982–1996 Insignificant Burlacu (2000) Convertible debt France 1981–1998 Insignificant Gajewski and Ginglinger (2002) Equity France 1986–1996 Negative Altinkilic and Hansen (2003) Equity USA 1990–1997 Insignificant Lewis et al. (2003) Convertible debt USA 1978–1992 Insignificant Bethel and Krigman (2004) Equity USA 1992–2001 Positive / insignificant Chang et al. (2004) Convertible debt Taiwan 1990–1999 Insignificant Ammann et al. (2006) Convertible debt Europe 1996–2003 Insignificant Dutordoir and Van de Gucht (2007) Convertible debt Europe 1990–2002 Negative / insignificant Autore et al. (2008) Equity USA 1990–2003 Insignificant Demiralp et al. (2011) Equity USA 1982–1985 Negative

97 Table 2 Descriptive statistics for security issues

The number and percentage of issues and average issue size (issue size/market value) by year for samples of seasoned equity, convertible debt, and straight debt issues by U.S. industrial companies between January 1999 and December 2011. Data are from the SDC Global New Issues Database. The equity sample is 1,160 issues, the convertible debt sample is 716 issues, and the straight debt sample is 1,951 issues. Seasoned equity Convertible debt Straight debt Number of Percentage Average issue Number of Percentage Average issue Number of Percentage Average issue Issue year issues size ($m) issues size ($m) issues size ($m) 1999 46 0.04 148 18 0.03 299 112 0.06 431 2000 66 0.06 205 27 0.04 494 80 0.04 535 2001 38 0.03 109 73 0.10 433 188 0.10 530 2002 65 0.06 168 31 0.04 381 190 0.10 331 2003 88 0.08 138 142 0.20 252 215 0.11 374 2004 121 0.10 138 96 0.13 215 146 0.08 420 2005 93 0.08 151 39 0.05 276 118 0.06 434 2006 83 0.07 149 58 0.08 540 93 0.05 630 2007 83 0.07 188 69 0.10 475 139 0.07 837 2008 49 0.04 285 39 0.05 363 119 0.06 1,026 2009 179 0.15 164 52 0.07 289 224 0.11 795 2010 125 0.11 128 34 0.05 368 185 0.09 831 2011 124 0.11 112 38 0.05 285 142 0.07 1,015 Total 1,160 100 160 716 100 359 1,951 100 630

No. of firms 762 497 702

98 Table 3 Variable descriptions

This table defines all variables used in the paper and their data sources. All variables are measured at the end of the fiscal year before the security offering, unless noted otherwise. Variable Definition Source ∑Acquisition Logarithmic transformation of the sum of all Compustat acquisition expenses up to and including the respective year after issuance for four years after issuance over total assets in the year before issuance plus one

∑CAPEX Logarithmic transformation of the sum of all Compustat capital expenditures up to and including the respective year after issuance for four years after issuance over total assets in the year before issuance plus one

ΔCash Logarithmic transformations of cash for each of Compustat the four years after issuance minus the value in the year before issuance scaled by total assets in the year before issuance plus one

Credit rating (index) Index following de Jong et al. (2012) assigning a Compustat value of one to an S&P AAA rating and adding one for each subsequent lower rating EBIT Earnings before interest and taxes scaled by total Compustat assets

Funding deficit Following the methodology of Frank and Goyal Compustat (2003), defined as the sum of cash dividends, net investment and change in working capital less the internal cash flow

Hotmarkets Issuancevolumeoftherespectivesecurityinthe SDC month before the offering date over issuance volume in months –4 to –2 before issuance

Income tax Income taxes divided by total assets Compustat

ΔInventory Logarithmic transformations of inventory for Compustat each of the four years after issuance minus the value in the year before issuance scaled by total assets in the year before issuance plus one

Issue size Logarithm of total proceeds SDC

Leading indicator Logarithmic growth in the composite leading Datastream indicator for the US economy over months –4 to –1 before issuance

Leverage Long-term debt over total assets Compustat

∑R&D Logarithmic transformation of the sum of all Compustat R&D expenses until and including the respective year after issuance for four years after issuance over total assets in the year before issuance plus one

99 Table 3 (continued)

Reputable underwriter A dummy taking the value one if the issue is Thomson ONE Banker underwritten by a reputable investment bank (top 8 investment banks in terms of market share for each security)

Stock price run-up Average daily stock return minus the average CRSP return for the CRSP equally-weighted market index over the window −76 to −2 before the announcement

Stock return volatility Daily stock return volatility over the window CRSP −240 to −40 before the announcement date

Trading volume Average monthly trading volume in the six CRSP months before the issue. Nasdaq trading volumes are divided by two to correct for double counting

Treasury Bond yield Three-month US Treasury Bill yield before the Datastream announcement date

Total assets (log) Logarithm of the book value of total assets Compustat

ΔTotal assets Logarithmic transformations of total assets for Compustat each of the four years after issuance minus the value in the year before issuance scaled by total assets in the year before issuance plus one

100 Table 4 Descriptive statistics for control variables

This table reports descriptive statistics for the set of control variables in the equity, convertible debt and straight debt samples. Panel A reports the mean and median for each control variable and security. Panel B reports t- statistics for pairwise differences in means of the control variables between equity and convertible debt (CD), equity and straight debt (SD), and convertible debt and straight debt. Table 3 gives the definition and source of all variables. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variable Equity Convertible debt Straight debt Funding deficit 0.24 (0.10) 0.07 (0.02) 0.05 (−0.01) EBIT −0.15 (0.01) 0.02 (0.06) 0.11 (0.10) Incometax 0.01(0.00) 0.02(0.01) 0.03(0.03) Leverage 0.25(0.18) 0.24(0.21) 0.29(0.26) Stock return volatility 0.05 (0.04) 0.04 (0.03) 0.03 (0.02) Creditrating(index) 19.06(22.00) 17.43(22.00) 10.15(9.00) TB yield (%) 2.17 (1.55) 2.34 (1.56) 2.52 (1.75) Stockpricerun-up(%) 0.27(0.17) 0.13(0.10) 0.01(0.00) Tradingvolume(thousands) 888(223) 1,753(557) 2,996(1,240) Total assets (millions) 1,685 (265) 3,338 (1,023) 15,560 (5638) Market-to-bookratio 4.33(2.61) 3.04(2.26) 3.71(2.41) Leading indicator 0.01 (0.01) 0.00 (0.01) 0.00 (0.00) Reputableunderwriter(dummy) 0.51(1.00) 0.78(1.00) 0.83(1.00) Hotmarkets 0.28(0.24) 0.33(0.27) 0.26(0.25) Panel B: Pairwise differences in means Variable Equity vs. CD Equity vs. SD CD vs. SD Funding deficit 0.17* 0.18** 0.01 EBIT −0.17*** −0.25*** −0.09*** Income tax −0.01*** −0.02*** −0.01*** Leverage 0.01 −0.05*** −0.05*** Stock return volatility 0.01*** 0.02*** 0.01*** Credit rating (index) 1.63*** 8.91*** 7.28*** TB yield (%) −0.16* −0.35*** −0.19** Stock price run-up (%) 0.14*** 0.26** 0.13*** Trading volume (thousands) −865*** −2,107*** −1,243*** Total assets (millions) −1,653*** −13,876*** −12,222*** Market-to-book ratio 1.28*** 0.61** −0.67** Leading indicator 0.00*** 0.01*** 0.00*** Reputable underwriter (dummy) −0.27*** −0.32*** −0.05*** Hot markets −0.05*** 0.03*** 0.08***

101 Table 5 OLS regressions of the determinants of issue size

This table reports coefficients and p-values of linear regressions of the determinants of issue size of seasoned equity, convertible debt and straight debt offerings. Table 3 gives the definition and source of all variables. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables Equity CD Debt Intercept 1.06*** 2.51*** 1.57*** (0.00) (0.00) (0.00) Funding deficit 0.03** 0.14* 0.03*** (0.04) (0.06) (0.00) EBIT −0.17** −0.21** 0.35 (0.02) (0.05) (0.37) Incometax 2.45*** 2.45*** 1.89** (0.00) (0.00) (0.03) Leverage −0.14* −0.14 0.05 (0.06) (0.24) (0.68) Stock return volatility −3.65*** −4.62*** −1.15 (0.00) (0.00) (0.38) Credit rating (index) 0.01* 0.00 0.03*** (0.08) (0.67) (0.00) TB yield 0.10*** 0.08*** −0.04*** (0.00) (0.00) (0.00) Stock price run-up 12.85** 4.21 −1.84 (0.03) (0.47) (0.80) Tradingvolume 0.00** 0.00*** 0.00*** (0.03) (0.00) (0.00) Total assets (log) 0.48*** 0.37*** 0.43*** (0.00) (0.00) (0.00) Market-to-book ratio 0.01*** 0.02*** −0.01*** (0.00) (0.00) (0.00) Leading indicator 4.75*** 2.28* −2.46*** (0.00) (0.07) (0.00) Reputable underwriter (dummy) 0.43*** 0.22*** 0.22*** (0.00) (0.00) (0.00) Hot markets −0.27 −0.07 0.70*** (0.12) (0.37) (0.00)

R 2 61.69 32.04 24.87 N 1,160 716 1,951

102 Table 6 Issue size impact and SEO announcement returns

This table reports coefficients and p-values of linear regressions of seasoned equity offering announcement returns. Regression (1) includes the selectivity correction term controlling for the endogeneity of issue size, and regression (2) adds the interaction between the selectivity correction term and issues size. Regression (3) adds the first stage residuals from the security choice model, and (4) shows the impact of issue size without controlling for endogeneity. Table 3 gives the definition and source of all variables. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) (3) (4) Intercept −2.18 −2.21 −4.69 2.32 (0.47) (0.46) (0.14) (0.23) Selectivity correction issue size −4.16* −4.23* −4.50** (0.06) (0.06) (0.04) Selectivity correction issue size x Issue size 0.03 (0.86) Selectivity correction security choice −2.83* (0.07) Issuesize 4.46** 4.42** 4.83** 0.33 (0.04) (0.04) (0.03) (0.24) Income tax −12.27 −12.21 −18.81* −2.41 (0.16) (0.16) (0.07) (0.75) EBIT 1.09** 1.08* 1.21** 0.53 (0.05) (0.06) (0.04) (0.38) Leverage 2.00* 2.00* 2.04* 1.47 (0.06) (0.06) (0.05) (0.13) Stock return volatility −21.51* −21.74* −15.71 −36.04*** (0.06) (0.06) (0.18) (0.00) Credit rating (index) −0.13** −0.13** 0.01 −0.08* (0.02) (0.02) (0.88) (0.09) TB yield −0.55** −0.54** −0.61** −0.13 (0.03) (0.04) (0.02) (0.27) Stock price run-up 3.38 3.25 19.64 55.15* (0.94) (0.94) (0.64) (0.09) Trading volume −0.00 −0.00 −0.00 0.00 (0.61) (0.62) (0.98) (0.77) Total assets (log) −2.46** −2.43** −3.19*** −0.50** (0.02) (0.03) (0.01) (0.05) Market-to-book ratio −0.09*** −0.08*** −0.08** −0.04* (0.01) (0.01) (0.02) (0.07) Leading indicator −2.74 −2.27 −6.33 16.35 (0.85) (0.88) (0.67) (0.13) Reputable underwriter (dummy) −1.55 −1.52 −1.81* 0.22 (0.13) (0.15) (0.08) (0.58) Hotmarkets 1.57 1.55 2.27 0.47 (0.34) (0.36) (0.18) (0.76)

R 2 3.09 3.10 3.31 2.92 N 1,160 1,160 1,160 1,160

103 Table 7 Impact of issue size and convertible debt announcement returns

This table reports coefficients and p-values of linear regressions on convertible debt offering announcement returns. Regression (1) includes the selectivity correction term controlling for the endogeneity of issue size, and regression (2) adds the interaction between the selectivity correction term and issues size. Regression (3) adds the first stage residuals from the security choice model, and (4) shows the impact of issue size without controlling for endogeneity. Table 3 gives the definition and source of all variables. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) (3) (4) Intercept −41.73** −41.62** −57.48*** −5.50* (0.02) (0.02) (0.00) (0.08) Selectivity correction issue size −14.32** −14.89** −15.08** (0.04) (0.03) (0.03) Selectivity correction issue size x Issue size 0.11 (0.81) Selectivity correction security choice −10.20* (0.08) Issuesize 15.69** 15.66** 16.60** 1.43*** (0.02) (0.02) (0.02) (0.00) Income tax −32.88* −32.86* −16.42 0.92 (0.07) (0.08) (0.42) (0.91) EBIT 6.61*** 6.60*** 13.62*** 3.15* (0.01) (0.01) (0.00) (0.06) Leverage 3.20** 3.20** 4.54*** 1.39 (0.03) (0.03) (0.01) (0.22) Stock return volatility −32.47 −32.21 −68.75* −96.17*** (0.33) (0.33) (0.09) (0.00) Credit rating (index) −0.08 −0.09 −0.58** −0.05 (0.13) (0.13) (0.04) (0.33) TB yield −0.91 −0.91 −0.78 0.24 (0.11) (0.11) (0.19) (0.12) Stock price run-up −20.36 −19.73 −229.91* 37.20 (0.82) (0.83) (0.10) (0.65) Trading volume −0.00 −0.00 −0.00* 0.00** (0.46) (0.47) (0.06) (0.02) Total assets (log) −5.67** −5.67** −2.58 −0.40 (0.03) (0.03) (0.41) (0.15) Market-to-book ratio 27.60* −0.27* −0.35** 0.05 (0.08) (0.08) (0.05) (0.28) Leading indicator 0.28 27.66 41.43* 57.99*** (0.20) (0.20) (0.05) (0.00) Reputable underwriter (dummy) −4.11** −4.10** −3.16* −1.02* (0.01) (0.01) (0.07) (0.08) Hot markets 1.05 1.03 −2.93 0.02 (0.34) (0.34) (0.24) (0.98)

R 2 16.41 16.43 14.89 15.49 N 716 716 716 716

104 Table 8 Impact of issue size and straight debt announcement returns

This table reports coefficients and p-values of linear regressions on straight debt offering announcement returns. Regression (1) includes the selectivity correction term controlling for the endogeneity of issue size, and regression (2) adds the interaction between the selectivity correction term and issues size. Regression (3) adds the first stage residuals from the security choice model, and (4) shows the impact of issue size without controlling for endogeneity. Table 3 gives the definition and source of all variables. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) (2) (4) Intercept 6.08** 5.96** 5.24 3.15*** (0.02) (0.02) (0.13) (0.01) Selectivity correction issue size 1.77 1.17 2.17 (0.21) (0.43) (0.26) Selectivity correction issue size x Issue size 0.15** (0.04) Selectivity correction security choice 0.90 (0.38) Issue size −2.13 −2.26 −2.54 −0.36*** (0.13) (0.11) (0.19) (0.00) Incometax 8.27 8.82 9.76 3.45 (0.18) (0.16) (0.12) (0.32) EBIT −1.32 −1.49 −0.82 −1.06 (0.28) (0.23) (0.54) (0.36) Leverage −1.78*** −1.71*** −1.66*** −1.93*** (0.00) (0.00) (0.00) (0.00) Stock return volatility −4.14 −4.60 −6.65 −1.50 (0.65) (0.62) (0.36) (0.87) Credit rating (index) 0.04 0.05 0.03 −0.02 (0.43) (0.29) (0.69) (0.46) TB Yield −0.04 −0.06 −0.05 0.03 (0.59) (0.43) (0.55) (0.51) Stock price run-up 67.67* 66.40 56.52* 70.55* (0.09) (0.10) (0.06) (0.08) Tradingvolume 0.00 0.00 0.00 0.00 (0.13) (0.11) (0.32) (0.43) Total assets (log) 0.74 0.81 1.07 −0.00 (0.21) (0.18) (0.23) (0.97) Market-to-book ratio −0.04** −0.04** −0.05* −0.02* (0.02) (0.02) (0.06) (0.06) Leading indicator −4.88 −5.70 −5.47 −0.54 (0.41) (0.34) (0.36) (0.91) Reputable underwriter (dummy) 0.14 0.25 0.30 −0.25 (0.71) (0.50) (0.55) (0.26) Hot markets 0.89 1.17 1.07 −0.33 (0.56) (0.44) (0.53) (0.78)

R 2 2.59 3.06 2.63 2.55 N 1,951 1,951 1,951 1,951

105 Table 9 Average post-issue earnings by issue size residuals

This table reports average post issue operating income before depreciation (OIbD) scaled by total assets for quartiles of issue size residuals. Residuals on issue size reflect managerial private information regarding a shortfall in earnings. t is the year after issuance. The columns Diff(4 – 1) and t(diff) report the difference and statistical significance, respectively, in OIbD between quartiles 4 and 1. Variable t Low 2 3 High Diff(4 − 1) t(diff) OIbD / total assets 1 0.043 0.081 0.078 0.055 0.012 0.492 2 0.066 0.085 0.077 0.070 0.004 0.761 Average 0.059 0.084 0.077 0.071 0.012 0.368

