THE PENNSYLVANIA STATE UNIVERSITY SCHREYER HONORS COLLEGE DEPARTMENT OF FINANCE

Transcription

1 THE PENNSYLVANIA STATE UNIVERSITY SCHREYER HONORS COLLEGE DEPARTMENT OF FINANCE THE EFFECT OF CONTINGENT CONVERTIBLE DEBT ISSUANCE ON THE VOLATILITY OF A COMPANY S RETURNS WILLIAM LUCAS SPRING 2017 A thesis submitted in partial fulfillment of the requirements for a baccalaureate degree in Finance with honors in Finance Reviewed and approved* by the following: Jingzhi Huang Professor of Finance, McKinley Professor of Business, and Professor of Mathematics Thesis Supervisor Brian Davis Clinical Associate Professor of Finance Honors Adviser and Faculty Reader * Signatures are on file in the Schreyer Honors College.

2 i ABSTRACT This study focuses on Eurozone contingent convertible debt issuances from the period 2009 to 2015 and attempts to identify trends in stock volatility and firm risk in a one-year postannouncement period. More specifically, the study focuses on contingent convertible debt issuances of companies that trade on primary exchanges that utilize the Euro as the main currency. This study employed four different analytical techniques. The first three analyses were similar in nature and attempted to find a link between how a company s equity volatility changes after contingent convertible debt issuance. This was measured by comparing the volatility of a company s returns versus those of a benchmark during a one-year period prior to the announcement date and a one-year period following the announcement date (post-announcement period). Equity return data of three different frequencies (daily, weekly, and monthly) were utilized in order to conduct this study. The fourth study employed a slightly different analysis. This study investigated how the idiosyncratic (firm-specific) risk of a company s returns changed in the post-announcement period rather than how overall volatility of returns changed. In theory, by removing market effects, one could see how a firm s inherent risk-profile changed from a contingent convertible debt issuance. Idiosyncratic risk was measured by comparing the average standard deviations of weekly residual return data in the one-year periods prior to and following the announcement date.

3 ii TABLE OF CONTENTS LIST OF FIGURES... iii LIST OF TABLES... iv ACKNOWLEDGEMENTS... v Chapter 1 Introduction... 1 Chapter 2 Literature Review... 2 CoCos and the Post-Recession Regulatory Landscape... 4 Problems with Contingent Convertible Debt... 6 Chapter 3 Methodology Data Sourcing for Volatility Tests Post-Issuance Volatility Change Analysis Post-Issuance Changes in Idiosyncratic Risk Chapter 4 Research Findings and Summary Statistics Study Summary Statistics Daily Volatility Study Results Weekly Volatility Study Results Monthly Volatility Study Results Idiosyncratic Risk Study Results Chapter 5 Conclusion Chapter 6 Suggested Further Study Link Between Stock Return Volatility and CoCo Issuance Contingent Convertible Debt and Market Shocks Benchmark Control Selection Sample Expansion Financial Statement Analysis Breakdown by CoCo Features Appendix A Volatility Studies Sub-sample Result Figures Daily Volatility Study Weekly Volatility Study Monthly Volatility Study Idiosyncratic Risk Study Expanded Time Period Studies... 43

4 iii Appendix B CoCo Issuance Summary Statistics References... 48

5 iv LIST OF FIGURES Figure 1: Structure of CoCos (Avdjiev, Kartasheva, & Bogdanova, 2013)... 3 Figure 2: Pricing of CoCo Bonds (Avdjiev, Kartasheva, & Bogdanova, 2013)... 4 Figure 3: CoCos and Basel III (Avdjiev, Kartasheva, & Bogdanova, 2013)... 6 Figure 4: Sample Issuances by Year Figure 5: Post-Announcement Change in Daily Volatility Figure 6: Companies with Market Cap > 10,000 mil Euros (Daily) Figure 7: Post-Announcement Changes in Weekly Volatility Figure 8: Companies with Market Cap > 10,000 mil Euros (Weekly) Figure 9: Post-Announcement Change in Monthly Volatility Figure 10: Companies with Total Assets > 500,000 mil Euros (Monthly) Figure 11: Post-Announcement Change in Volatility of Residual Returns Figure 12: Companies with Total Assets > 500,000 mil Euros (Daily) Figure 13: Companies with Total Assets > 500,000 mil Euros (Daily) Figure 14: Companies with Total Assets < 500,000 mil Euros (Daily) Figure 15: Companies with Total Assets > 100,000 mil Euros (Daily) Figure 16: Companies with Total Assets < 100,000 mil Euros (Daily) Figure 17: Companies with Market Cap > 10,000 mil Euros (Weekly) Figure 18: Companies with Total Assets > 500,000 mil Euros (Weekly) Figure 19: Companies with Total Assets < 500,000 mil Euros (Weekly) Figure 20: Companies with Total Assets > 100,000 mil Euros (Weekly) Figure 21: Companies with Total Assets < 100,000 mil Euros (Weekly) Figure 22: Companies with Market Cap > 10,000 mil Euros (Monthly) Figure 23: Companies with Market Cap < 10,000 mil Euros (Monthly)... 38

6 v Figure 24: Companies with Total Assets < 500,000 mil Euros (Monthly) Figure 25: Companies with Total Assets > 100,000 mil Euros (Monthly) Figure 26: Companies with Total Assets < 100,000 mil Euros (Monthly) Figure 27: Companies with Market Cap > 10,000 mil Euros (Idiosyncratic) Figure 28: Companies with Market Cap < 10,000 mil Euros (Idiosyncratic) Figure 29: Companies with Total Assets > 500,000 mil Euros (Idiosyncratic) Figure 30: Companies with Total Assets < 500,000 mil Euros (Idiosyncratic) Figure 31: Companies with Total Assets > 100,000 mil Euros (Idiosyncratic) Figure 32: Companies with Total Assets < 100,000 mil Euros (Idiosyncratic)... 43

7 vi LIST OF TABLES Table 1: Summary Research Results Table 2: Sub-sample Daily Volatility Summary Table 3: Sub-sample Weekly Volatility Summary Table 4: Sub-sample Monthly Volatility Summary Table 5: Sub-Sample Idiosyncratic Risk Study Results Table 6: Daily Study Result Summary by Company Table 7: Weekly Study Result Summary by Company Table 8: Monthly Study Result Summary by Company Table 9: Idiosyncratic Risk Study Result Summary by Company Table 10: CoCo Issuances (Bloomberg Data) Table 11: CoCo Issuances by Currency ( January 2017 Bloomberg Data)... 47

8 vii ACKNOWLEDGEMENTS I would like to take the time to thank a number of mentors who have helped me throughout the writing process. First, I would like to thank my adviser, Professor Huang, for all his support and guidance. His flexibility and advice have made this process as smooth and enjoyable as possible. Second, I would like to thank Professor Vanden. Although he is not my official thesis adviser or reader, he has been a key source of knowledge and ideas for this thesis, and I cannot thank him enough for that. Finally, I would like to thank Professor Davis, who has helped guide me through this process since my junior year. On top of helping me to brainstorm ideas, he has served as a source of assurance in times where the writing process was difficult.

9 1 Chapter 1 Introduction The purpose of this thesis is to provide a comprehensive overview of a new and growing tool in the financial sector known as contingent convertible debt (also known as CoCos ). This paper will attempt to explain what contingent convertible debt is, how it works, and who is using it, and will provide a history of the precluding financial turmoil and regulatory changes that led to its inception and creation. The paper will look at this financial instrument from the perspectives of the issuer, investor, and regulator in order to provide an unbiased look at the benefits and disadvantages of contingent convertible debt and the controversy surrounding its implementation and utilization. The paper will then attempt to tackle several questions surrounding the instrument. The two main questions from which this thesis has evolved are: Has CoCo debt issuance at current levels reduced risk for issuing banks? Has CoCo debt issuance reduced investors perceptions of firm risk? The paper will serve as a starting and reference point for an investigation into the contingent convertible debt world, as well as spark new ideas and create a preliminary foundation for identifying and addressing a number of future potential questions.