106 Table 10 Use of proceeds for the different security types

This table reports the effect of issue proceeds and residual issue proceeds on subsequent increases in assets and expenditures for equity (panel A), convertible debt (panel B) and straight debt (panel C). The dependent variable for asset-based variables (cash, inventory and total assets) is the log of the difference between the values of the variable in the respective year and in the year of issuance scaled by total assets in the year of issuance plus one. For expenditure-based variables, it is the log of the sum of values in the years after issuance up to and including the respective year, scaled by total assets plus one. All regressions include year fixed effects and the log of total assets for which coefficients are omitted for brevity. Bold letters indicate statistical significance. t is the year after issuance. Panel A: Equity issues Variable t N Issue p-value Residual p-value Other p-value size issue sources size ΔCash 1 923 0.70 0.00 0.21 0.00 1 923 5.25 0.00 0.17 0.00 2 734 0.77 0.00 0.23 0.00 2 734 4.53 0.09 0.21 0.00 3 518 0.70 0.00 0.19 0.00 3 518 3.98 0.28 0.16 0.00 4 442 0.51 0.01 0.18 0.00 4 442 2.64 0.40 0.17 0.00 ∑CAPEX 1 923 −0.02 0.61 0.07 0.00 1 923 0.16 0.67 0.07 0.00 2 734 −0.12 0.08 0.14 0.00 2 734 −0.42 0.55 0.15 0.00 3 518 −0.14 0.11 0.16 0.00 3 518 −1.19 0.30 0.16 0.00 4 442 −0.09 0.40 0.21 0.00 4 442 −1.07 0.43 0.21 0.00

∑Acquisition 1 923 0.13 0.06 0.05 0.00 1 923 2.06 0.17 0.04 0.00 2 734 0.23 0.01 0.09 0.00 2 734 3.24 0.05 0.09 0.00 3 518 0.29 0.01 0.12 0.00 3 518 4.21 0.01 0.12 0.00 4 442 0.54 0.00 0.19 0.00 4 442 5.24 0.00 0.17 0.00

∑R&D 1 923 0.29 0.00 −0.02 0.04 1 923 1.18 0.23 0.04 0.00 2 734 0.51 0.00 −0.06 0.00 2 734 1.57 0.47 −0.08 0.00 3 518 0.78 0.00 −0.05 0.03 3 518 1.66 0.65 −0.08 0.00 4 442 0.74 0.01 −0.05 0.11 4 442 0.38 0.92 −0.07 0.01

107 Table 10 (continued)

ΔInventory 1 923 0.06 0.00 0.01 0.00 1 923 0.82 0.01 0.01 0.00 2 734 0.05 0.06 0.02 0.00 2 734 0.72 0.04 0.02 0.00 3 518 0.07 0.03 0.03 0.00 3 518 0.21 0.49 0.03 0.00 4 442 0.09 0.07 0.04 0.00 4 442 0.21 0.65 0.04 0.00

ΔTotal assets 1 923 1.21 0.00 0.36 0.00 1 923 9.97 0.00 0.30 0.00 2 734 1.32 0.00 0.51 0.00 2 734 9.16 0.00 0.47 0.00 3 518 1.35 0.00 0.52 0.00 3 518 6.17 0.07 0.47 0.00 4 442 1.47 0.00 0.63 0.00 4 442 5.19 0.22 0.58 0.00 Panel B: Convertible debt issues Variable t N Issue p-value Residual p-value Other p-value size issue sources size ΔCash 1 610 0.41 0.00 0.16 0.00 1 610 18.89 0.02 0.14 0.00 2 543 0.40 0.01 0.20 0.00 2 543 29.15 0.00 0.19 0.00 3 452 0.27 0.06 0.14 0.00 3 452 7.20 0.61 0.14 0.00 4 380 0.35 0.04 0.17 0.00 4 380 12.26 0.65 0.17 0.00

∑CAPEX 1 610 0.02 0.42 0.07 0.00 1 610 6.58 0.05 0.03 0.00 2 543 0.08 0.16 0.05 0.00 2 543 18.39 0.06 0.06 0.00 3 452 0.09 0.35 0.07 0.00 3 452 28.10 0.09 0.08 0.00 4 380 0.05 0.62 0.11 0.00 4 380 26.21 0.12 0.11 0.00 ∑Acquisition 1 610 0.42 0.02 0.15 0.00 1 610 12.15 0.17 0.12 0.00 2 543 0.41 0.02 0.10 0.00 2 543 18.21 0.13 0.09 0.00 3 452 0.35 0.07 0.10 0.00 3 452 9.70 0.45 0.10 0.00

108 Table 10 (continued)

4 380 0.50 0.02 0.16 0.00 4 380 31.31 0.19 0.15 0.00 ∑R&D 1 610 0.26 0.00 −0.06 0.04 1 610 10.06 0.03 −0.07 0.00 2 543 0.52 0.00 −0.05 0.00 2 543 29.11 0.00 −0.06 0.00 3 452 0.74 0.00 −0.05 0.00 3 452 41.83 0.00 −0.06 0.00 4 380 0.99 0.00 −0.05 0.03 4 380 51.42 0.00 −0.05 0.05

ΔInventory 1 610 0.05 0.00 0.02 0.00 1 610 1.51 0.11 0.02 0.00 2 543 0.07 0.00 0.02 0.00 2 543 3.67 0.03 0.02 0.00 3 452 0.06 0.04 0.03 0.00 3 452 2.79 0.27 0.03 0.00 4 380 0.04 0.20 0.04 0.00 4 380 0.35 0.91 0.04 0.00 ΔTotal assets 1 610 1.19 0.00 0.43 0.00 1 610 51.07 0.00 0.35 0.00 2 543 1.25 0.00 0.46 0.00 2 543 81.21 0.00 0.45 0.00 3 452 1.06 0.00 0.42 0.00 3 452 62.94 0.00 0.42 0.00 4 380 1.15 0.00 0.51 0.00 4 380 63.21 0.01 0.50 0.00 Panel C: Straight debt issues Variable t N Issue p-value Residual p-value Other p-value size issue sources size ΔCash 1 1,587 −0.06 0.43 0.02 0.22 1 1,587 −2.52 0.70 0.03 0.10 2 1,332 −0.16 0.07 0.00 0.67 2 1,332 −9.23 0.21 0.01 0.51 3 1,079 −0.1 0.37 0.02 0.13 3 1,079 −6.34 0.47 0.02 0.10 4 916 −0.14 0.17 0.00 0.67 4 916 −4.46 0.57 0.01 0.56

∑CAPEX 1 1,587 0.12 0.01 0.12 0.00 1 1,587 16.15 0.00 0.13 0.00 2 1,332 0.08 0.35 0.24 0.00 2 1,332 14.67 0.03 0.25 0.00

109 Table 10 (continued)

3 1,079 0.04 0.77 0.32 0.00 3 1,079 13.65 0.20 0.32 0.00 4 916 0.01 0.93 0.38 0.00 4 916 10.25 0.42 0.38 0.00 ∑Acquisition 1 1,587 0.23 0.00 0.11 0.00 1 1,587 17.42 0.00 0.11 0.00 2 1,332 0.35 0.00 0.19 0.00 2 1,332 22.91 0.00 0.18 0.00 3 1,079 0.35 0.00 0.23 0.00 3 1,079 29.38 0.03 0.23 0.00 4 916 0.38 0.01 0.24 0.00 4 916 28.75 0.05 0.24 0.00 ∑R&D 1 1,587 0.05 0.00 0.00 0.23 1 1,587 2.00 0.00 0.00 0.80 2 1,332 0.11 0.00 0.02 0.11 2 1,332 6.22 0.00 0.02 0.13 3 1,079 0.10 0.02 0.02 0.05 3 1,079 6.70 0.00 0.02 0.06 4 916 0.12 0.04 0.02 0.04 4 916 7.44 0.01 0.02 0.04 ΔInventory 1 1,587 −0.01 0.65 0.02 0.00 1 1,587 1.14 0.36 0.02 0.00 2 1,332 −0.03 0.38 0.03 0.00 2 1,332 0.05 0.98 0.04 0.00 3 1,079 −0.04 0.40 0.05 0.00 3 1,079 2.62 0.68 0.06 0.00 4 916 −0.01 0.92 0.06 0.00 4 916 3.29 0.65 0.06 0.00

ΔTotal assets 1 1,587 0.42 0.00 0.39 0.00 1 1,587 45.39 0.00 0.41 0.00 2 1,332 0.40 0.00 0.60 0.00 2 1,332 42.74 0.00 0.61 0.00 3 1,079 0.40 0.03 0.72 0.00 3 1,079 49.93 0.00 0.73 0.00 4 916 0.37 0.09 0.77 0.00 4 916 46.54 0.01 0.77 0.00

110 Chapter 4 The costs of raising capital: New evidence

Abstract We provide new evidence on the magnitude and determinants of the direct costs of raising capital by studying a comprehensive data set of security offerings made between 1999 and 2011 by U.S. firms. Our analysis includes non-shelf, shelf, and 144a offerings and controls for firms’ self-selection into different security classes (seasoned equity, convertibles, and straight bonds). We find that underwriter spreads are larger for equity(-linked) security offerings than debt offerings, and larger for non-shelf than 144a and shelf offerings. We furthermore show that underwriter spreads for particular security types are affected by proxies for underwriter effort and by unobservable characteristics guiding the security choice. As such, they are not directly comparable across security types.

111 4.1. Introduction When companies raise external capital, they face a range of financing costs. Indirect costs include the potentially negative stock price reaction to the security offering announcement and underpricing of the newly offered securities. A large number of studies examines the magnitude and determinants of these costs.1 This paper turns the focus on the direct costs of security issuance. These costs are generally negotiated prior to the offering and represent the direct compensation to the underwriting syndicate (underwriter spreads) and other direct expenses (registration fees and printing, legal and auditing costs).2 In their often-cited study based on a data set of U.S. securities issues over the period 1990 to 1995, Lee et al. (1996) report average total direct costs (sum of underwriting spreads and other expenses) of 7.1 percent of offering proceeds for seasoned equity offerings (SEOs), 3.8 percent for convertible bonds, and 2.2 percent for straight bonds. The goals of this paper are twofold. First, we want to update existing knowledge on underwriter spreads by using a recent sample of U.S. securities issues spanning 1999 until 2011. Our data set includes shelf and 144a security offerings, which nowadays constitute a major part of public securities issuance. Lee et al. (1996) exclude shelf offerings, stating that these offerings were extremely rare at the time of their analysis, and do not mention 144a offerings, probably because the SEC only approved Rule 144a in April 1990. Our analysis also controls for the potential impact of the Global Financial Crisis (GFC) on underwriting spreads. Second, we want to examine to what extent observed direct issuance costs are conditional on observable and unobservable differences in security issuer characteristics. As such, we extend the descriptive analysis of Lee et al. (1996), who acknowledge (p. 59) that “we do not address why firms issue the securities they do.” To control for firms’ self-selection into security types, we analyze security choices and associated underwriting fees in an integrated framework using a switching regression framework (Li and Prabhala, 2007). This gives us unbiased estimates of the impact of a range of determinants of underwriter spreads and of counterfactual fees for alternative security choices. Our main findings are as follows. In line with previous studies we find higher direct

1 Studies examining security offering announcement returns include Asquith and Mullins (1986), Eckbo (1986), Mikkelson and Partch (1986), Chaplinsky and Hansen (1993), and Bayless and Chaplinsky (1996). Studies examining underpricing include Corwin (2003) and Kim et al. (2010) for seasoned equity offerings (SEOs), Wasserfallen and Wydler (1988) and Cai et al. (2007) for straight bonds, and Amman et al. (2003) for convertible bonds. 2 As in previous studies, we use the terms underwriting fees and spreads interchangeably.

112 issuance costs for equity issues than for debt issues. Convertible bond underwriting fees fall between equity and debt. However, we find that direct issuance costs have decreased for all three security classes compared to the findings of Lee at al. (1996), possibly as a result of increased competition among underwriters due to the entry of commercial banks in the underwriting market in 1999.3 For the different distribution mechanisms, we find that direct issuance costs are lower for shelf and 144a offerings than for non-shelf offerings. With respect to economies of scale, on a univariate basis, we only find evidence for scale economies for equity offerings and shelf convertible debt. Moreover, we do not find any evidence of diseconomies of scale for larger security offerings for any security type. The latter finding changes in a multivariate setting controlling for both fixed and marginal costs of the offering following Altinkilic and Hansen (2000). Specifically, we find evidence for economies of scale, i.e. decrease in fixed costs, for both non-shelf equity and shelf straight bonds. With respect to diseconomies of scale, we find an increase in marginal costs for both shelf and 144a straight bonds. As there is less legal due diligence for shelf and 144a offerings compared to non-shelf offerings, the latter confirms the finding of Altinkilic and Hansen (2000) that there are larger diseconomies of scale for riskier bonds. Our analysis of the determinants of underwriting fees shows that fees increase with proxies for underwriters’ due diligence, pricing, and marketing efforts associated with the offering. Significant determinants of fees differ across security and distribution mechanisms. For example, for non-shelf equity issues we find a significant negative impact of pre-issuance stock run-ups on fees, whilst underwriting spreads of debt-type securities are significantly positively influenced by pre-issuance stock return volatility. Underwriting fees of convertible bond issues are significantly higher for offerings underwritten by reputable underwriters, which is consistent with Fang (2005) who shows that underwriters charge a premium for offerings carrying a higher amount of risk. 4 We also find that fees are higher for shelf offerings made during the Global Financial Crisis (GFC). This can potentially be explained by increased concerns about reputational capital during this time, which may have led underwriters to charge an additional risk premium for this security type, as they have little time to adequately certify an “off-the-shelf” issue (Denis, 1991 and Sherman, 1999). We furthermore find evidence of a significant impact of firms’ self-selection into security

3 From 1999 onwards, the Financial Services Modernization Act allowed subsidiaries of commercial banks to have unlimited underwriting powers. 4 Brennan and Kraus (1987) and Brennan and Schwartz (1988) predict that convertibles are issued by firms with higher uncertainty about their risk, compared with straight bonds. 113 types on the observed fees for some security classes.5 In particular, we find that non-shelf and shelf equity issuers have unobservable characteristics that lead to higher underwriter spreads than if a random firm from the universe of security offering firms had issued these security types. Conversely, we find a negative selection effect for shelf and 144a straight bond issuers, resulting in lower observed underwriter spreads than if security choice was exogenous. The switching regression model allows us to calculate unconditional average underwriting fees that would pertain with random assignment of firms to security types. Our estimates suggest that the unconditional average underwriting fee for non-shelf equity is 5.04%, significantly lower than the actual (conditional) fee of 5.34%. Similarly, for shelf equity, we find an unconditional average fee of 3.86%, which is significantly lower than the actual fee for shelf equity issuers of 4.69%. The unconditional average underwriting fee for convertibles, shelf straight bonds, and 144a bonds, in turn, are significantly higher than the actual observed fees for these security types. 6 Our analysis thus suggests that security issuance costs are not readily comparable across security and distribution classes, as they are conditional on observable and unobservable characteristics that differ across issuer types. Apart from building on the Lee et al. (1996) analysis, our paper complements a number of other studies on the direct issuance costs of securities issuance. These papers tend to focus on a limited set of underwriting fee determinants, and on individual security types. For SEOs, Hansen et al. (1987) examine whether overallotment options impact underwriter spreads, but find no evidence. Butler et al. (2005) show a negative relationship between stock liquidity and underwriter spreads, while Lee and Masulis (2009) find a negative relationship between firms’ accounting quality and underwriter spreads. Finally, Kim et al. (2010) report a positive relation between underwriter spreads and underpricing for equity issues. Turning to debt issues, Livingston and Miller (2000) show for a sample of non- convertible debt issues that investment bank reputation and repeat business with the same underwriter lead to lower underwriter spreads. Livingston and Zhou (2002) find that underwriter spreads for industrial and utility 144a bond issues do not differ from those for publicly issued bonds. For Eurobonds, Esho et al. (2004) show that underwriter spreads decrease if issued under English law and increase with investment bank reputation, when

5 We exclude non-shelf and shelf convertible bonds and non-shelf straight bonds from our sample for this analysis due to having less than 100 observations for each security type. 6 For convertible debt we find an actual fee of 2.84%, which is significantly lower than the unconditional average fee of 4.36%. For shelf and 144a straight bonds the actual (unconditional average) fees are 0.82% (1.29%) and 1.13% (1.58%) respectively.