10 2 Chapter 2 Literature Review The term contingent convertible debt refers to a group of hybrid securities that are issued predominantly by banks and act as loss-absorbing tools for financial institutions in times of financial distress. When the capital of an issuing bank falls below a certain level, the intrinsic loss-absorbing mechanism of these instruments is activated, effectively reducing the firm s debt levels and recapitalizing its balance sheet. CoCos can absorb losses and recapitalize banks in one of two ways: by converting into common equity or by suffering a principal write-down. A conversion to equity increases Tier 1 Common Equity capital by converting at a predetermined rate. The Bank for International Settlement s Quarterly Review (2013) claims that this rate can be based on the stock price at time of trigger, a pre-specified contractual price, or a combination of the two. On the other hand, CoCos with a write-down feature can be written down in part or full (usually full) to raise equity. The trade-offs between equity conversion versus a write-down will be discussed later in the review. Additionally, there can be one or more triggers of this instrument s loss-absorbing nature. Avdjiev, Kartasheva, and Bogdanova (2013) categorize these triggers as either mechanical or discretionary. Mechanical triggers activate the loss-absorbing mechanism when the capital of the bank falls below a predetermined ratio. This ratio involves a comparison between either (i) the bank s book value and its risk-weighted assets or (ii) its market value and its risk-weighted

11 assets. Discretionary triggers, on the other hand, involve the triggering of the loss absorbing 3 mechanisms by a bank s supervisors or regulators. Figure 1: Structure of CoCos (Avdjiev, Kartasheva, & Bogdanova, 2013) In terms of investment potential, CoCos provide a relatively attractive investment in a low-interest rate environment. These instruments are subordinate in the capital structure, and as a result, the yields associated with CoCos tend to be greater than those of other debt instruments. Avdjiev, Kartasheva, and Bogdanova (2013) showed that on average, yields on CoCos were 2.8% higher than those of other subordinated debt and 4.7% higher than those of senior unsecured debt of the same issuer. Yields of high-trigger CoCos are generally higher than those of low-trigger CoCos, as the former are more likely to incur early loss absorption.

12 4 Figure 2: Pricing of CoCo Bonds (Avdjiev, Kartasheva, & Bogdanova, 2013) CoCos and the Post-Recession Regulatory Landscape Even though the new capital requirements put in place by Basel III will most likely bring increased stability to the financial system and will help to improve banks abilities to handle crises, higher equity-capital requirements raise the cost of capital for banks, increasing the potential for credit stress for banks. Bolton and Samama (2011) believe that an answer to this problem is the convertible debt instrument, and state: If ex ante deleveraging by banks is not sufficient to avoid banks in another crisis, then ex post deleveraging through debt restructuring should be facilitated. One way of achieving such ex post deleveraging is to require banks to hold debt instruments that convert into equity in the event of a crisis, or in the event that an institution approaches a dangerous leverage ratio. (p. 278) Such an instrument would basically reduce leverage through a debt-equity swap, similar to a debt restructuring, while also being less costly than a pure common equity requirement. Many academics support the idea that properly implemented contingent convertible debt can reduce financial distress while increasing the value of the firm. Albul, Jaffee, and Tchistyi

13 (2013) found that properly implemented contingent convertible bonds can increase the overall 5 value of a firm, while also reducing the risk of bankruptcy or bailout, depending on the state of the regulatory environment (p. 24). Koziol and Lawrenz (2012) additionally found that in properly regulated environments in which risk shifting incentives are controlled and minimized, CoCos can increase firm value, as well as reduce the risk of financial distress. Goes, Schiozer, and Sheng (2016) furthered the research of Koziol and Lawrenz and applied their model to the ten biggest banks in Brazil, and their findings supported Koziol and Lawrenz s original claims. Furthermore, CoCos meet the main criteria of contingent capital requirements outlined in Basel III, while other types of contingent capital do not. In addition to meeting the necessary requirements to be considered going-concern capital (tier 1), these instruments also possess either a trigger point or write-down feature, which are necessities for contingent securities under Basel III (Basel Committee on Banking Supervision, 2011, p. 17). Equity AT1 instruments, like preferred shares, do not meet the going-concern requirement called for under the Basel III framework. This makes CoCos a better candidate for banks looking for inventive ways to minimize costs while meeting capital requirements. CoCos can qualify as either Additional Tier 1 or Tier 2 capital under the Basel framework, and this distinction depends largely on the trigger level that a firm sets for its contingent debt. CoCos with lower triggers are less costly to issue, but have lower loss-absorbing capacity and only qualify as Tier 2 capital. CoCos with higher trigger levels, in contrast, can qualify as AT1 capital, but are more expensive. Under Basel III, the minimum trigger level of Common Equity Tier 1 to risk-weighted assets required to qualify as AT1 is 5.125%, and banks tend to set their CoCos trigger levels at exactly this point, in order to maximize capital while minimizing costs.

14 6 Figure 3: CoCos and Basel III (Avdjiev, Kartasheva, & Bogdanova, 2013) Problems with Contingent Convertible Debt The purpose of this section is to give a brief overview of the many questions and issues surrounding the implementation of CoCos. I have split the section into a series of smaller parts detailing individual issues (and the order of these problems as presented has no significance). Credit Rating Credit rating information on many CoCo bonds is currently incomplete, as credit rating agencies have shown considerable reluctance to place ratings on these instruments. Credit rating agencies have found it difficult to apply a consistent process to rating these hybrids due to the wide variety of features that they may possess, as well as the differential regulatory and tax treatments that may be applied. S&P has rated some CoCos (but has left a large portion unrated) and the firm has generally ranked these instruments two to three notches below the issuer s credit rating.

15 Price Spiral 7 Academics have studied the concept of a CoCo death spiral, which applies to CoCos with market-value-based triggers. This term refers to stock price manipulation that can occur in a number of ways. Albul, Jaffee, and Tchistyi (2013) pose two instances in which such manipulation can occur. First, they offer a case in which CoCo holders would actually possess incentives to trigger conversion. In this case, an investor would buy a contingent convertible bond above its conversion level and then drive down the stock price of the firm to the point of conversion by spreading negative news or short selling. The investor would then obtain the firm s equity at a low price, giving the investor the chance to sell the equity when the firm s stock price corrects. The second case involves manipulation on part of the equity holders. If the overall conversion value of the contingent convertible debt of a firm (the total value of the amount of equity that the bonds will be converted into) is low in comparison to the value that CoCo holders would receive from future coupons of the CoCos before conversion, equity holders may want the conversion to take place. Since it is essentially cheaper for equity holders to convert CoCos than keep paying out a substantially large coupon, they may possess incentives to negatively manipulate the firm s stock price in order to trigger conversion. In the case that manipulation causes the market value trigger to be breached and conversion from debt to equity is dilutive, the swap could push stock prices even lower. Additionally, Koziol and Lawrenz (2011) state, It is not unlikely that in a situation of asymmetric information, the observation of mandatory conversion will be interpreted as a negative signal by the market, which in turn can put additional pressure on the stock price of the bank (p. 101).

16 Koziol and Lawrenz also provide another interesting case where a negative price spiral 8 could occur. If large institutional investors who invest under restrictions that bar them from holding equity invest in CoCos under the assumption that conversion will not take place, and conversion actually does occur, they will be forced to sell the shares they receive. If the overall value of the forced stock sale is large enough to considerably impact stock price, it can push prices further downward. Conflict of Interest Bank Regulators vs. Investor Regulators. The regulators of issuing banks prefer instruments with greater loss-absorption capacity (higher trigger points) in order to insure a high source of quality capital for banks. On the other side, regulators of buyers of CoCo bonds prefer CoCo issues with lower loss-absorbing capacity (lower trigger points) because they are less costly and the likelihood of conversion or write-down is smaller. This divergence can potentially lead to a mismatch of supply and demand in the market. CoCo Holders vs. Equity Holders. CoCo bond buyers find contingent convertibles with a convert-to-equity feature more attractive than those with a write-down feature, since they recoup partial compensation. Equity holders, on the other hand, obviously prefer CoCos to be written down because this imposes loss strictly on the bondholders and avoids equity dilution. Coupon Interestingly, academic research seems to avoid investigating the possible implications of a CoCo issuer that skips a coupon payment. Issuers of CoCo bonds are allowed to halt coupon payments in times of financial stress, and stakeholders in the company s debt cannot subsequently call for bankruptcy. Fears over Deutsche Bank s ability to pay coupons on its contingent convertible debt sparked a sell-off in the company s CoCo bonds in February 2016,

17 and this was the first real case where these hypothetical instruments were tested in a time of 9 distress. Risk Shifting The idea that CoCo issuance may induce risk shifting incentives on part of a bank s management is somewhat contradictory, and academic research on the subject is somewhat scarce. Early research done by Flannery (2005, 2009) suggests the idea that CoCos strictly reduce the possibility of financial distress while also increasing overall firm value. However, a later study done by Koziol and Lawrenz (2011) contradicts Flannery s findings. Their findings show that CoCos ability to reduce distress risk for a bank is dependent on, if bank managers have substantial discretion over the bank s business risk. (p. 101). Koziol and Lawrenz identify two different cases: if complete contracts can be written and if complete contracts cannot be written (incomplete contracts). They define a complete contract setting as an environment in which the investment policy of a bank is given, or contractible, and managers cannot change the risk technology of the bank. In such a setting, CoCos are unambiguously beneficial (p. 101), and will increase overall firm value while also reducing the probability of financial distress. However, Koziol and Lawrenz found this is not necessarily the case when a bank possesses an incomplete contract setting, where investment policy is not contractible and managers can change investment policy. They found that CoCo financing can actually distort risk taking incentives of managers in non-contractible environments, increasing the probability of financial distress. So how does CoCo financing actually increase a bank s distress risk? This lies in the fact that, straight bonds impose lower risk-shifting incentives on the equity holders than CoCo bonds (Koziol and Lawrenz, p. 98). In the case of default with normal bonds, bondholders lose