114 privately placed or issued in less frequent currencies. Ang and Zhang (2006) provide evidence that frequent issuers obtain lower underwriter spreads in the floating rate debt market. Finally, a handful of studies examine more than one security type, Kidwell et al. (1987) examine whether underwriter spreads declined after the introduction of shelf registrations in 1983. They find a reduction in underwriter spreads for both industrial straight debt and seasoned equity issues. Examining whether investment bank loyalty leads to lower underwriter spreads, Burch et al. (2005) find a negative relation between investment bank loyalty and underwriter spreads for common stock, but the reverse for straight debt. Finally, Kim et al. (2008) show that, since the early 1990s, underwriter spreads decreased for both debt and equity issues due to increased competition after the entry of commercial banks in the underwriting market. A limitation of these studies is that they do not allow for underwriting spreads being truncated by firms’ self-selection into security types and distribution mechanisms. Underwriting fees may not be directly comparable across security types and distribution mechanisms, as the underlying issuer characteristics driving the spreads may differ. Moreover, to the extent that there is self-selection, not controlling for this self-selection leads to biased estimates of the impact of underwriting spread determinants on fees.7 Our paper also contributes to a growing literature in corporate finance controlling for firms’ self-selection.8 Most closely related to our research are Dunbar (1995), Fang (2005) and Dutordoir and Hodrick (2012). Accounting for self-selection, Dunbar (1995) shows that underpricing and total offering costs of IPOs are reduced for security issuing firms using warrants as underwriter compensation. Fang (2005) controls for endogeneity in issuer– underwriter matching and shows that reputable underwriters of straight bond issues obtain lower yields and larger fees than less reputable underwriters. Dutordoir and Hodrick (2012) study security offering announcement returns conditional on firms self-selecting into a particular security type and show that firms select the security with the least negative announcement effect. The remainder of the paper continues as follows. The next section reviews the prior literature and develops our research questions. Section 3 describes our dataset and research methodology. Section 4 reports and discusses our empirical findings. Section 5 concludes.

7 Eckbo et al. (1990) show that OLS and GLS estimators are inconsistent if managers initiate corporate events. 8 Li and Prabhala (2007) survey self-selection models in corporate finance, reviewing both econometric models and empirical applications. 115 4.2. Hypothesis development We examine three interrelated research issues. First, we compare the magnitude of underwriter spreads and total direct costs 9 across three major security classes (seasoned equity, convertible bonds, and straight bonds) and three distribution mechanisms for non- privately issued securities (non-shelf, shelf, and 144a).10 With respect to differences across security types, following previous empirical evidence, we predict direct issuance costs are larger for equity than for convertible bond offerings, which in turn are larger than for straight bond offerings (e.g., Lee et al., 1996; Gande et al. 1999; Kim et al., 2008; Kim et al., 2010). These differences are largely due to higher asymmetric information and a lack of pricing related information, i.e. ratings, for equity issues, increasing the underwriting effort for equity-like issues. The SEC permanently adopted shelf registrations (Rule 415) in 1982. This rule allows companies to make multiple offerings for two years with a single prospectus. We have no clear prediction for the difference in costs across shelf and non-shelf public offerings. Firms can choose from a large list of potential underwriters for each separate offering due to a considerably less costly bidding process. The latter allows firms to restart the bidding process if the issuer does not like the terms of any underwriter, which may increase underwriter competition (Kidwell et al, 1987; Autore et al., 2008; Lee and Masulis, 2009). Direct issuance costs for shelf offerings may thus be lower than for non-shelf offerings. On the other hand, underwriters have little time to adequately certify an “off-the-shelf” issue (Denis, 1991 and Sherman, 1999). This weakens due diligence and increases the riskiness of the issue, for which underwriters may demand additional compensation (Booth and Smith, 1986; Tinic, 1988). The effect of reduced due diligence may be more pronounced from July 2005, as the SEC allowed automatic registration for shelf filings from this date, eliminating any prior legal approval.11 Rule 144a registrations allow companies to sell securities without having to file a public registration statement with the SEC. However, these securities can only be sold to, and traded among, qualified institutional investors. Direct issuance costs may therefore be higher, as

9 Following Lee at al. (1996) we define total direct costs as the sum of underwriter spreads and other direct expenses (including registration fee and printing, legal, and auditing costs) as a percentage of total proceeds. 10 We exclude private issues, as Gomes and Phillips (2012) show that there are fundamental differences between public and private markets in terms of asymmetric information, risk, and market timing. 11 With automatic registration, shelf registration statements become effective automatically upon filing without any prior approval by the SEC as for regular shelf registrations. This modification to rule 415 applies to “well- known seasoned issuers” (WKSIs). A WKSI is a firm that has filed all financial reports in a timely manner and has either a market capitalization greater than $700 million or has issued debt offerings greater than $1 billion over the previous three years.

116 weaker regulatory demands increase underwriters’ certification costs. On the other hand, Rule 144a issuers do not have to file a registration statement for SEC approval before proceeding with the offering. This results in a quicker issuance process. 12 This in turn may lower underwriting firms’ workload and associated fees (Livingston and Zhou, 2002). We furthermore predict that direct issuance costs are lower for all security types for larger offering sizes due to economies of scale effects. This prediction follows from a large literature showing that larger issues have lower spreads than smaller issues (e.g., for SEO spreads: Bhagat and Frost, 1986, Booth and Smith, 1986 and Lee et al., 1996; for convertible bond spreads: Lee at al., 1996; and for straight bonds: Blackwell and Kidwell, 1988 and Lee et al., 1996). These studies argue that this observation is a result of most of the underwriter expenses being fixed costs. However, Hansen and Torregrosa (1992) and Altinkilic and Hansen (2000) show that for very large offering sizes, there are diseconomies of scale, as after a decrease in fixed costs marginal costs of the offering increase. We thus predict both underwriter spreads and total direct costs to first decrease and then to increase as issue size increases, for all security types. Second, we study differences in underwriter spread determinants across different security types and for different distribution mechanisms.13 We predict these determinants influence underwriter spreads depending on whether they increase or decrease the underwriter effort in bringing the issue to the market. Underwriter effort increases with greater difficulty in respect of due diligence, i.e., assessing the issuing firm, and the pricing and marketing of the issue. The difficulty of all three tasks depend on firm- and issue-specific characteristics, capturing the conditions under which a firm raises capital and thus determining the underwriter effort in placing the issue. In particular, we predict underwriter spreads for equity-like security offerings increase with equity-related adverse selection costs, and spreads for debt-like security offerings increase with debt-related financing costs. We predict underwriter spreads for shelf and Rule 144a issues may be more strongly affected by these costs compared with non-shelf issues, as the reduced time for due diligence for these issues increases underwriter risk. Finally, underwriter spreads may be higher in the GFC, as increased concerns about reputational capital may have led underwriters to charge a risk premium.

12 It is estimated that 144a debt issues take half the time of a traditional registered offering to issue (Investment Dealers Digest, 1997 and Livingston and Zhou, 2002). 13 Previous studies that examine this issue focus on underwriter spreads rather than total direct costs, as firm characteristics do not influence other direct expenses, i.e., registration fees and printing costs do not depend on firm characteristics (see, for example, Kidwell et al., 1987; Altinkilic and Hansen, 2000; Livingston and Miller, 2000; Livingston and Zhou, 2002; Esho et al., 2004; Burch et al., 2005; Butler et al., 2005; Ang and Zhang, 2006; Kim et al., 2008; Lee and Masulis, 2009; Kim et al., 2010). To allow for comparability, we follow this precedent. 117 Third, we combine the study of security choice and underwriting fees by controlling for firms’ self-selecting into security classes and flotation types when analyzing underwriter spreads. We use a switching regression model, which gives unbiased estimates of the impact of underwriting fee determinants on each security type. We include a selectivity correction term from a first stage multinomial logit regression in the second stage underwriter spread regressions. If this term has a positive (negative) impact on underwriter spreads for a specific security class and distribution mechanism, this implies that observed fees are higher (lower) than if firms randomly select security types and distribution mechanisms. We do not make a prediction on whether the self-selection correction term has a positive or negative impact for particular security and distribution types, as this is mainly an empirical issue. In a final stage of our empirical study, we conduct a counterfactual analysis of what underwriter spreads would have been if the same firm had chosen a different security or distribution type. We cannot predict the magnitude of the counterfactual underwriter spreads, as they depend both on (i) differences in issuer characteristics across security classes, and (ii) differences in the determinants of underwriting spreads, which are both an empirical issue.

4.3. Data and methodology In this section, we discuss the sample of security offerings and the methodology for examining underwriter spread determinants and counterfactual underwriter spreads. We first discuss the sample selection criteria and present descriptive statistics on our dataset. We then explain the switching regression model and calculation of counterfactual underwriter spreads. Finally, we motivate the determinants of underwriter spreads.

4.3.1. Sample selection We download seasoned equity, convertible bond, and straight bond offerings by US firms between January 1999 and December 2011 from the SDC Global New Issues database. In line with prior studies, our search algorithm excludes utility and financial companies and non- 144a private issues. We also exclude mortgage- and asset-backed debt, secondary equity offerings, pass-through securities, rights offerings, unit offerings, convertible preferred stock, and exchangeable bonds. After aggregating multiple tranches of the same bond offering by the same firm on the same day into one offering, we obtain an initial sample of 3,364 seasoned equity offerings, 1,426 convertible bond offerings and 14,727 straight bond offerings. From this sample, we retain all issues for which company accounts data are available from the Compustat Fundamentals Annual database, stock price data are available

118 from the Center for Research in Security Prices (CRSP), and deal specific data are available from SDC. Applying these criteria, our final sample consists of 1,195 seasoned equity issues, of which 308 are non-shelf and 887 are shelf offerings, 294 convertible bond issues, of which 40 are non-shelf, 94 are shelf, and 160 are 144a offerings, and 1,363 straight bond issues, of which 62 are non-shelf, 1,200 are shelf, and 101 are 144a offerings.14 From these offerings the sample contains 292, 586, 38, 86, 150, 52, 396, 89 unique firm observations for non-shelf and shelf equity, non-shelf, 144a and shelf convertible bonds, and non-shelf, 144a and shelf straight bonds respectively. Table 1 provides an overview of the annual frequencies and offering volumes of non- shelf equity, shelf equity, non-shelf convertible bond, shelf convertible bond, 144a convertible bond, non-shelf straight bond, shelf straight bond, and 144a straight bond offerings. Consistent with previous studies we find that the number and issue volume of security offering fluctuates over time (e.g., Huang and Ramirez, 2010; Erel et al., 2011; Dutordoir and Hodrick, 2012). For example, for all security types the number of issues and the issuance volume drop at the start of the GFC in 2008. Moreover, there is a shift in popularity of equity shelf-registrations over time. This shift is also apparent in issuance volumes (expressed in $ millions), which are decreasing over time for non-shelf equity, but have an increasing trend for shelf equity. Similarly, shelf convertible and straight bond offerings have an increasing trend, which is also shown in total issuance volumes. Finally, 144a straight bond offerings experience a fall after 2004, whereas non-shelf straight bond offerings drastically increase in the final two years of the sample

4.3.2. Switching regression model We use a switching regression model, following the methodology Li and Prabhala (2007) describe and used in several studies of self-selection in corporate finance (e.g., Dunbar, 1995; Fang, 2005; Dutordoir and Hodrick, 2012). The model consists of a selection equation modeling firms’ security choice and outcome equations analyzing underwriter spreads for five security types, namely non-shelf equity, shelf equity, 144a convertible bonds, shelf straight bonds and 144a straight bonds. We exclude non-shelf and shelf convertible debt and non-shelf straight debt from the sample for this analysis due to having less than 100 observations for each security type. For the selection equation, we estimate a multinomial logit model in which we assign the

14 For the analysis of underwrite spread determinants, we exclude the 40 non-shelf and 94 shelf convertible bond offerings and the 62 non-shelf straight bond offerings due to a shortage of observations. 119 value one for non-shelf equity issues, two for shelf equity issues, three for 144a convertible bond issues, four for shelf straight bond issues, and five for 144a straight bond issues to the dependent variable. Prior to estimating this model, we test whether the independence of irrelevant alternatives (IIA) assumption holds for firms’ security choice. The IIA assumption demands that the log odds ratio of any two of the five security choice alternatives does not depend on the availability of any of the other alternatives. Only if this property holds, can we safely estimate the security choice with a multinomial logit model (Train, 2009). A Hausman test indicates that the IIA assumption is not violated in our dataset. Formally the selection equation is = + (1) ∗ ᇱ where is a latent variableܻ that௜ describesߛܼ௜ ߝ௜ the security choice and depends on measurable ∗ 15 determinants,ܻ , and unobservable factors, i.e. e. As we have five security choices the latent variable hasܼ four unknown thresholds to , such that when , the firm chooses ∗ ∗ non-shelfܻ௜ equity, when , theߠ firmଵ ߠ choosesସ shelf equity,ܻ௜ when≤ ߠଵ , the ∗ ∗ firm chooses 144a convertibleߠଵ ≤ ܻ௜ debt, ≤ ߠଶ when , the firm chooses shelfߠଶ ≤ straightܻ௜ ≤ ߠଷ debt, ∗ 16 and when > , the firm chooses 144a straightߠଷ ≤ ܻ௜ debt.≤ ߠସ, is the respective security choice. ∗ To controlܻ௜ forߠସ endogeneity of the security choice, we݅ allow the error terms in equation (1) to correlate with the error terms in the outcome equations determining underwriter spreads for the five security types, = + , = 1, … , 5 (2) where is the underwriterܷܹ spreadܵ௜ ߚ௜ forܺ theݑ݅ chosen݅ security offering, which depends on the determinants,ܷܹ ܵ௜ . As a result, the conditional expectations of the error terms, , are non-zero. To produce consistentܺ estimates, we apply an extension of the binary two-stepݑ௜ Heckman (1979) regression model (Lee, 1982). We estimate the multinomial logit model to obtain consistent estimates of and augment equation (2) with the selectivity correction term from this model, where the selectivity correction term is defined as (Lee, 1982; Hamilton and ߛ Nickerson, 2003)

( ) = for non-shelf equity, (3a) షభ ః ௣భ ଵ (భ ) ߣ = −ߔ ቀ ௣ ቁ for shelf equity, (3b) షభ ః ௣మ ߣଶ −ߔ ቀ ௣మ ቁ 15 Erel et al. (2012) and Dutordoir et al. (2013) use a similar approach in their security choice analysis. 16 We use robust standard errors in all regression analyses. Following the security choice literature, we do not use standard errors clustered at firm level as it is common in studies on using panel data (Petersen, 2009). Security issuance is rare for firms, so that most firms only appear once in the sample. Confirming this argument, our results remain unchanged using standard errors clustered at firm level.

120 ( ) = for 144a convertible debt, (3c) షభ ః ௣య ଷ (య ) ߣ = −ߔ ቀ ௣ ቁ for shelf straight debt, (3d) షభ ః ௣ర ସ ర( ) ߣ = −ߔ ቀ ௣ ቁ for 144a straight debt, (3e) షభ ః ௣ఱ ହ ఱ where is theߣ standard−ߔ ቀ normal௣ ቁ density function and the , , , , are estimated from equation

(1). We߶ can interpret the selectivity correction term as capturing݌ଵ ଶ ଷସ ହ underwriters’ updated beliefs on managerial private information influencing the security and distribution choice. Including them in the second stage underwriter spread regressions allows us to obtain consistent estimates of the impact of potential underwriter spread determinants.

4.3.3. Counterfactual analysis This analysis asks, what would the underwriter spread have been if a firm with particular characteristics had issued an alternative security instead of the security it actually issued? For example, how much would the underwriter spread paid by a non-shelf equity issuer have been if it had instead issued shelf equity, 144a convertible debt, shelf straight debt, or 144a straight debt? Formally, this can be expressed by the following four counterfactuals, ( = 0) (4a)

ܧ(ܷܹ ܵଶ௝ ∣ ܻ௜ = 0) (4b) ܧ(ܷܹ ܵଷ௝ ∣ ܻ௜ = 0) (4c) ܧ(ܷܹ ܵସ௝ ∣ ܻ௜ = 0) (4d) whereܧ equationܷܹ ܵହ௝ ∣ (4a)ܻ௜ describes the underwriter spread that equity issuer j would have paid for issuing shelf equity, equation (4b) describes the underwriter spread the equity issuer would have paid for a 144a convertible debt issue, equation (4c) describes the underwriter spread the equity issuer would have paid for issuing shelf debt, and equation (4d) describes the underwriter spread the equity issuer would have paid for a 144a debt issue. To calculate these four counterfactuals, we follow the methodology of Fang (2005) and Dutordoir and Hodrick (2012). Both studies suggest evaluating the issuer attributes using the parameter estimates obtained for the relevant outcome equations, e.g., to evaluate equation (4a), we multiply the parameter estimates of equation (4a) by equity issuer firm characteristics.