18 complete control rights. The authors argue that this possibility of loss of control actually 10 disciplines managers to prevent what they call poor states, or in other words, environments in which bankruptcy can occur. However, with CoCo financing in a state of default, bondholders lose the cash flow they received from coupons, but still maintain ownership rights since they receive an equity stake in the firm through conversion. Managers, recognizing this inherent protection that CoCos provide, may be inclined to increase the riskiness of the bank s strategy in order to increase overall returns. Equity holders will possess similar incentives, since this increases their returns. This increased riskiness can manifest in a wide range of forms, from investing in riskier securities to decreased monitoring of the bank s risk-taking activities. All these actions, however, have the potential to lead to increased risk for the firm. This effect is compounded for banks that are considered too-big-to-fail, since they essentially possess a guarantee that they will receive government support in situations of extreme financial distress and are seen by investors as risk-free. Goes, Schiozer, and Sheng (2014) support this idea that too-big-to-fail banks tend to assume more risk in order to increase returns, as well as claim that investors tend to favor these bigger banks because they believe they are safer. If such risk shifting does occur, it is then possible that there are conditions in which CoCos can still increase firm value through tax shield benefits, while also increasing the probability of financial distress. While Flannery did acknowledge the possibility of a risk shift of this caliber, he argued that the higher premium on CoCo coupons internalizes and fully captures this risk. Koziol and Lawrenz, however, stated this only holds in complete contract settings. In order to counteract this theoretical risk-shifting nature, Koziol and Lawrenz claim that regulators need to pose stricter

19 financial regulation. However, they also acknowledge that in a setting with stricter regulation, 11 CoCo financing can actually decrease a firm s value more than straight bond financing.

20 12 Chapter 3 Methodology After conducting initial due diligence on contingent convertible debt and prior studies around the topic, I hypothesized that despite some fears about the theoretical effectiveness of CoCos, contingent convertible debt issuances should reduce the underlying risk of an issuing bank. Further, I hypothesized that investors, in general, believe that these tools are effective in reducing risk, and as a result, this should be reflected in an equity s performance through a drop in volatility of equity returns. In order to test this hypothesis, I conducted four studies that are detailed in the next few sections of this chapter. Data Sourcing for Volatility Tests Using Bloomberg data, I collected a sample of 93 public companies that have issued contingent convertible debt instruments since the financial crisis. Similarly, stock price, dividend, and total return data on these companies from the beginning of 2007 until the end of 2016 was similarly collected, using FactSet software. Returns were calculated assuming dividends were reinvested and were calculated using the following formula: Return = (Period 2 closing price + dividends issued) / Period 1 closing price Weekly: calculated from Wednesday to following Wednesday* Monthly: calculated from end of month to end of following month *If a company did not trade on a Wednesday, the previous trading day was used

21 13 Next, control groups were created based on the currency that these companies use and the primary stock exchanges on which they trade. The research sample was then narrowed to just Eurozone companies with sufficient pricing data to conduct the study. This left a sample of 21 companies. Five European financial benchmarks were then considered for the study. They included the Euro STOXX / Banks SS, the Euro STOXX / Financials IND, Euro STOXX TMI / Financials IND, the STOXX Europe 600 / Banks SS, and the STOXX Europe 600 / Financials IND. After further review, the STOXX Europe 600 Banks was chosen as the proper benchmark to use as a control for the analysis. This benchmark is simply the banking sub-sector of the larger STOXX Europe 600 index. The total returns for this index were pulled from FactSet. Post-Issuance Volatility Change Analysis To run this analysis, the average daily standard deviation of the sample companies returns over several time periods was calculated. These time periods included: one-year before and after announcement, six months before and after announcement, three months before and after announcement, one month before and after announcement, and one week before and after announcement. The same analysis was then run on the control benchmark. The change in daily standard deviation between the corresponding time periods (e.g., one year prior to announcement versus one year after announcement) was then calculated in order to assess how volatility changed relative to the benchmark. A greater decrease in volatility versus the benchmark or a lesser increase in volatility versus the benchmark could be the first sign of a possible boost in

22 14 financial stability of a company s stock due to a CoCo issuance, all things equal, and thus, could be considered a successful result of the experiment. These same steps were then applied to both weekly and monthly returns. Post-Issuance Changes in Idiosyncratic Risk After comparing overall volatility, the experiment was enhanced by investigating idiosyncratic volatility. A firm s risk can be broken down into systematic and idiosyncratic risk. While systematic risk essentially reflects events that affect the whole market, idiosyncratic risk reflects the risk associated with the firm itself. Idiosyncratic risk could give a better look at how a company s inherent risk-profile changes with a CoCo issuance without market effects skewing results. The overall goal of this analysis was to compare changes in the changes in volatility of residual returns of a firm s stock before and after a CoCo issuance. Based on the work of Shakya (2016), residual returns were calculated as follows: Residual return = Ri αi,m - βi,mrm Ri = weekly risk-adjusted return of firm stock Rm = weekly risk-adjusted return of the STOXX Europe 600 Banks Index αi,m = y-intercept of regression of stock returns against benchmark returns βi,m = overall volatility of stock returns versus the benchmark returns In order to conduct this study, the same data from the overall volatility analysis was used. However, returns were adjusted for risk by using the daily 3-month LIBOR yield, which is the standard risk-free rate applied in Europe. After adjusting returns, a regression was run on a single company s returns against the index returns over time periods of one-year before announcement

23 15 and one-year after announcement. This regression yielded the alpha and beta associated with a company s stock return during the time period specified. This was then repeated for every CoCo bond that a company issued. After finding each alpha and beta, residual returns for every week during each particular time period could be calculated. The final result of the testing was the change in the standard deviations of residual returns, or in other words, the change in idiosyncratic risk after issuance. A fall in idiosyncratic risk would indicate that a CoCo issuance decreased a specific firm s systematic risk, all things equal.

24 # of Issuance 16 Chapter 4 Research Findings and Summary Statistics This section will summarize the findings of the four different studies detailed in Chapter 3. After reviewing initial results, a year of return data seemed to yield the most consistent results, and accordingly, the results in this chapter will discuss the results of the studies over a one-year post-announcement period. A summary table detailing some results of shorter time-period tests can be found in Appendix A. Study Summary Statistics The sample in this study includes 46 total contingent convertible debt issuances by 21 companies that trade on exchanges that use the Euro as the primary currency over the time period of 2010 to A majority of the issuances used in the study were issued between 2013 and Number of Issuances Per Year (Sample) Year Figure 4: Sample Issuances by Year

25 17 Success rates of the four analytical tests were found, with the study of the change in overall weekly standard deviation over a one-year period after announcement yielding the highest success rate of 73.9%. This means that for 73.9% of the issuances analyzed in the sample, the change in volatility of a specific firm s equity returns was less than the change in the volatility of the benchmark s returns. Table 1 summarizes the overall results of all four analytical tests: Daily σ Study Weekly σ Study # of Issuances # of Succesful Success # of Succesful Success Year per Year Results Rate Results Rate % 0 0.0% % % % 0 0.0% % % % % % % % % Monthly σ Study Idiosyncratic Risk Study # of Issuances # of Succesful Success # of Succesful Success Year per Year Results Rate Results Rate % % % % % % % % % % % % % % Table 1: Summary Research Results Comparing the overall volatility (standard deviation) studies, we see varied results, with higher frequencies of return data yielding better results. This was contrary to what I initially believed, as I thought lower frequency return data would help to smooth the volatility of returns

26 Change in Daily σ 18 and provide a better sample for the study. In terms of the change in idiosyncratic risk analysis, the evidence, unfortunately, did not support the initial hypothesis, and further research would need to be done. Firm-specific risk fell across only 56.5% of issuances. In other words, the volatility of residual weekly returns of an individual firm only decreased across 56.5% of sample issuances during the one-year post-announcement study period from the one-year period prior to the announcement date. Daily Volatility Study Results The daily volatility study yielded an overall success rate of 71.7% across all issuances, as depicted in Table 1. Below, Figure 5 illustrates the changes in daily standard deviation of the stock returns of each company in the sample one-year after announcement versus one-year before announcement against the adjusted returns of STOXX Europe 600 banks index during the same time periods: Annual Post-Announcement Change in Daily Volatility STOXX Europe 600 Banks Firm % % % % 50.00% 0.00% % % 7/14/2010 7/14/2011 7/14/2012 7/14/2013 7/14/2014 7/14/2015 Announcement Date Figure 5: Post-Announcement Change in Daily Volatility