4.3.4. Determinants of underwriter spreads for different security and distribution types Underwriter spreads are hypothesized to be related to the effort underwriters spend on due diligence, pricing, and selling of the issue, which in turn depend on the level of firm-

121 specific asymmetric information costs for equity-like issues and debt-related financing costs for debt-like issues. If these costs are high, underwriters have to spend more time on due diligence to protect their reputational capital. Moreover, pricing and selling these issues is more difficult, as investors are more reluctant to buy these security offerings. We group and discuss these financing costs under: (1) equity-related adverse selection costs, (2) debt-related financing costs, and (3) general financing costs. Moreover, as theory predicts higher equity- related adverse selection costs for debt-type securities, and higher debt-related financing costs for equity-type securities, we follow the security choice literature and use the same variables as determinants in our first stage security choice model (e.g., Marsh, 1982; Bayless and Chaplinsky, 1991; Lewis et al., 1999). In addition, we add the difference between the firms’ debt ratio and the SIC 2-digit median debt ratio as exclusion restriction. As firms tend to move towards their target debt-ratio, we predict this variable to have a negative impact on firms’ likelihood to issue a debt-like security and a positive impact on firms’ likelihood to issue an equity-like security. Moreover, we do not expect it to have an impact on underwriter spreads, as underwriter spreads are more likely to be affected by proxies measuring equity- related adverse selection costs and debt-related risk proxies. Nevertheless it may capture some effects not fully controlled for in the second stage regressions.17 However, Li and Prabhala (2007) state that exclusion restrictions may not be strictly necessary in the baseline Heckman selection procedure, as the model could be identified by the non-linearity of the selectivity correction terms. It is possible, however, that the function of the generalized residuals is linear in parts of its domain and thus has very little variation relative to the remaining variables in the second stage equation. Nevertheless, Golubov at al. (2012) suggest testing whether the model is valid even without an exclusion restriction. We obtain robust results when testing our model without the exclusion restriction. Table 2 provides a detailed description and definition of all determinants. A first proxy for equity-related financing costs is pre-issue stock price run-up. A large run-up may signal valuable growth opportunities (Lucas and McDonald, 1990), making it easier for underwriters to place an offering. But it may signal firm over-valuation, leading to increased underwriter spreads. This may be particularly the case for shelf-equity issues, for which Autore et al. (2008) argue that smaller stock price run-ups can serve as an alternative

17 Industry target debt ratio differences may correlate with underwriter spreads through issue size, as equity (debt) issue size increases depending on how much firms’ debt ratio is below (above) the industry target. To alleviate this concern we substitute residual issue size for actual issue size. We obtain residual issue size following the methodology described in chapter 2. Untabulated results show that all of our results remain robust to this specification.

122 form of certification, offsetting the negative market reaction to weaker due diligence. Lee and Masulis (2009) show that equity-related adverse selection costs vary with the quality of a firm’s accounting information, which we proxy for, inversely, by the magnitude of abnormal accruals.18 A NYSE listing is associated with lower information asymmetry and underwriting fees, as NYSE firms need to fulfill strict listing criteria. As intermediation costs decline with stock liquidity (Demsetz, 1968), we expect this determinant, measured by trading volume, to have a negative impact on equity-related financing costs. Hot equity and convertible markets may offer windows of opportunity to lower equity financing costs, as investors are more likely to buy these issues during these periods. But they may also be associated with higher underwriter spreads due to increased underwriter workloads (Altinkilic and Hansen, 2000). Underwriter spreads of debt-like securities increase with debt-related financing costs. They increase with stock return volatility reflecting increased asymmetric information, uncertainty about future cash flows, and potential asset substitution costs (Green, 1984). We expect this effect to be particularly pronounced for shelf issues, as higher firm risk aggravates the issue of reduced due diligence (Blackwell et al., 1990; Denis, 1991). Debt-related financing costs, and hence underwriter spreads of debt-like securities, also increase with financial leverage, which is associated with a higher risk of financial distress and debt overhang costs (Myers, 1977). Conversely, debt-related financing costs and thus underwriter spreads should be lower for better performing firms, measured by EBIT, as higher pre-issue profitability makes it easier for a company to pay debt interest. Moreover, we expect these three proxies to influence firms’ choice to issue 144a convertible debt. Brown et al. (2012) show that firms resort to selling convertibles to hedge funds when costs for issuing seasoned equity (relative to the costs of establishing and maintaining a short position) are high. This is particularly the case for firms with higher stock return volatility and a higher likelihood of financial distress proxied for by financial leverage and profitability. Treasury Bond yields control for the level of external financing costs at the time of the offering. As for equity-like securities, the expected impact of hot debt markets is ambiguous. For debt-like securities, we include additional determinants suggested by empirical studies (Rogowski and Sorensen, 1985; Livingston and Miller, 2000; Livingston and Zhou, 2002). Lower credit ratings19 and longer maturities increase risk, while bond seniority and longer call protection reduce the risk

18 The appendix describes how we measure absolute abnormal accruals. 19 Following de Jong et al. (2012) we apply a numerical credit rating transformation, assigning a value of one to a Standard & Poors (S&P) AAA rating and adding the value one for each subsequent lower rating. If the S&P rating is unavailable, we use the Moody’s rating instead. For unrated convertibles, we assign the rating of the lowest-rated convertible in the sample, as unrated convertibles should have the highest risk; our results are robust to assigning the median credit rating value instead. 123 of a bond issue to a potential investor.20 As riskier debt requires more underwriting effort, we expect debt with lower credit ratings and longer maturities to have higher underwriter spreads and and debt with longer call protection to have lower underwriter spreads. Finally, general financing costs affect both equity- and debt-like securities. Following Altinkilic and Hansen (2000), we include the ratio of issue proceeds to market value and the inverse of proceeds to account for marginal and fixed direct issuance costs. We predict that marginal costs increase and fixed costs fall with issue size for all security types. We expect underwriter spreads to be lower for larger firms, as better accessibility of financial information reduces asymmetric information (MacKie-Mason, 1990). Following Myers and Majluf’s (1984) asymmetric information framework, we expect this effect to be stronger for equity-like than debt-like securities. Frequent issuers are less risky, as they are regularly evaluated by financial analysts. Moreover, underwriters may already have a fixed clientele for these issuers and have lower information costs. We follow Livingston and Zhou (2002) and control for the total amount of security offerings of each issuer in the sample.21 We expect this variable to have a negative impact on underwriter spreads, especially for shelf and 144a issues, as frequent issuance can be an alternative form of certification, offsetting the positive impact on underwriter spreads of less stringent due diligence and weaker information disclosure for these security types. Market return volatility controls for the macro-economic environment in which the issue takes place. Higher market return volatility is associated with times of greater uncertainty and information costs (Bhagat and Frost, 1986). We thus expect a positive impact on underwriter spreads, particularly for shelf and 144a issues, as firms can time these better. We include a dummy taking the value one for the years of and after the GFC (2008–2011). We expect this dummy to have a significant positive impact on fees, as increased concerns about reputational capital may have led underwriters to charge a risk premium. Finally, we control for underwriter reputation by including a dummy taking the value one if a reputable underwriter underwrites the issue. Reputable underwriters may charge a premium for their certification services and thus increase direct financing costs (Esho et al., 2004). Comparing investment grade and junk bonds, Fang (2005) shows that reputable

20 We do not use these variables in the regressions controlling for self-selection as these variables are debt- specific, so that there is no data for these variables for equity offerings. This makes it impossible to run the multinomial logit model and thus to calculate the selectivity correction terms. It also does not allow calculating counterfactual underwriter spreads based on these variables. We thus only include these variables for completeness in a regression examining debt underwriter spread determinants. 21 We test our findings for robustness to other specifications, such as total security offerings of each issuer in the two years before each security issue. This specification avoids a hindsight bias of the Livingston and Zhou (2002) specification. We also test both variables splitting them for the three security types. Our results are robust to all specifications.

124 underwriters have a stronger certification effect on junk bond issues than investment grade bonds, which can be attributed to junk bonds carrying a higher level of asymmetric information. As Brennan and Kraus (1987) and Brennan and Schwartz (1988) predict a higher level of asymmetric information for convertible bond issuers than for straight bond issuers, we expect the certification effect to be particularly pronounced for 144a convertible bonds. We also predict that reputable underwriters charge higher fees for shelf and 144a registrations, as lower levels of external due diligence for these registration types demand higher levels of internal due diligence to protect the underwriters’ reputation.

4.4. Empirical results In this section we discuss our main empirical findings. We first examine average underwriter spreads and total direct costs for all eight security types. We then present descriptive statistics on underwriter spread and security choice determinants, followed by a discussion of the results of a multinomial logit model of the security choice between non-shelf equity, shelf equity, 144a convertible bonds, shelf straight bonds and 144a straight bonds. Finally, we discuss the results of our underwriter spread regression models, which control for selection using selectivity correction from the first-stage logit model, and examine counterfactual underwriter spreads.

4.4.1. Average underwriter spreads and total direct costs Table 3 reports average underwriter spreads, other direct expenses and total direct costs for three security classes, namely seasoned equity, convertible bonds and straight bonds, and three distribution mechanisms, namely non-shelf, shelf and 144a. Following Lee et al. (1996) we also split the issues into proceeds categories. 22 This lets us distinguish between the average direct issuance costs for different proceeds amounts across the different security types (rows) and the extent of economies or diseconomies of scale as issue size increases for a particular security type (columns). We find that total direct costs are substantially larger for both types of equity offerings, non-shelf and shelf, at 6.79% and 5.51% compared with 1.23%, 1.01% and 1.81% for the three debt issue types (non-shelf, shelf, and 144a). Total direct costs for convertible bond issues lie in between with means of 3.67%, 2.98% and 6.34% for non-shelf, shelf, and 144a convertible bonds. Underwriter spreads, which exclude other direct expenses, show a similar

22 Following Lee et al. (1996) we use nominal values of issue size for comparability purposes. Untabulated results show that all finding are robust when correcting for inflation. 125 picture with substantially larger underwriter spreads for both non-shelf and shelf equity (5.34% and 4.69%) than for non-shelf, shelf, and 144a straight bonds (0.91%, 0.82%, and 1.13%). Convertible bond underwriter spreads again lie between with average values of 3.77%, 2.64%, and 2.84% for offerings of non-shelf, shelf and 144a convertible bonds. These findings suggest that mean underwriter spreads have fallen substantially compared with the beginning of the 1990s, as reported by Lee at al. (1996), especially for straight bond securities. A potential explanation is the entry of commercial banks as underwriters in 1999, the start year of our sample (Gande et al., 1999; Kim et al., 2008). Moreover, we find substantial differences between direct issuance costs of the three flotation methods. In particular, both mean total direct costs and underwriter spreads are larger for non-shelf than for shelf offerings for all three security classes, which is consistent with increased underwriter competition for shelf registrations due to a more flexible and less costly bidding process decreasing direct issuance costs. Direct issuance costs for 144a registrations mostly fall between the costs of non-shelf and shelf registrations, but evidence is slightly mixed.23 Moreover, the results show that there are scale economies effects for both types of equity issues (non-shelf and shelf) for both total direct costs und underwriter spreads. Total direct costs (underwriter spreads) range from 11.89% (6.51%) and 9.10% (6.31%) for offerings between $2 and $9.99 million to 3.48% (3.64%) and 3.07% (2.98%) for offerings above $500 million for non-shelf and shelf equity respectively. Turning to the debt-like security types, we only find some evidence of economies of scale for shelf convertible bonds with total direct costs (underwriter spreads) ranging from 4.80% (3.69%) for offerings between $40 and $59.99 million to 2.98% (2.64%) for offerings larger than $500 million. We thus find no evidence for diseconomies of scale as suggested by Hansen and Torregrosa (1992) and Altinkilic and Hansen (2000). For the other security types, the evidence does not suggest a clear pattern, again due to a very low number of observations. Figures 1 and 2 plot the data presented in table 3 for underwriter spreads and total direct costs respectively classified by the different proceeds categories.24 Both figures confirm the finding of economies of scale for both equity security types and of higher direct issuance costs for non-shelf than shelf equity. Moreover, the plots show a parallel movement of these economies of scale for the different proceeds categories. Similarly, for convertible and straight bonds both figures show a parallel movement of both total direct costs and underwriter spreads for the different proceeds

23 For both 144a convertible and 144a straight bonds, data on direct issuance costs are only available for roughly 20% of the issues. Thomson One Banker confirms that for these security types, managers do normally not disclose underwriter spread information. 24 We do not include 144a convertible and straight bonds in this figure due to a lack of observations.

126 categories within each security class, just on a different level for the different distribution mechanisms.

4.4.2. Descriptive statistics for security choice determinants Table 4 reports descriptive statistics for the control variables for samples of non-shelf equity, shelf equity, 144a convertible bonds, shelf straight bonds, and 144a straight bonds. We exclude shelf convertible and straight bonds and non-shelf convertible bonds from the subsequent analysis due to having less than 100 observations for each security type. Consistent with theory, equity-like security issuers have a significantly larger average stock price run-up than debt-like security issuers, with the largest run-up for non-shelf equity issuers. This confirms Autore et al.’s (2008) findings that firms offer shelf equity after a smaller stock price run-up, as this serves as an alternative form of certification. Inconsistent with expectation is a significantly higher average magnitude of abnormal accruals for equity- like security issuers than debt-like security issuers. A possible explanation is that the Jones (1991) model does not fully control for growth and equity issuers tend to be high growth firms. Moreover, equity issuers are on average smaller and less mature firms, which explains equity-like security issuers being on average less often listed on the NYSE than debt-like security issuers and equity-like issuers having lower average trading volumes. Consistent with hot markets offering a window of opportunity for security issuance, average market volumes are significantly higher for both equity security types and 144a convertible bond issues. For debt-related financing costs, stock return volatility is significantly lower for shelf than non-shelf equity issues and significantly lower for shelf straight bond issues than for 144a straight bond and 144a convertible bond issues. Financial leverage is highest for 144a straight bond issuers, which is not in line with expectation, as an additional debt issue aggravates financial distress and debt-overhang costs (Myers, 1977). As predicted, debt-like security issuers have significantly higher average earnings than equity-like security issuers, consistent with higher earnings making it easier to repay debt interest. Average earnings of 144a convertible bond issuers lie between those of equity- and debt-like security issuers. We expect Treasury Bond yields to be lower for debt-like than for equity-like security issues, as low yields offer a window of opportunity for issuing debt. However, we find them to be lowest for shelf security types, consistent with their more flexible timing. Inconsistent with prediction, we find no difference in the average market volumes of debt issues, measured by hot debt markets, between the different security types. Finally, 144a convertible bonds have on average the lowest issue rating and the largest maturity, followed by 144a straight bonds

127 and shelf straight bonds. The same order holds for years of call protection, but reverses for debt seniority. In line with expectations, issue proceeds are significantly higher for both debt issue types than for convertible debt, which in turn has significantly higher issue proceeds than shelf and non-shelf equity. Shelf straight bond issuers are the largest firms, measured by total assets, followed by 144a straight bond issuers and 144a convertible bond issuers. Both equity type issuing firms are significantly smaller with non-shelf equity issuers being smaller than shelf equity issuers. Mean market return volatility is significantly higher for shelf equity, shelf straight bond and 144a straight bond issues than for non-shelf equity and 144a convertible bond issues. Shelf straight bond issues are more likely to be underwritten by reputable underwriters, followed by 144a convertible bond issues, 144a straight bond issues, and shelf equity issues. Issue frequency is significantly higher for both types of debt securities than for equity-like security types. Finally, industry target debt ratio differences are largest for shelf equity and 144a convertible bond issuers. The latter finding is intuitive as convertible bonds are a vehicle for flexible capital structure adjustment.

4.4.3. Determinants of the choice between different security types Table 5 reports the results of the multinomial logit model. The output of this model consists of four pairwise regressions, each comparing a security choice alternative with non- shelf equity (the base outcome). Stock return volatility, earnings, Treasury Bond yields, and issue size have a significant negative impact on the choice of shelf over non-shelf equity. Conversely, consistent with the findings of Autore et al. (2008), financial leverage and firm size have a significant positive impact on the choice of shelf over non-shelf equity. As shelf registrations allow for greater flexibility over timing, the significant negative impact of hot debt markets and market return volatility is in line with expectations. Finally, the GFC dummy has a significant positive impact on firms’ choice to issue shelf over non-shelf equity, consistent with firms preferring more flexibility over timing in these times of higher market uncertainty. For the choice of 144a convertible bonds over a non-shelf public equity offering, we find a significant negative impact of abnormal accruals, the NYSE dummy, financial leverage and both issue size proxies. Conversely, the likelihood of a 144a convertible bond issue increases with total assets. The significant positive impact of hot convertible debt markets and the significant negative impact of hot equity markets and Treasury Bond yields are in line with prediction, as 144a registrations can be brought to the market comparatively quickly. Finally,

128 higher deviations from industry target debt levels make a 144a convertible bond issue more likely. Weaker accounting information, a NYSE listing, profitability, and larger issue proceeds increase firms’ likelihood to issue shelf straight bonds over non-shelf equity. Conversely, higher stock price run-ups and stock return volatility reduce this likelihood. The negative impact of Treasury bond yields on firms’ likelihood to issue shelf straight bonds over non- shelf equity is in line with expectation, especially for shelf registrations. The GFC dummy has a significantly positive coefficient, likely indicating again a shift to flotation types with more flexibility over timing the market in these uncertain conditions. As for shelf straight bonds, a stock price run-up has a significant negative impact on the choice to issue 144a straight bonds over non-shelf equity. We also find a significant positive impact of EBIT, issue size and total assets on this choice. Proxying for market conditions, hot equity markets reduce, while market return volatility increases firms’ likelihood to issue 144a straight bond offerings over non-shelf equity. The GFC dummy has a negative impact on firms’ likelihood to issue 144a straight bonds over non-shelf equity, which may be explained by the debt market crisis following the GFC. Finally, reputable underwriters do not pick up these issues, as indicated by a significant negative coefficient.25 To confirm the credibility of our estimates in the second stage underwriter spread analysis, we test the strength of our exclusion restriction, industry target debt ratio differences. The overall significance of this instrument in the model (taken all four equations together) measured by the  2 -statistic is 15.2, significant at the 1% level. This implies that our instrument fulfills the condition to be correlated with the endogenous variable, even though it is only significant in one of the four equations of the multinomial logit model.