27 Change in Daily σ 19 Any point below the line indicates a positive result, while those above the line indicate a negative result. However, by breaking down the sample in sub-samples based on market value and asset level criteria, we start to see slight changes in results: Annual Post-Announcement Change in Daily Volatility (Market Cap > 10,000 mil) STOXX Europe 600 Banks Firm % 80.00% 60.00% 40.00% 20.00% 0.00% % % % 7/14/2010 7/14/2011 7/14/2012 7/14/2013 7/14/2014 7/14/2015 Announcement Date Figure 6: Companies with Market Cap > 10,000 mil Euros (Daily) Figure 6 illustrates annual post-announcement changes in daily volatility of the firms in the sample that possessed a market capitalization of 10,000 million euros or more at the time of issuance. The sample was broken down into multiple other subsets, similar to the one above. These include: firms with market capitalizations greater than 10,000 million euros at time of announcement, firms with market capitalizations less than 10,000 million euros at time of announcement, firms with total asset values greater than 500,000 million euros at time of announcement, firms with total asset values less than 500,000 million euros at time of announcement, firms with total asset values greater than 100,000 million euros at time of announcement, and firms with total asset values less than 100,000 million euros at time of

28 announcement. The charts of these results can be found at the end of the chapter, and the 20 summary table of these result can be found in Table 2 below: Financial Criteria Value # of # of Successful Success Metric (millions of euros) Issuances Results Rate Market Cap > 10, % Market Cap < 10, % Total Asset Value > 500, % Total Asset Value < 500, % Total Asset Value > 100, % Total Asset Value < 100, % Table 2: Sub-sample Daily Volatility Summary An individual can infer that exclusion of smaller companies, in terms of both market cap and asset value, seems to boost overall results, suggesting that the existence of relatively undersized companies skews overall results. Weekly Volatility Study Results The weekly volatility study yielded the highest success rate of all tests, boasting a rate of 73.9% across all issuances in the sample. These results can be viewed on the following page in Figure 7:

29 Change in Weekly σ 21 Annual Post-Announcement Change in Weekly Volatility STOXX Europe 600 Banks Firm % % % % % % 50.00% 0.00% % % Announcement Date Figure 7: Post-Announcement Changes in Weekly Volatility The graph can be interpreted in the same way as that in the daily volatility study, with any points below the line representing a lesser change in a firm s standard deviation than the benchmark s standard deviation over the one-year post-announcement period. Also, similar to the daily volatility study, breaking down the sample by size-oriented financial metrics yields better results. A breakdown by market value, using a requirement of market capitalization greater than 10,000 euros, again, yielded the best results, and a graphical representation of these results can be found in Figure 8 on the following page:

30 Change in Weekly σ 22 Annual Post-Annoouncement Change in Weekly Volatility (Market Cap > 10,000 mil) STOXX Europe 600 Banks Firm 80.00% 60.00% 40.00% 20.00% 0.00% % % % Announcement Date Figure 8: Companies with Market Cap > 10,000 mil Euros (Weekly) This market capitalization analysis produced close to a 10% boost in the success of results, yielding a success rate of 83.3%. The summary table of the sample breakdown results can be found below in Table 3: Financial Criteria Value # of # of Successful Success Metric (millions of euros) Issuances Results Rate Market Cap > 10, % Market Cap < 10, % Total Asset Value > 500, % Total Asset Value < 500, % Total Asset Value > 100, % Total Asset Value < 100, % Table 3: Sub-sample Weekly Volatility Summary The graphs of the rest of the results of these sub-samples can be found in Appendix A. Again, as in the daily volatility study, the exclusion of smaller companies seems to yield a greater success rate, suggesting the existence of smaller companies in the sample skews the data.