4.4.4. Determinants of underwriter spreads Tables 6–10 present the results of a cross-sectional regression analysis of non-shelf equity, shelf equity, 144a convertible bond, shelf straight bond and 144a straight bond underwriter spreads. In these tables, column (1) presents the results for a regression analysis of underwriter spreads controlling for self-selection by including the selectivity correction term from the multinomial logit model, whereas column (2) shows the regression results not controlling self-selection. In tables 8–10, column (3) presents regression results including the

25 We acknowledge that this variable is not strictly exogenous, as firms may self-select into working with a particular underwriter. Repeating our empirical analysis omitting this variable confirms the robustness of the results in all respects. We also repeat our analysis including a dummy taking the value one if the issue is underwritten by an underwriting syndicate. Again, our results are robust. 129 four debt-specific variables, the rating index, the log of years to maturity, a dummy taking the value one for senior debt, and the percentage of years with call protection. Column (1) shows that stock price run-up has a negative effect on non-shelf equity underwriter spreads, consistent with a larger run-up making it easier for underwriters to place an offering and with Lucas and McDonald’s (1990) argument that a larger run-up signals growth opportunities. Leverage has a positive impact on spreads, in line with higher leverage increasing financial distress risk and, therefore, underwriter risk. The significant positive impact of leverage is thus in line with our predictions, whereas the significant negative impact of Treasury Bond yields is unexpected. Consistent with the predicted decline in fixed costs as issue size increases (Altinkilic and Hansen, 2000), we find a negative impact of inverse proceeds. Firm size has a negative impact, in line with a reduction of asymmetric information (MacKie-Mason, 1990) making it easier to sell the offering. All other underwriter spread determinants are insignificant for this security type. The selectivity correction term has a significant positive impact on non-shelf equity underwriter spreads. This indicates that unobservable characteristics raising the likelihood of a non-shelf equity issue are associated with higher underwriter spreads. This private information may relate to equity over-valuation, increasing the selling effort for underwriters. Results in column (2) show, that leverage and inverse proceeds are insignificant, while stock return volatility has a significant positive impact and the GFC dummy has a significant negative impact on underwriter spreads. Controlling for self-selection thus leads to different inferences on the impact of underwriter spread determinants. Table 7 repeats the analysis for shelf equity offerings. Column (1) shows that stock return volatility has a significant positive impact on shelf equity underwriter spreads. This is in line with expectations, as higher volatility increases the risk of the issue and aggravates the reduced due diligence for shelf registrations (Blackwell et al., 1990 and Denis, 1991). As for non-shelf equity, firm size has a significant negative impact. The GFC dummy has a significant positive impact, consistent with underwriters charging a risk premium for reputational concerns during this time period. Finally, issue frequency has a significant negative impact on underwriter spreads for this security type, confirming Livingston and Zhou’s (2002) findings of a reduction in asymmetric information through regular evaluation by financial analysts. All other underwriter spread determinants are insignificant for this security type. The insignificant impact of reputable underwriters is surprising due to higher demands of internal due diligence to protect reputation capital for shelf registrations. For this security type the effect of the selectivity correction term is not statistically significant.

130 Comparing the results of columns (1) and (2) shows that self-selection does not bias the regression results for this security type. Table 8 analyzes the underwriter spreads for 144a convertible bonds. In column (1) we find a significant positive impact of the NYSE dummy, which is unexpected, as a NYSE listing adds extra certification to the issue, reducing the underwriting effort. Conversely, the significant positive impact of the hot equity markets indicator variable is in line with expectations. Consistent with our predictions, we find a significant positive impact of the reputable underwriter dummy. This result confirms the findings of Fang (2005) also for convertible bonds, which are a security type mostly offered by firms with higher levels of asymmetric information. She shows that the certification effect of a reputable underwriter is particularly pronounced for offerings carrying a higher amount of asymmetric information, in her case junk bonds. The significant positive impact indicates that reputable underwriters can earn a premium for this certification effect. All other equity- and debt-related underwriter spread determinants are insignificant for this security type, which is to some extent not surprising as convertible debt is a hybrid security. The impact of the selectivity correction term is statistically insignificant. Moreover, self-selection does not bias the findings for this security type, as not controlling for selectivity correction leaves the statistical results from the model largely unchanged. The sole difference is the NYSE dummy is now insignificant. Results in column (3) show that we do not find a significant impact of any of the four debt-specific variables, which may be due to the fact that convertible debt is a hybrid security. For the other underwriter spread determinants we do not find any major differences compared to the other regression models. We repeat the analysis for shelf straight bond offerings. Table 9 presents the results, which show, in column (1), a significant negative impact of abnormal accruals which is not in line with our predictions, as a lower quality of firms’ accounting information increases asymmetric information costs. Conversely, the significant positive impact of stock return volatility on the underwriter spreads is consistent with predictions. Higher financial leverage, indicating higher financial distress costs, increases underwriter spreads, while higher earnings decrease them, as it is comparatively easier for companies with higher earnings to repay debt interest. Consistent with our predictions we find a significant positive impact of marginal issuance costs, as captured by proceeds over market value and a significant negative impact of fixed issuance cost on shelf straight bond underwriter spreads. Also in line with our expectations is the negative impact of firm size. The same holds for the negative impact of market return volatility and the positive impact of issue frequency. Finally, as for shelf equity,

131 a dummy taking the value one for the years of and after the GFC has a significant positive impact on shelf straight bond underwriter spreads, again suggesting that underwriters charge a risk premium for these registration types during this time. All other underwriter spread determinants are insignificant for this security type. As for shelf equity this is surprising for reputable underwriters, as higher internal due diligence requires increased underwriting fees. The selectivity correction term has a significant negative impact on shelf straight bond underwriter spreads. Shelf straight bond issuers self-selecting into this security type thus pay significantly lower underwriter spreads on average than the underwriter spreads that would obtain if firms randomly selected this security type. Turning to the results in column (2), we find that trading volume has a significant positive impact on shelf straight bond underwriter spreads, which is unexpected, as we predict it to have a significant negative impact on underwriter spreads for equity-like security types. For the other underwriter spread determinants, the results remain unchanged. Column (3) shows a positive impact for the rating index, debt maturity and years of call protection and a negative impact of a dummy taking the value one for senior debt. These results are in line with expectations, but for the positive effect of the years of call protection, as longer call protection reduces the risk of the bond offering and thus decreases the underwriter’s selling effort. For the other underwriter spread determinants, abnormal accruals, leverage, EBIT, market return volatility and the GFC dummy are insignificant in this framework. Finally, Table 10 presents the results for 144a straight bond issues. As for 144a convertible bonds, column (1) shows a significant positive impact of the NYSE dummy. The positive impact of hot convertible debt markets on underwriter spreads is in line with prediction, but the positive impact of hot debt markets is not, as they offer a window of opportunity for debt securities. On the other hand, as pointed out by Altinkilic and Hansen (2000), there is an increased work load in investment banks during hot market periods, increasing underwriter spreads. Treasury Bond yields have a positive impact on the 144a straight bond underwriter spreads, as this registration type allows for a better timing of the market. Moreover, we find significant positive impact of financial leverage and fixed issuance costs and a significant negative impact of firm size and EBIT, which are all in line with our expectations. More uncertain market conditions, proxied by market return volatility make it harder to place an offering and thus increase underwriter spreads. Finally, we find a significant negative impact of the GFC dummy on underwriter spreads for this security type. This again suggests a shift in preference towards shelf registrations leading to lower fees for alternative security types as demand decreases. All other underwriter spread determinants are

132 insignificant for this security type, which is particularly surprising for issue frequency, as frequent issuance can serve as an alternative form of certification for Rule 144a registrations. The selectivity correction term has a significant negative impact on the 144a straight bond underwriter spreads. This indicates that unobservable private information raising the likelihood of a 144a straight bond issue is associated with lower underwriter spreads for this security type. There is thus negative selection, suggesting that 144a straight bond issuers have characteristics which lead to lower underwriter spreads than if a random firm from the universe of security offering firms had issued this security type. Turning to the results in column (2), we find that trading volume is now significant positive and hot convertible and debt markets, Treasury Bond yields, firm size and the GFC dummy are all insignificant. Finally, in column (3), we find a significant positive impact of lower credit ratings, which is in line with our expectations. The other underwriter spread determinants remain mostly unchanged in comparison to the other models. In our main model we do not control for overallotment, as there is no overallotment option for debt securities. Including a dummy taking the value one for issues with an overallotment option and repeating our analysis on the determinants of underwriter spreads, correcting for self-selection, for non-shelf equity, shelf equity and convertible debt, leaves our main finding unchanged. We also check the robustness of our results to controlling for the modification to rule 415 stating that from July 2005 shelf registrations for “well known seasoned issuers” were effective automatically upon filing without any prior SEC approval as for regular shelf offerings. We include a dummy taking the value one for issues made after July 2005 and repeat the empirical analysis. Untabulated results show that our results are robust, except for a now insignificant impact of the selectivity correction term for 144a straight bonds.

4.4.5. Counterfactual analysis Table 11 presents the average counterfactual underwriter spreads that firms would have paid if they had issued alternative security types. The main diagonal cells show the actual underwriter spreads paid on average by issuers of each security type. The table allows us to compare the actual and counterfactual announcement effects had security issuers instead issued an alternative security type (comparison within security issuer types, done by comparing within a row) as well as had a specific security type instead been issued by other security issuer types (comparison within security types, done by comparing within a column). The latter allows us to show whether actual issuers have an advantage or disadvantage in

133 terms of fees for their security choice compared to other firms if they had chosen this security type. Counterfactual underwriter spreads are calculated by multiplying the coefficients of the alternative security type regression by the values of the explanatory variables for a particular issuer. E.g., in row (1), column (2), the counterfactual underwriter spread if a non-shelf equity issuer had instead issued shelf equity, is calculated by multiplying the coefficients from table 6, column (1) by the actual values of the corresponding explanatory variables for non-shelf equity issuers. The other counterfactual values are calculated analogously. The table also provides t-statistics, which are calculated with respect to the actual average underwriter spread for each security type. Column (1) indicates that the unconditional average underwriter spread for non-shelf equity (average fee over all security issuers) is 5.04%, compared with an actual (conditional) fee of 5.34%. Thus, non-shelf equity issuers pay larger fees than if a random firm from the universe of security offering firms had issued this security type. Similarly, for shelf equity, we find an unconditional average fee of 3.86%, which is significantly lower than the actual fee for shelf equity issuers of 4.69%. Turning to the underwriting fees that the various security issuers would have obtained when issuing 144a convertible bonds in column (3), we find a significantly lower actual fee (2.84%) than the average unconditional fee (4.36%). Column (4) presents the results for shelf straight bonds. We observe that shelf straight bond issuing firms with an actual underwriting fee of 0.82% compared to an unconditional average fee of 1.29% pay the comparatively lowest fees for this security type. This finding confirms the result from the underwriter spread regression analysis suggesting that shelf straight bond issuers have characteristics that result in lower underwriting fees. Similarly, in column (5) we find a significantly lower actual fee of 1.13%, which is significantly lower than the fees for all other security issuers (1.58%). Turning to the counterfactual underwriter spreads that one security issuer would have obtained had it issued another security type instead, row (1) indicates that non-shelf equity issuing firms could have obtained lower underwriting fees by issuing debt-like securities. However, for non-shelf equity issuers, these firms may not have any other choice in terms of security classes, due to limited debt capacity.26 We make a similar observation when looking at the underwriter spread that shelf equity issuers would have obtained, had they issued any other security type instead (row 2). Conversely, 144a convertible bond issuing firms would

26 Non-shelf equity issuers are the smallest in terms of total assets (1,192 million) compared to an average size of 9,002 million across issuers of the other security types. Moreover, only 13% of the non-shelf equity issuing firms have a credit rating compared to 65% for all other security issuers.

134 pay significantly larger fees when issuing any equity security type and significantly lower fees for the two debt security types. The finding with respect to equity confirms the results of Brown et al. (2012). However, as Brennan and Kraus (1987) and Brennan and Schwartz (1988) predict a higher level of asymmetric information for convertible bond issuers than for straight bond issuers, these firms may not have the opportunity to issue straight bond securities. Both shelf straight bond issuers (row 4) and 144a straight bond issuers (row 5) obtain the lowest fees for the actual security type, i.e., these firms would have obtained significantly larger underwriting fees, had they issued any other security type instead. In summary, we only find limited evidence for firms self-selecting into the security type that offers them the absolutely lowest direct financing. This finding contrasts with the evidence of Dutordoir and Hodrick (2012) suggesting that firms actively choose the financing type with the least negative announcement returns and Dunbar (1995) who shows that firms choose compensation contracts which minimize costs. We explain this finding by direct issuance costs being comparatively low compared to indirect issuance costs.

4.5. Conclusion We examine the magnitude and determinants of underwriter spreads for a sample of corporate US non-shelf equity, shelf equity, 144a convertible bond, shelf straight bond and 144a straight bond issues between 1999 and 2011 controlling for firms self-selecting into these security types. This also allows us to construct counterfactual underwriter spreads that would have been obtained if the same firm had issued another security type instead. We find significant differences in the magnitude of direct issuance costs for the different security types and a reduction of the overall level of fees in recent years. We find average costs to be larger for equity than for convertible bonds, for which direct issuance costs are again larger than for straight bonds. For the different distribution mechanisms, we find smaller direct costs for shelf equity than for non-shelf equity and smaller costs for shelf bonds than for 144a bonds. On a univariate basis, we only find evidence for economies of scale for equity offerings and shelf convertible debt, but no evidence for diseconomies of scale. The latter changes in a multivariate setting controlling for both fixed and marginal costs of the offering following Altinkilic and Hansen (2000). Specifically, we find evidence for economies of scale, i.e. decrease in fixed costs, for both non-shelf equity and shelf straight bonds and diseconomies of scale, i.e. increase in marginal costs, for both shelf and 144a straight bonds. As there is less legal due diligence for shelf and 144a offerings compared to non-shelf offerings, the latter confirms the result of Altinkilic and Hansen (2000) that there

135 are larger diseconomies of scale for riskier bonds. Moreover, there are significant differences in the determinants of the underwriter spreads for the different security types. In particular, we show that underwriter spreads depend on the overall underwriting effort, i.e., due diligence, pricing and selling, which increases with equity-related adverse selection costs and debt-related financing costs. Controlling for self-selection we find that non-shelf and shelf equity issuers’ observable and unobservable characteristics lead to higher underwriter spreads than for any other firm from the universe of security offering firms. Conversely, convertible and straight bond issuers have characteristics leading to lower conditional observed underwriter spreads. These results suggest that fees are partially conditional on observable and unobservable issuer characteristics, so that any comparison of fees across security classes and distribution mechanisms should take into account firms self-selecting into these security types. However, underwriter spreads also seem to have a fixed component similar to the 7% solution of Chen and Ritter (2000) for IPOs, which may explain the sticky fee differential between equity, convertible and straight bonds for both actual and counterfactual underwriter spreads. As Chen and Ritter (2000) argue the fixed component is a result of underwriters avoiding to compete on fees as security underwriting is a very profitable business area.

136 Appendix: Calculating abnormal accruals

We measure abnormal accruals using the modified Jones (1991) model. Abnormal accruals are the difference between total and normal accruals. Following Francis et al. (2005), we measure firm j’s total accruals in year t as TAj,t = ∆CAj,t − ∆CLj,t − ∆Cashj,t + ∆STDEBTj,t −

DEPNj,t. ∆CAj,t , ∆CLj,t, ∆Cashj,t, and ∆STDEBTj,t are firm j’s change in current assets, current liabilities, cash, and debt in current liabilities in the fiscal year before the announcement date and DEPNj,t is firm j’s depreciation and amortization expense. Normal accruals are estimated as,

, , ) , , = + + (A1) ଵ , (∆ோ௘௩ೕ೟ି,∆஺ோೕ೟ ௉௉ாೕ, ೟ ௝௧ ଵ ೕ೟షభ ଶ ೕ೟షభ ଷ ೕ೟షభ where Assetj,t−1 is firmܰܣ j’s totalߚ෢ ஺ assets௦௦௘௧ at theߚ෢ fiscal஺௦ year-end௦௘௧ twoߚ෢ years஺௦௦௘ before௧ the announcement date, ∆ARj,t is firm j’s change in accounts receivable in the fiscal year before the announcement date, ∆REVj,t is firm j’s change in revenues, measured by EBIT in the fiscal year before the announcement date and PPEj,t is firm j’s property plant and equipment at the fiscal year-end before the announcement date. , , and are estimates from a cross- sectional regression of total accruals ߚ෢ଵ ߚ෢ଶ ߚ෢ଷ , = + , + , + (A2) ்஺ೕ,೟ ଵ , ∆ோ௘௩,ೕ೟ ௉௉ாೕ, ೟ ೕ೟షభ ଵ ೕ೟షభ ଶ ೕ೟షభ ଷ ೕ೟షభ We use the measures஺௦௦௘௧ for totalߚ and஺௦௦ normal௘௧ accrualsߚ ஺௦௦௘௧ to calculateߚ ஺௦௦ abnormal௘௧ ߝ accruals (AAj,t), as the absolute value of TAj,t /Assetj,t−1 − NAj,t.

137 References

Altinkilic, O. and Hansen, R.S., 2000. Are there economies of scale in underwriting fees? Evidence of rising external financing costs. Review of Financial Studies 13, 191–218. Ammann, M., Kind, A. and Wilde, C., 2003. Are convertible bonds underpriced? An analysis of the French market. Journal of Banking and Finance 27, 635–653. Ang, J.S. and Zhang, S., 2006. Underwriting relationships: Information production costs, underwriting fees, and first mover advantage. Review of Quantitative Finance and Accounting 27, 205–229. Asquith, P. and Mullins Jr., D.W., 1986. Equity issues and offering dilution. Journal of Financial Economics 15, 61–89. Autore, D.M., Kumar, R. and Shome, D.K., 2008. The revival of shelf-registered corporate equity offerings. Journal of Corporate Finance 14, 32–50. Bayless, M. and Chaplinsky, S., 1991. Expectation of security and the information content of debt and equity offers. Journal of Financial Intermediation 1, 195–214. Bayless, M. and Chaplinsky, S., 1996. Is there a window of opportunity for seasoned equity issuance? Journal of Finance 51, 253–278. Bhagat, S. and Frost, P.A., 1986. Issuing costs to existing shareholders in competitive and negotiated underwritten public utility equity offerings. Journal of Financial Economics 15, 233–259. Blackwell, D.W. and Kidwell, D.S., 1988. An investigation of cost differences between public sales and private placements of debt. Journal of Financial Economics 15, 253–278. Blackwell, D.W., Marr, M.W. and Spivey, M.F., 1990. Shelf registration and the reduced due diligence argument: implications of the underwriter certification and the implicit insurance hypothesis. Journal of Financial and Quantitative Analysis 25, 245–259. Booth, J.R. and Smith II, R.L., 1986. Capital raising, underwriting and the certification hypothesis. Journal of Financial Economics 15, 261–281. Brennan, M.J. and Kraus, A., 1987. Efficient financing under asymmetric information. Journal of Finance 25, 1225–1243. Brennan, M.J. and Schwartz, E.S., 1988. The case for convertibles. Journal of Applied Corporate Finance 1, 55–64. Brown. S.J., Grundy, B.D., Lewis, C.M. and Verwijmeren, P., 2012. Convertibles and hedge funds as distributors of equity exposure. Review of Financial Studies 25, 3077–2112. Burch, T.R., Nanda, V. and Warther, V., 2005. Does it pay to be loyal? An empirical analysis of underwriting relationships and fees. Journal of Financial Economics 77, 673–699. 138 Butler, A.W., Grullon, G. and Weston, J.P., 2005. Stock market liquidity and the cost of issuing equity. Journal of Financial and Quantitative Analysis 40, 331–348. Cai, N.K., Helwege, J. and Warga, A., 2007. Underpricing in the corporate bond market. Review of Financial Studies 20, 2021–2046. Chaplinsky, S. and Hansen, R.S., 1993. Partial anticipation, the flow of information and the economic impact of corporate debt sales. Review of Financial Studies 6, 709–732. Chen, H-C. and Ritter, J.R., 2000. The seven percent solution. Journal of Finance 55, 1105– 1131. Corwin, S.A., 2003. The determinants of underpricing for seasoned equity offers. Journal of Finance 58, 2249–2279. De Jong, A., Dutordoir, M., van Genuchten, N. and Verwijmeren, P., 2012. Convertible arbitrage price pressure and short-sale constraints. Financial Analysts Journal 68, 70–88. Demsetz, H., 1968. The cost of transacting. The Quarterly Journal of Economics 82, 33–53. Denis, D.J., 1991. Shelf registration and the market for seasoned equity offerings. Journal of Business 64, 189–212. Dunbar, C.G., 1995. The use of warrants as underwriter compensation in initial public offerings. Journal of Financial Economics 38, 59–78. Dutordoir, M. and Hodrick, L.S., 2012. Self-selection and stock returns around corporate security offering announcements. Working Paper. Dutordoir, M., Strong, N. and Ziegan, M.C., 2013. Does corporate governance influence convertible bond issuance? Journal of Corporate Finance (forthcoming). Eckbo, B.E., 1986. Valuation effects of corporate debt offerings. Journal of Financial Economics 15, 119–151. Eckbo, B.E., Maksimovic, V. and Williams, J., 1990. Consistent estimation of cross-sectional models in event studies. Review of Financial Studies 3, 343–365. Erel, I., Julio, B., Kim, W. and Weisbach, M.S., 2012. Macroeconomic conditions and capital raising. Review of Financial Studies 25, 341–376. Esho, N., Kollo, M.G. and Sharpe, I.G., 2004. Eurobond underwriter spreads. Accounting and Finance 46, 71–95. Fang, L.H., 2005. Investment bank reputation and the price and quality of underwriting services. Journal of Finance 60, 2729–2761. Gande, A., Puri, M. and Saunders, A., 1999. Bank entry, competition, and the market for corporate securities underwriting. Journal of Financial Economics 54, 165–195.

139 Golubov, A., Petmezas, D. and Travlos, N.G., 2012. When it pays to pay your investment banker: New evidence of the role of financial advisors in M&As. Journal of Finance 67, 271–311. Gomes, A. and Phillips, G., 2012. Why do public firms issue private and public securities. Journal of Financial Intermediation 21, 619–658. Green, R.C., 1984. Investment incentives, debt and warrants. Journal of Financial Economics 13, 115–136. Hamilton, B.H. and Nickerson, J.A., 2003. Correcting for endogeneity in strategic management research. Strategic Organization 1, 51–78. Hansen, R.S., Fuller, B.R. and Janjigian, V., 1987. The over-allotment option and equity financing flotation costs: An empirical investigation. Financial Management 16, 24–32. Hansen, R.S. and Torregrosa, P., 1992. Underwriter compensation and corporate monitoring. Journal of Finance 47, 1537–1555. Heckman, J.J., 1979. Sample selection bias as a specification error. Econometrica 47, 153– 161. Investment Dealer’s Digest, 1997. As rule 144a offerings multiply, SEC considers fine-tuning system. Huang, R. and Ramirez, G.G., 2010. Speed of issuance, lender specialization, and the rise of the 144a debt market. Financial Management 39, 643–673. Jones, J.J., 1991. Earnings management during important relief investigations. Journal of Accounting Research 29, 193–228. Kidwell, D.S., Marr, M.W. and Thompson, G.R., 1987. Shelf registration: competition and market flexibility. Journal of Law and Economics 30, 181–206. Kim, D., Palia, D. and Saunders, A., 2008. The impact of commercial banks on underwriting spreads: Evidence from three decades. Journal of Financial and Quantitative Analysis 43, 975–1000. Kim, D., Palia, D. and Saunders, A., 2010. Are initial returns and underwriting spreads in equity issues complements or substitutes? Financial Management 39, 1403–1423. Lee, L-F., 1982. Some approaches to the correction of selectivity bias. Review of Financial Studies 49, 355–372. Lee, G. and Masulis, R.W., 2009. Seasoned equity offerings: Quality of accounting information and expected flotation costs. Journal of Financial Economics 92, 443–469. Lee, I., Lochhead, S., Ritter, J. and Zhao, Q., 1996. The costs of raising capital. Journal of Financial Research 19, 59–74.

140 Lewis, C., Rogalski, R. and Seward, J., 1999. Is convertible debt a substitute for straight debt or for common equity? Financial Management 28, 5–27. Li, K. and Prabhala, N.R., 2007. Self-selection models in corporate Finance, in Eckbo, B.E., (ed.): Handbook of corporate finance: Empirical corporate finance, Vol. 1, Handbook of finance series, North-Holland, Elsevier. Livingston, M. and Miller, R.E., 2000. Investment bank reputation and the underwriting of nonconvertible debt. Financial Management 29, 21–34. Livingston, M. and Zhou, L., 2002. The impact of rule 144a debt offerings upon bond yields and underwriter fees. Financial Management 31, 5–27. Lucas, D.J. and McDonald, R.L., 1990. Equity issues and stock price dynamics. Journal of Finance 45, 1019–1043. Marsh, P., 1982. The choice between equity and debt: An empirical study. Journal of Finance 37, 121–144. McKie-Mason, J., 1990. Equity issues and stock price dynamics. Journal of Finance 45, 1019–1045. Mikkelson, W.H. and Partch, M.M., 1986. Valuation effects of security offerings and the issuance process. Journal of Financial Economics 15, 31–60. Myers, S.C., 1977. Determinants of corporate borrowing. Journal of Financial Economics 5, 147–175. Myers, S.C. and Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13, 187–221. Petersen, M.A., 2009. Estimating standard errors in finance panel datasets: Comparing approaches. Review of Financial Studies 22, 435–480. Rogowski, R.J. and Sorensen, E.H., 1985. Deregulation in : Shelf registrations, structure and performance. Financial Management 16, 5–15. Sherman, A.E., 1999. Underwriter certification and the effect of shelf registration on due diligence. Financial Management 28, 5–19. Tinic, S.M., 1988. Anatomy of initial public offerings of common stock. Journal of Finance 43, 789–822. Train, K., 2009. Discrete choice models with simulation, 2nd Ed. Cambridge University Press. Wasserfallen, W. and Wydler, D., 1988. Underpricing of newly issued bonds: Evidence from the Swiss capital market. Journal of Finance 43, 1177–1191.

141 Underwriter spreads (%) 7

6

5

4

3

2 Non-shelf equity Shelf equity 1 Non-shelf convertibles 144a convertibles 0 Shelf convertibles Non-shelf bonds 144a bonds Shelf bonds

Figure 1: Underwriter spreads Underwriter spreads for samples of non-shelf equity, shelf equity, non-shelf convertibles, shelf convertibles 144a convertibles, non-shelf bonds, 144a bonds and shelf bonds. The numbers plotted are reported in table 3 for issues from 1999 to 2011.

142 Total direct costs (%)

12

10

8

6

4

2 Non-shelf SEOs Shelf SEOs 0 Shelf convertibles Non-shelf convertibles Non-shelf bonds Shelf bonds

Figure 2: Total direct costs Total direct costs as a percentage of gross proceeds for samples of non-shelf equity, shelf equity, non-shelf convertibles, shelf convertibles, non-shelf bonds and shelf bonds. The numbers plotted are reported in table 3 for issues from 1999 to 2011.

143 Table 1 Descriptive statistics for security issues

The table reports the number and percentages of issues by year and offering volumes for samples of non-shelf equity, shelf equity, non-shelf convertible bond, shelf convertible bond, 144a convertible bond, non-shelf straight bond, shelf straight bond, and 144a straight bond issues offered by U.S. industrial companies between January 1999 and September 2011. Data on all issues are from the SDC Global New Issues database. Issue year Traditional equity Shelf equity Non-shelf convertible bonds Number of Percentage Offering Number of Percentage Offering Number of Percentage Offering issues volume ($ issues volume ($ issues volume ($ millions) millions) millions) 1999 53 0.17 5,310 10 0.01 1,519 3 0.08 571 2000 67 0.22 13,247 14 0.02 2,802 5 0.13 3,043 2001 23 0.07 1,886 30 0.03 4,962 3 0.08 1,116 2002 32 0.10 6,529 38 0.04 4,614 0 0.00 0 2003 25 0.08 2,331 68 0.08 9,712 0 0.00 0 2004 22 0.07 2,070 104 0.12 14,906 1 0.03 35 2005 10 0.03 375 84 0.09 13,401 0 0.00 0 2006 12 0.04 822 67 0.08 11,237 2 0.05 259 2007 15 0.05 878 73 0.08 17,985 10 0.25 3,171 2008 5 0.02 765 42 0.05 13,459 8 0.20 4,984 2009 20 0.06 3,052 157 0.18 26,404 3 0.08 840 2010 10 0.03 606 101 0.11 15,109 2 0.05 795 2011 14 0.05 1013 99 0.11 12,409 3 0.08 1,253 N 308 100 38,884 887 100 148,519 40 100 16,067

No. of firms 292 586 38

144 Issue year Shelf convertible bonds 144a convertible bonds Non-shelf straight bonds Number of Percentage Offering Number of Percentage Offering Number of Percentage Offering issues volume ($ issues volume ($ issues volume ($ mllions) mllions) mllions) 1999 0 0.00 0 15 0.09 5,265 5 0.08 1,857 2000 1 0.01 500 20 0.13 7,965 1 0.02 499 2001 5 0.05 2,933 34 0.21 17,655 2 0.03 315 2002 1 0.01 403 18 0.11 4,673 1 0.02 123 2003 7 0.07 2,794 30 0.19 6,868 0 0.00 0 2004 6 0.06 1,588 17 0.11 3,822 0 0.00 0 2005 5 0.05 1,542 5 0.03 945 2 0.03 1,243 2006 8 0.09 3,121 3 0.02 5,609 1 0.02 975 2007 13 0.14 4,895 6 0.04 5,354 2 0.03 3,952 2008 10 0.11 2,335 1 0.01 173 0 0.00 0 2009 24 0.26 6,335 6 0.04 2,973 4 0.06 10,050 2010 9 0.10 3,089 4 0.03 1,248 13 0.21 16,563 2011 5 0.05 963 1 0.01 60 32 0.51 32,949 N 94 100 30,498 160 100 62610 63 100 68,526

No. of firms 86 150 52

145 Issue year Shelf straight bonds 144a straight bonds Number of issues Percentage Offering volume ($ millions) Number of issues Percentage Offering volume ($ millions)

1999 40 0.03 26,041 26 0.26 11,125 2000 55 0.05 22,567 12 0.12 13,274 2001 97 0.08 52,170 37 0.37 38,672 2002 113 0.09 44,048 14 0.14 4,929 2003 95 0.08 44,653 9 0.09 3,419 2004 46 0.04 26,060 1 0.01 350 2005 43 0.04 23,626 0 0.00 0 2006 69 0.06 47,418 0 0.00 0 2007 96 0.08 91,015 0 0.00 0 2008 107 0.09 117,737 0 0.00 0 2009 185 0.15 152,885 1 0.01 243 2010 164 0.14 132,588 0 0.00 0 2011 90 0.08 94,206 1 0.01 235 N 1,200 100 875,014 101 100 72,247

No. of firms 396 89

146 Table 2 Variable descriptions

This table defines the underwriter spread and security choice determinants and their data sources. Variables are in alphabetical order and are measured at the end of the fiscal year before the security offering, unless noted otherwise. Variable Definition Source Abnormal accruals Abnormal accruals obtained using the Compustat modified Jones (1991) model. See Appendix 1.

Debt maturity Logarithm of the years to final SDC maturity

EBIT Earnings before interest and taxes Compustat scaled by total assets

GFC dummy Dummy variable taking the value SDC one if the offering falls in the time of and after the Global Financial Crisis (2008 to 2011)

Hotmarkets Issuancevolumeofthesecuritytype SDC in the month before the offering over issuance volume in months −4 to −2 before issuance

Industry target debt ratio difference Difference between the firm’s debt Compustat ratio and the SIC 2-digit industry median debt ratio

Issue frequency Logarithm of the total number of SDC security issues of a firm in the sample period (1999–2011)

1/issue proceeds 1/total proceeds SDC

Issue proceeds/market value Total proceeds scaled by market value SDC

Issue rating Index following de Jong et al. SDC (2012) assigning a value of one to an S&P AAA rating and adding one for each subsequent lower rating

Market return volatility (%) Daily market return volatility over CRSP the window −240 to −40 days before issuance

NYSE dummy Dummy variable taking the value SDC one if the issuing company lists on the NYSE Reputable underwriter Dummy taking the value one if a SDC reputable underwriter (top 8 underwriters in terms of market share for each security) underwrites the issue

Senior debt Dummy variable taking the value SDC one if the issue is senior debt

147 Table 2 (continued)

Stock price run-up Average daily stock return minus CRSP the average return for the CRSP equally-weighted market index over the window −76 to −2 days before issuance

Stock return volatility Daily stock return volatility over CRSP the window −240 to −40 days before issuance

Trading volume Average monthly trading volume CRSP in the six months before the issue. Nasdaq trading volumes are divided by two to correct for double counting

Treasury Bond yield Three-month US Treasury Bill Datastream yield before the offering date

Total assets Logarithm of the book value of Compustat total assets

Years of call protection (%) Years with call protection relative SDC to the bond’s maturity

148 Table 3 Direct costs as a percentage of gross proceeds for different security classes and distribution mechanisms

This table reports descriptive statistics for the direct costs as a percentage of gross proceeds for the security offerings of all companies in the non-shelf equity, shelf equity, non- shelf convertible bond, shelf convertible bond, 144a convertible bond, shelf bond, and 144a bond samples. It reports differences in direct costs for different levels of offering proceeds. N is the number of observations in the respective category, GS is the underwriter spread, E are other direct expenses and TDC are total direct costs. Equity Non-shelf Shelf Proceeds N GS N E TDC N GS N E TDC (millions)

2−9.99 23 6.51 21 5.43 11.89 22 6.31 21 2.77 9.10 10−19.99 14 6.22 14 2.86 9.08 61 5.80 56 3.10 8.90 20−39.99 35 5.95 31 1.55 7.49 109 5.37 100 0.90 6.27 40−59.99 42 5.36 38 1.03 6.48 113 5.00 102 0.87 5.93 60−79.99 53 5.53 39 0.79 6.46 96 5.01 89 0.58 5.66 80−99.99 30 5.07 24 0.90 5.99 78 5.10 70 0.70 5.90 100−199.99 65 5.10 44 0.56 5.64 217 4.49 200 0.38 4.85 200−499.99 36 4.47 24 0.20 4.67 134 3.36 117 0.26 4.07 500−up 10 3.64 5 0.06 3.48 57 2.98 51 0.11 3.07

Total 308 5.34 240 1.35 6.79 887 4.69 806 0.77 5.51

149 Convertible bonds Non-shelf Shelf 144a Proceeds N GS N E TDC N GS N E TDC N GS N E TDC (millions)

2−9.99 0 0 0 0 0 0 10−19.99 0 0 0 0 0 0 20−39.99 1 3.25 1 0.87 4.12 0 0 0 0 40−59.99 0 0 2 3.69 2 1.11 4.80 0 0 60−79.99 0 0 1 3.00 1 0.38 3.38 9 3.12 0 80−99.99 1 3.13 0 6 2.61 5 0.49 2.97 7 2.38 0 100−199.99 10 2.69 8 0.48 3.27 21 2.90 18 0.38 3.24 59 2.96 3 2.31 5.22 200−499.99 20 5.04 16 0.32 4.32 47 2.57 36 0.35 2.93 49 2.67 2 0.35 3.23 500-up 8 2.06 5 0.10 2.15 17 2.36 12 0.16 2.45 36 2.90 2 0.14 11.11

Total 40 3.77 30 0.35 3.67 94 2.64 74 0.36 2.98 160 2.84 7 1.13 6.34

Straight bonds Non-shelf Shelf 144a Proceeds N GS N E TDC N GS N E TDC N GS N E TDC (millions)

2−9.99 0 0 9 0.59 0 1 0.63 0 10−19.99 0 0 9 0.62 0 1 3.89 0 20−39.99 0 0 11 0.74 1 0.58 2.22 2 1.53 0 40−59.99 1 3.50 1 0.50 4.00 14 0.64 7 0.25 1.06 0 0 60−79.99 1 0.63 0 5 0.67 1 0.16 0.74 0 0 80−99.99 0 0 9 0.78 6 0.22 1.04 2 0.48 0 100−199.99 6 1.53 4 0.57 2.21 102 0.97 64 0.36 1.36 17 1.39 0 200−499.99 23 0.80 16 0.34 1.11 543 0.86 439 0.20 1.05 46 1.15 1 0.40 2.90 500-up 31 0.79 20 0.11 0.99 498 0.75 424 0.17 0.91 32 0.91 1 0.07 0.72

Total 62 0.91 41 0.25 1.23 1200 0.82 942 0.20 1.01 101 1.13 2 0.23 1.81

150 Table 4 Descriptive statistics for underwriter spread determinants

This table reports descriptive statistics for the underwriter spread determinants of all companies in the non-shelf equity, shelf equity, 144a convertible bond, shelf bond, and 144a bond samples (panel A). Panel B reports t-statistics for pairwise differences in means of the underwriter spread determinants. Table 2 gives the definition and source of all variables. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variable (1) Non-shelf equity (2) Shelf equity (3) 144a convertible bonds (4) Shelf straight bonds (5) 144a straight bonds Panel A Stock price run-up (%) 0.32 (0.18) 0.22 (0.18) 0.24 (0.18) −0.01 (−0.01) 0.01 (−0.01) Abnormalaccruals 0.20(0.09) 0.14(0.07) 0.07(0.05) 0.05(0.03) 0.07(0.04) NYSEdummy 0.18(0.00) 0.37(0.00) 0.31(0.00) 0.92(1.00) 0.76(1.00) Tradingvolume(thousands) 376(144) 1,003(242) 1,120(484) 3,809(1916) 1,424(526) Hotequitymarkets 0.28(0.25) 0.28(0.24) 0.28(0.25) 0.26(0.25) 0.26(0.25) Hotconvertiblemarkets 0.29(0.26) 0.29(0.25) 0.35(0.29) 0.28(0.23) 0.31(0.30) Stockreturnvolatility(%) 4.97(4.58) 4.31(3.70) 4.36(4.21) 2.35(2.03) 3.28(2.96) Leverage 0.17(0.09) 0.27(0.22) 0.25(0.18) 0.25(0.23) 0.28(0.26) EBIT −0.18 (−0.01) −0.13 (0.02) 0.02 (0.05) 0.11 (0.11) 0.10 (0.10) TBYield 3.17(3.83) 1.68(0.95) 2.89(2.19) 2.01(1.41) 3.76(4.37) Hotdebtmarkets 0.26(0.25) 0.26(0.25) 0.26(0.25) 0.25(0.25) 0.26(0.25) Issuerating(Index) n/a n/a 17.74(21.00) 8.26(8.00) 10.36(9.00) Debt maturity (years log) n/a n/a 2.25 (1.95) 2.19 (2.30) 2.09 (2.30) Seniordebt(dummy) n/a n/a 0.54(1.00) 0.96(1.00) 0.85(1.00) Years of call protection (%) n/a n/a 0.51 (0.43) 0.22 (0.00) 0.34 (0.00) Issueproceeds/marketvalue 0.44(0.26) 0.32(0.20) 0.23(0.18) 0.10(0.05) 0.26(0.12) 1/issueproceeds 0.04(0.01) 0.02(0.01) 0.01(0.00) 0.01(0.00) 0.00(0.00) Totalassets(millions) 1,192(135) 1,657(325) 3,715(894) 20,148(8294) 10,487(2187) Marketreturnvolatility(%) 1.00(0.95) 1.23(0.99) 1.01(1.08) 1.21(1.07) 1.06(1.10) Reputableunderwriter(dummy) 0.44(0.00) 0.55(1.00) 0.70(1.00) 0.88(1.00) 0.68(1.00) Issuefrequency 0.67(0.69) 1.10(1.10) 0.99(0.90) 1.89(1.95) 1.30(1.39) Industrytargetdebtratiodifferences 0.09(0.00) 0.16(0.08) 0.16(0.07) 0.09(0.07) 0.11(0.07)

151 Panel B Variable Difference (1) and (2) Difference (1) and (3) Difference (1) and (4) Difference (1) and (5) Difference (2) and (3) Stock price run-up (%) 0.00*** 0.00 0.00*** 0.00*** −0.52 Abnormalaccruals 0.06*** 0.12*** 0.15*** 0.13** 0.07*** NYSE dummy −0.19*** −0.12*** −0.73*** −0.58*** 0.07* Trading volume (thousands) −627*** −744*** −3,412*** −1,048*** −117 Hot equity markets 0.00 0.00 0.02*** 0.02* −0.00 Hot convertible markets −0.00 −0.06*** 0.00 −0.02 −0.05*** Stock return volatility (%) 0.01*** 0.01*** 0.03*** 0.02*** −0.00 Leverage −0.10*** −0.08*** −0.08*** −0.11*** 0.02 EBIT −0.05** −0.19*** −0.29*** −0.27*** −0.14*** TB Yield 1.49*** 0.28 1.16*** −0.58*** −1.21*** Hotdebtmarkets 0.00 0.00 0.01* 0.00 0.00 Issue rating (Index) n/a n/a n/a n/a n/a Debt maturity (years log) n/a n/a n/a n/a n/a Senior debt (dummy) n/a n/a n/a n/a n/a Years of call protection (%) n/a n/a n/a n/a n/a Issueproceeds/marketvalue 0.12*** 0.21*** 0.34*** 0.18** 0.09*** 1/issueproceeds 0.02*** 0.03*** 0.03*** 0.03*** 0.02*** Total assets (millions) −466 −2,523** −18,957*** −11,137*** −2,058*** Market return volatility (%) −0.00*** −0.00 −0.00*** −0.00 0.00*** Reputable underwriter (dummy) −0.10*** −0.26*** −0.44*** −0.24*** −0.15 Issue frequency −0.43*** −0.32*** −1.22*** −0.63*** 0.11* Industry target debt ratio differences −0.07*** −0.07*** −0.00 −0.02 −0.00

152 Variable Difference (2) and (4) Difference (2) and (5) Difference (3) and (4) Difference (3) and (5) Difference (4) and (5) Stock price run-up (%) 0.00*** 0.00*** 0.00*** 0.00*** −0.00 Abnormal accruals 0.09*** 0.07*** 0.02 0.00 −0.01 NYSE dummy −0.54*** −0.39*** −0.61*** −0.46*** 0.15*** Trading volume (thousands) −2,805,319*** −420,863 2,688,637*** −304,180 2,384,456*** Hot equity markets 0.02*** 0.02* 0.02** 0.02 −0.00 Hot convertible markets 0.01 −0.02 0.06*** 0.04 −0.03 Stock return volatility (%) 0.02*** 0.01*** 0.02*** 0.01*** −0.01*** Leverage 0.02** −0.01 −0.00 −0.04 −0.03** EBIT −0.24*** −0.22*** −0.10*** −0.08*** 0.03* TB Yield −0.33*** −2.07*** 0.88*** −0.86*** −1.74*** Hot debt markets 0.01 −0.00 0.00 −0.00 −0.01 Issue rating (Index) n/a n/a 9.48*** 7.39*** −2.09*** Debt maturity (years log) n/a n/a 0.07 0.16** 0.09 Senior debt (dummy) n/a n/a −0.42*** −0.31*** 0.11*** Years of call protection (%) n/a n/a 0.29*** 0.18*** −0.11*** Issue proceeds / market value 0.22*** 0.06 0.13*** −0.03 −0.16*** 1/issueproceeds 0.02* 0.02*** 0.00 0.00 0.00 Total assets (millions) −18,491*** −10,671*** −16,433*** −8,614*** 7,820** Market return volatility (%) 0.00 0.00** −0.00*** −0.00 0.00** Reputable underwriter (dummy) −0.33*** −0.14*** −0.18*** 0.02 0.20*** Instruments Industry targetdebt ratio differences 0.07*** 0.05* 0.07*** 0.05 −0.02 Issue frequency −0.79*** −0.20** −0.90*** −0.31*** 0.59***

153 Table 5 Multinomial logistic regression analysis of the choice between non-shelf equity, shelf equity, 144a convertible bonds, shelf straight bonds and 144a straight bonds

This table reports coefficients and p-values from multinomial logit regressions for the choice between non-shelf equity, shelf equity, 144a convertible bonds, shelf straight bonds and 144a straight bonds. Non-shelf public equity is the base outcome. Table 2 gives the definition and source of all variables. The McFadden R 2 and observation numbers refer to the entire multinomial model. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Relative to a base outcome of non-shelf equity Variables Shelf equity 144a convertibles Shelfbonds 144abonds Intercept 1.97*** 1.20 −9.51*** −11.98*** (0.00) (0.40) (0.00) (0.00) Stock price run-up −0.88 6.34 −72.80*** −124.16*** (0.95) (0.81) (0.01) (0.01) Abnormal accruals −0.09 −1.79* 1.21*** 0.68 (0.58) (0.07) (0.00) (0.17) NYSE dummy 0.32 −0.92*** 1.42*** 0.45 (0.13) (0.00) (0.00) (0.19) Trading volume 0.00 −0.00 −0.00 −0.00 (0.23) (0.36) (0.59) (0.11) Hot equity markets −0.13 −2.09* −1.59 −3.52** (0.88) (0.10) (0.12) (0.02) Hot convertible markets 0.38 1.98*** 0.53 0.97 (0.51) (0.00) (0.42) (0.27) Stock return volatility −14.16*** 10.62 −43.66*** 12.33 (0.00) (0.13) (0.00) (0.20) Leverage 1.43* −3.47*** 0.01 1.53 (0.10) (0.01) (0.99) (0.29) EBIT −0.54* 0.29 3.88*** 4.91*** (0.07) (0.68) (0.00) (0.00) TB Yield −0.31*** −0.35*** −0.15** 0.07 (0.00) (0.00) (0.02) (0.43) Hot debt markets -2.10* −2.89 −2.14 2.45 (0.10) (0.15) (0.17) (0.30) Issue proceeds / market value −0.38*** −1.33*** 0.71* 1.00*** (0.01) (0.01) (0.08) (0.00) 1 / issue proceeds −5.34*** −207.68*** 2.24 −38.69 (0.00) (0.00) (0.33) (0.32) Total assets (log) 0.16* 0.42*** 1.60*** 1.28*** (0.06) (0.01) (0.00) (0.00) Market return volatility (%) −39.74** 39.41 −9.91 175.70*** (0.05) (0.32) (0.70) (0.00) GFC dummy 1.32*** −2.26*** 1.66*** −4.83*** (0.00) (0.00) (0.00) (0.00) Reputable underwriter (dummy) -0.09 −0.22 0.16 −0.58** (0.60) (0.38) (0.49) (0.05) Industry target debt ratio differences -0.57 3.27** −0.08 −2.02 (0.52) (0.01) (0.94) (0.19)

McFadden R2 45.05 45.05 45.05 45.05 N 2,656 2,656 2,656 2,656

154 Table 6 Regression analysis of underwriter spreads of non-shelf equity issues

This table reports coefficients and p-values of linear regressions of underwriter spreads of non-shelf equity issues. The dependent variable is the underwriter spread. Regression (1) includes the selectivity correction term from the multinomial model to control for self-selection. Table 2 gives the definition and source of all variables. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) Intercept 8.79*** 7.96*** (0.00) (0.00) Selectivity correction 0.89*** (0.00) Stock price run-up −35.47*** −34.60*** (0.00) (0.00) Abnormal accruals −0.04 −0.09 (0.60) (0.19) NYSE dummy −0.07 −0.19 (0.68) (0.27) Trading volume 0.00 −0.00 (0.62) (0.62) Hot equity markets −0.02 0.12 (0.98) (0.83) Hot convertible markets 0.57 0.39 (0.19) (0.37) Stock return volatility 2.57 6.41*** (0.29) (0.01) Leverage 0.56** 0.42 (0.05) (0.18) EBIT −0.03 0.16 (0.85) (0.24) TB Yield −0.19*** −0.10*** (0.00) (0.00) Hot debt markets −0.94 −0.09 (0.43) (0.94) Issue proceeds / market value −0.04 0.02 (0.47) (0.72) 1 / issue proceeds −2.29** −1.23 (0.02) (0.23) Total assets (log) −0.28*** −0.46*** (0.00) (0.00) Issue frequency −0.07 −0.10 (0.41) (0.27) Market return volatility (%) −23.19 −8.67 (0.19) (0.60) GFC dummy −0.07 −0.53** (0.79) (0.02) Reputable underwriter (dummy) −0.11 −0.08 (0.33) (0.45)

R2 53.64 52.17 F-test 15.39*** 14.91*** N 308 308

155 Table 7 Regression analysis of underwriter spreads of shelf equity issues

This table reports coefficients and p-values of linear regressions of underwriter spreads of shelf equity issues. The dependent variable is the underwriter spread. Regression (1) includes the selectivity correction term from the multinomial model to control for self-selection. Table 2 gives the definition and source of all variables. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) Intercept 7.41*** 7.31*** (0.00) (0.00) Selectivity correction −0.12 (0.41) Stock price run-up −3.93 −4.37 (0.61) (0.57) Abnormal accruals −0.27 −0.26 (0.15) (0.16) NYSE dummy 0.04 0.04 (0.75) (0.71) Trading volume −0.00 −0.00 (0.23) (0.21) Hot equity markets 0.47 0.44 (0.23) (0.26) Hot convertible markets −0.17 −0.15 (0.46) (0.50) Stock return volatility 4.79** 4.83*** (0.01) (0.01) Leverage 0.20 0.15 (0.24) (0.34) EBIT 0.13 0.12 (0.28) (0.31) TB Yield −0.02 −0.01 (0.67) (0.86) Hot debt markets −0.34 −0.30 (0.56) (0.60) Issue proceeds / market value −0.00 0.01 (0.99) (0.90) 1/issueproceeds 1.35 1.70 (0.30) (0.17) Total assets (log) −0.50*** −0.47*** (0.00) (0.00) Issue frequency −0.18*** −0.18*** (0.00) (0.00) Market return volatility (%) 3.37 4.07 (0.67) (0.61) GFC dummy 0.25* 0.22 (0.08) (0.11) Reputable underwriter (dummy) 0.05 0.05 (0.61) (0.59)

R2 39.77 39.72 F-test 30.99*** 32.79*** N 887 887

156 Table 8 Regression analysis of underwriter spreads of 144a convertible bond issues

This table reports coefficients and p-values of linear regressions of underwriter spreads of 144a convertible bond issues. The dependent variable is the underwriter spread. Regression (1) includes the selectivity correction term from the multinomial model to control for self-selection. Regression (3) includes four debt-specific determinants (senior debt dummy, debt maturity, years of call protection and the rating index). Table 2 defines the variables and gives their source. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) (3) Intercept 2.05** 2.54*** −1.33 (0.02) (0.01) (0.64) Selectivity correction 0.88 (0.17) Stock price run-up −7.04 −1.50 −5.51 (0.72) (0.94) (0.79) Abnormalaccruals 3.71 2.84 2.75 (0.39) (0.47) (0.44) NYSE dummy 0.49* -0.05 −0.01 (0.08) (0.86) (0.97) Trading volume 0.00 −0.00 −0.00 (0.84) (0.59) (0.41) Hotequitymarkets 1.93* 1.78* 1.87** (0.07) (0.07) (0.03) Hot convertible markets −0.48 −0.09 −0.13 (0.15) (0.82) (0.77) Stock return volatility −14.35 −4.64 2.97 (0.41) (0.69) (0.64) Leverage 0.37 0.14 0.20 (0.23) (0.54) (0.40) EBIT 0.62 1.09 0.88 (0.37) (0.22) (0.26) TB Yield 0.14 0.06 0.16 (0.16) (0.37) (0.18) Hot debt markets −0.80 −1.43 −1.51 (0.57) (0.35) (0.38) Issuerating(Index) 0.02 (0.31) Debt maturity (years log) 0.69 (0.32) Seniordebts(dummy) 0.21 (0.25) Years of call protection (%) 1.67 (0.16) Issueproceeds/marketvalue 3.34 2.73 2.13 (0.12) (0.13) (0.13) 1 / issue proceeds 69.08 −21.48 4.16 (0.10) (0.73) (0.93) Total assets (log) 0.09 −0.03 −0.03 (0.53) (0.69) (0.73) Issue frequency −0.19 −0.24 −0.19 (0.33) (0.28) (0.34) Market return volatility (%) −62.94 −45.66 −9.51 (0.23) (0.30) (0.74)

157 Table 8 (continued

GFC dummy 1.37 0.16 −0.46 (0.15) (0.53) (0.27) Reputable underwriter (dummy) 0.68** 0.61** 0.51** (0.02) (0.02) (0.02)

R2 20.64 19.04 24.85 F-test 1.97** 1.85** 1.96*** N 160 160 160

158 Table 9 Regression analysis of underwriter spreads of shelf straight bond issues

This table reports coefficients and p-values of linear regressions of underwriter spreads of shelf straight bond issues. The dependent variable is the underwriter spread Regression (1) includes the selectivity correction term from the multinomial model to control for self-selection. Regression (3) includes four debt-specific determinants (senior debt dummy, debt maturity, years of call protection and the rating index). Table 2 gives the definition and source of all variables. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) (3) Intercept 1.02*** 1.96*** 0.35 (0.00) (0.00) (0.15) Selectivity correction −0.34*** (0.00) Stock price run-up 0.29 4.90 1.17 (0.96) (0.45) (0.84) Abnormal accruals −0.10** −0.20*** 0.05 (0.03) (0.00) (0.10) NYSE dummy 0.04 −0.06 −0.05 (0.54) (0.30) (0.37) Tradingvolume 0.00 0.00*** 0.00** (0.43) (0.00) (0.04) Hot equity markets −0.00 0.04 0.01 (0.99) (0.76) (0.94) Hot convertible markets 0.06 0.08 0.03 (0.45) (0.31) (0.66) Stock return volatility 5.14*** 8.60*** 2.83* (0.01) (0.00) (0.05) Leverage 0.33*** 0.42*** −0.14 (0.01) (0.00) (0.21) EBIT −0.88*** −1.17*** −0.19 (0.00) (0.00) (0.38) TB Yield −0.01 −0.01 −0.00 (0.39) (0.28) (0.58) Hot debt markets 0.11 0.11 0.12 (0.56) (0.55) (0.41) Issuerating(Index) 0.08*** (0.00) Debt maturity (years log) 0.18*** (0.00) Senior debt (dummy) −0.38*** (0.00) Years of call protection (%) 0.13*** (0.00) Issueproceeds/marketvalue 0.43*** 0.41*** 0.29*** (0.00) (0.01) (0.00) 1 / issue proceeds −0.98* −1.81*** −1.15*** (0.06) (0.00) (0.01) Total assets (log) −0.06** −0.14*** −0.05** (0.03) (0.00) (0.02) Issuefrequency 0.05** 0.04** 0.04** (0.02) (0.03) (0.01) Market return volatility (%) −8.52*** −10.00*** −0.24 (0.01) (0.00) (0.93)

159 Table 9 (continued)

GFC dummy 0.14*** 0.07* −0.03 (0.00) (0.08) (0.29) Reputable underwriter (dummy) −0.00 −0.05 0.01 (0.96) (0.27) (0.88)

R2 34.46 33.50 53.55 F-test 32.66*** 33.05*** 61.68*** N 1,200 1,200 1,200

160 Table 10 Regression analysis of underwriter spreads of 144a straight bond issues

This table reports coefficients and p-values of linear regressions of underwriter spreads of 144a straight bond issues. The dependent variable is the underwriter spread. Regression (1) includes the selectivity correction term from the multinomial model to control for self-selection. Regression (3) includes four debt-specific determinants (senior debt dummy, debt maturity, years of call protection and the rating index). Table 2 gives the definition and source of all variables. Coefficients are expressed as percentages. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Variables (1) (2) (3) Intercept −0.14 2.40*** 0.73 (0.92) (0.01) (0.45) Selectivity correction −0.67* (0.07) Stock price run-up −30.99 −7.49 −13.99 (0.17) (0.70) (0.46) Abnormal accruals −0.50 −0.51 −0.24 (0.29) (0.29) (0.61) NYSEdummy 0.40** 0.49** 0.48** (0.05) (0.02) (0.01) Tradingvolume 0.00 0.00** 0.00** (0.61) (0.03) (0.04) Hot equity markets −1.57 −0.67 −0.88 (0.12) (0.51) (0.36) Hot convertible markets 0.96* 0.53 0.26 (0.05) (0.27) (0.58) Stock return volatility 1.69 −3.20 −3.18 (0.79) (0.56) (0.58) Leverage 1.53*** 1.55*** 0.94** (0.00) (0.00) (0.04) EBIT −2.12** −2.92*** −1.45 (0.04) (0.00) (0.18) TB Yield 0.10* 0.03 0.01 (0.08) (0.53) (0.78) Hot debt markets 2.29* 1.09 1.39 (0.08) (0.35) (0.16) Issuerating(index) 0.10*** (0.01) Debt maturity (years log) −0.00 (0.99) Seniordebt(dummy) 0.05 (0.87) Years of call protection (%) 0.10 (0.60) Issueproceeds/marketvalue 1.01*** 0.96*** 0.81*** (0.00) (0.00) (0.00) 1 / issue proceeds 0.26 9.51 −5.57 (0.98) (0.35) (0.67) Total assets (log) −0.30*** −0.36*** −0.27*** (0.00) (0.00) (0.00) Issue frequency −0.02 −0.03 0.00 (0.85) (0.78) (0.99) Market return volatility (%) 109.69*** 48.64** 58.67*** (0.00) (0.04) −0.01

161 Table 10 (continued)

GFC dummy −1.41* −0.11 −0.49 (0.08) (0.83) (0.37) Reputable underwriter (dummy) −0.16 0.04 −0.00 (0.38) (0.78) (0.99)

R2 66.01 64.19 68.74 F-test 8.28*** 8.16*** 7.80*** N 101 101 101

162 Table 11 Counterfactual underwriter spreads

This table presents the actual and counterfactual underwriter spreads for the samples of non-shelf equity, shelf equity, 144a convertible bond, shelf straight bond and 144a straight bond issues. Counterfactuals underwriter spreads are the underwriter spreads that would have obtained if the same firm had decided to issue another security type instead. They are calculated by multiplying the coefficients of the respective counterfactual security type with the actual corresponding explanatory variables of this firm. Cells (1) to (5) show the actual average underwriter spread for each security type. t-statisics are calculated with respect to the actual average underwriter spread for each security type. The unconditional average is the average fee for each security type across all security issuers. *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels. Issuer type Security type (Non-shelf equity) (Shelf equity) (144a convertibles) (Shelf bonds) (144a bonds) (1)Non-shelf equity issuer (1) 5.34% 5.22% 6.01% 1.70% 1.73% t-stat for difference across row −4.36*** 1.44 −71.20*** −48.40*** t-stat for difference across column 8.84*** 4.86*** 34.85*** 4.15***

(2) Shelf equity issuer 5.79% (2) 4.69% 5.05% 1.46% 0.98% t-stat for difference across row 25.56*** 3.90*** −93.31*** −75.11*** t-stat for difference across column 7.81*** 10.65*** 37.94*** −1.30

(3) 144a convertible bond issuer 4.81% 4.20% (3) 2.84% 1.43% 1.04% t-stat for difference across row 22.16*** 15.97*** −22.16*** −18.93*** t-stat for difference across column −6.27*** −6.30*** 22.60*** -0.83

(4) Shelf straight bond issuer 5.15% 2.80% 4.34% (4) 0.82% 0.19% t-stat for difference across row 278.87*** 76.50*** 56.17*** −34.59*** t-stat for difference across column −5.58*** −48.55*** 8.74*** −16.24***

(5) 144a straight bond issuer 4.06% 3.46% 3.15% 1.35% (5) 1.13% t-stat for difference across row 29.56*** 20.68*** 13.28*** 2.61*** t-stat for difference across column −14.08*** −12.80*** 2.45** 16.20***

Unconditional average 5.04% 3.86% 4.36% 1.29% 1.58%

163 Chapter 5 Conclusion

5.1. Summary of results This thesis presents three essays that individually improve our understanding on the determinants and costs of corporate security offerings. We investigate whether corporate governance influences convertible bond issuance in chapter 2, the signaling content of security offering proceeds in chapter 3 and provide new evidence on the costs of raising capital in chapter 4. In chapter 2, we test three hypotheses on the potential impact of corporate governance quality on convertible bond issuance and announcement returns on a sample of Western European security offerings made between 2000 and 2010. Specifically, we test whether the relationship between corporate governance quality and convertible debt issuance is substitutive or complementary, or whether entrenched managers use convertible debt to insulate themselves from market discipline (Isagawa, 2002). Our main finding is that companies with weaker corporate governance, measured by both internal and external mechanisms, are significantly more likely to issue convertible debt than straight debt and equity. For the internal corporate governance mechanisms, this result mainly holds for large blockholders with strong monitoring incentives. For the external corporate governance mechanisms, we find a negative impact for proxies measuring country-specific legal systems and capital market development. Moreover, we show this finding is robust to measuring corporate governance through composite indices rather than individual proxies. As this finding can either suggest a substitutive relationship between corporate governance quality and convertible debt issuance or management entrenchment, we analyze stock returns around convertible debt announcements. In line with the former, we find convertible debt announcements are negatively influenced by some measures of corporate governance quality. Firms thus seem to use convertible bonds to achieve lower agency and adverse selection costs. In chapter 3, we examine the determinants and signaling content of security offering proceeds for seasoned equity, convertible debt and straight debt offerings by U.S. firms between 1999 and 2011. Controlling for the endogeneity of issue size, we test the impact of expected and unexpected issue size on security offering announcement returns. In particular, we examine

164 whether issue size is related to an expected shortfall in earnings, firm overvaluation, or growth opportunities following the theoretical implications from the models of Miller and Rock (1984), Krasker (1986) and Ambarish et al. (1987). Our results show a positive impact of firms’ funding needs on issue size for all security types and a negative impact of debt- and equity-related financing costs on issue size of more debt- and equity-related securities respectively. For predicted issue size we find a positive impact on announcement returns for both equity and convertible debt. Further test on firms’ use of proceeds show that larger equity(-linked) offerings are more likely to be used to finance growth. Moreover, we find a negative impact of unpredicted issue size for equity and convertible debt and a positive impact for very large issues of straight debt on announcement returns. Further tests on firms’ post-issuance earnings and use of proceeds show that the negative impact of managerial private information on equity and convertible debt announcement returns is related to firm overvaluation, as firms use larger than expected proceeds to build up financial slack. For straight debt issues we find that firms use larger than expected proceeds to fund growth. Chapter 3 studies the magnitude and determinants of direct issuance costs of security offerings on a sample corporate U.S. equity, convertible and straight bond offerings between 1999 and 2011 controlling for firms self-selecting into these security classes and different flotation mechanisms, namely non-shelf, shelf and 144a. We find that direct issuance costs have dropped for all security classes in the last decade compared to previous evidence from the early 1990s by Lee et al. (1996). We find that direct issuance costs are larger for equity than for convertible debt, for which direct issuance costs are again larger than for straight debt. For the different distribution mechanisms our results show larger direct issuance costs for non-shelf than for shelf offerings. Direct issuance costs for 144a offerings are of intermediate magnitude. Moreover, we show that underwriter spreads depend on the overall underwriting effort, i.e. due diligence, pricing and selling, which increases with equity-related adverse selection costs and debt-related financing costs. Controlling for self-selection we find that non-shelf equity issuers have characteristics which make them self-select into this security type, but that these characteristics lead to higher underwriter spreads than any other issuer would have paid, had they offered this security. Conversely, our results show a negative selection effect for both issuers of shelf and 144a straight bonds. These results suggest that fees are partially conditional on observable and unobservable issuer characteristics, so that any comparison of fees across security classes and distribution mechanisms should take into account firms self-selecting into 165 these security types.

5.2. Implications and suggestions for future research This thesis contributes to improving our understanding of the determinants and costs of corporate security offerings. Our findings have important implications for firms’ management, analysts, investors, underwriters and academic researchers. We show that convertible debt can substitute for weak corporate governance in order to reduce firms’ financing costs. We show that investors and analysts can interpret larger than expected security offering proceeds as a signal of firm overvaluation for equity(-linked) security types and as a signal for growth opportunities for debt(-linked) security types. For managers this implies that they should be careful in using their discretion in determining security offering proceeds, as market may interpret issue sizes beyond those expected as a negative signal for certain security types, i.e. equity and convertible debt. We also show that underwriter spreads are not directly comparable across security types, as they are affected by proxies for underwriter effort and by unobservable characteristics guiding the security choice. This allows firms to potentially reduce financing costs with respect to underwriter spreads. The results are also important to academic researchers as they open up new fields for further research on the determinants and costs of corporate security offerings. Future research can examine further aspects of the relationship between convertible bond issuance and corporate governance. Research can examine whether a potential reason for the lack of evidence for Isagawa’s (2002) entrenchment rationale in our data lies in the focus on a Western European setting. We would expect to obtain more evidence for an entrenchment rational for a sample of security issue in emerging markets, where managers and families routinely employ pyramid ownership structures to give themselves control rights that far exceed their proportional cash flows (Harvey et al., 2004). Future research can also study whether there are differences with respect to the use of proceeds for different distribution mechanisms, e.g., it is possible that the overvaluation signal for equity(-linked) issues is stronger for shelf and 144a offerings due to higher timing flexibilities for these issue types. Moreover, research can address the total costs of security issuance considering both direct and indirect costs. We find limited evidence for firms self-selecting into the security type that offers them the absolutely lowest direct financing, whereas findings of Dutordoir and Hodrick (2012) suggest that firms actively choose the financing type with the least negative announcement returns. This leaves room to analyze whether firms try to minimize the aggregate of three cost components, i.e. underwriter 166 spreads, announcement returns, and underpricing. Finally, further research may expand on the models in chapter 3 and chapter 4 by including proxies for corporate governance quality in the respective security choice and announcement return models. As chapter 2 shows that these proxies are significantly related to security choice and announcement returns, our current models in these chapters may potentially suffer from an omitted variable bias.

167 References

Ambarish, R., Kose, J. and William, J., 1987. Efficient signalling with dividends and investments. Journal of Finance 42, 321–343. Dutordoir, M. and Hodrick, L.S., 2012. Self-selection and stock returns around corporate security offerings announcements. Working Paper. Harvey, C.R., Lins, K.V. and Roper, A.H., 2004. The effect of capital structure when expected agency costs are extreme. Journal of Financial Economics 74, 3–30. Isagawa, N., 2002. Callable convertible debt under managerial entrenchment. Journal of Corporate Finance 8, 255–270. Krasker, W.S., 1986. Stock price movements in response to stock issues under asymmetric information. Journal of Finance 41, 93–106. Lee, I., Lochhead, S., Ritter, J. and Zhao, Q., 1996. The costs of raising capital. Journal of Financial Research 19, 59–74. Miller, M.H. and Rock K., 1985. Dividend policy under asymmetric information. Journal of Finance 40, 1031–1051.

168