31 Change in Monthly σ Monthly Volatility Study Results 23 While decreasing the frequency of data used to conduct the study seemed to slightly boost the success rate of the weekly study against the daily study, there was a noticeable decline in the success rate of results when monthly data was used to conduct the study. While the daily and weekly studies yielded success rates of 71.7% and 73.9%, the standard deviation of equity returns in the 1-year post-announcement period only decreased more or increased less versus the benchmark across 58.7% of the issuances in the monthly study. Annual Post-Announcement Change in Monthly Volatility STOXX Europe 600 Banks Firm % % % % % % 0.00% % Announcement Date Figure 9: Post-Announcement Change in Monthly Volatility Again, any points below the line represent a successful result, while any above the line indicate an unsuccessful result. When comparing the Figures 5 and 7, one can also see a slight uptick in changes in volatility as a whole during the post-announcement period, for both the individual firms and the benchmark. Additionally, while breaking down the sample by size seemed to provide a broad boost to overall success rates in the daily and weekly volatility studies, this case is much weaker for the monthly study. While market cap seemed to be the best

32 Change in Monthly σ 24 financial metric to utilize in breaking down the sample in the two higher frequency data studies, a breakdown by assets provided the largest boost in this study. Annual Post-Announcement Change in Monthly Volatility (Total Assets > 500,000 mil) STOXX Europe 600 Banks Firm % % % 50.00% 0.00% % % Announcement Date Figure 10: Companies with Total Assets > 500,000 mil Euros (Monthly) While organization by total assets boosted results, this boost was approximately half the magnitude as the boosts in the first two studies. A review of all segmentation results can be found below in Table 4: Financial Criteria Value # of # of Successful Success Metric (millions of euros) Issuances Results Rate Market Cap > 10, % Market Cap < 10, % Total Asset Value > 500, % Total Asset Value < 500, % Total Asset Value > 100, % Total Asset Value < 100, % Table 4: Sub-sample Monthly Volatility Summary One can see that across the board, organizing by size-oriented financial metrics did not seem to help results very much. Additionally, success rates for small-sized firm subsets were comparable to success rates of large companies, which was not the case for the daily and weekly

33 Change in Weekly σ (Residual) 25 studies where small company issuances seemed to skew results. This contradicts the idea that the study is more effective for larger companies. Idiosyncratic Risk Study Results While the overall standard deviation event studies provided a holistic look at how a company s risk changed in comparison to the market s risk over a one-year period following a contingent convertible debt issuance, the idiosyncratic event study attempted to control for the overall market s influence on individual firms equity volatilities. This was achieved by comparing the standard deviations of each firm s weekly residual returns in the one-year beforeannouncement period and the one-year post-announcement period. Any decrease in a particular volatility would indicate a decrease in idiosyncratic risk, and thus, a result that supports my hypothesis. This analysis can be summarized in Figure 11 below: Annual Post-Announcement Change in Idiosyncratic Risk % % % % % % 50.00% 0.00% % % Change in Idiosyncratic Risk Announcement Date Figure 11: Post-Announcement Change in Volatility of Residual Returns

34 Any non-positive change on the scatter plot (any point above zero-line) indicates a 26 successful result. As depicted in Table 1, this success rate was only 56.5%. The sample was additionally broken down into the same sub-samples as in the three volatility studies, and the results of this breakdown are summarized in Table 5 below: Table 5: Sub-Sample Idiosyncratic Risk Study Results The results of this study were comparable to those of the monthly standard deviation study, and actually yielded the overall lowest success rate of 56.5%. Breaking down the sample by market cap slightly boosted the success rate to 60% for the larger market cap companies. However, breaking down the sample by assets actually decreased the overall success rate for large-sized companies, while increasing the success rate for the samples that included relatively smaller-sized companies. Again, this contradicts the idea that the study works better for larger companies. Financial Criteria Value # of # of Successful Success Metric (millions of euros) Issuances Results Rate Market Cap > 10, % Market Cap < 10, % Total Asset Value > 500, % Total Asset Value < 500, % Total Asset Value > 100, % Total Asset Value < 100, %

35 27 Chapter 5 Conclusion CoCo issuances, in essence, are supposed to reduce the risk of an issuing bank by acting as loss-absorbing tools in times of financial crisis, and the instruments are attractive to issuing banks, investors, and regulators alike. However, CoCos are new instruments, and therefore are largely untested in times of distress, which has led some to theorize that these instruments have the potential to have adverse effects on a company s equity performance and balance sheet. This study attempts to dive deeper into this theory, and looks at how investors, as whole, perceive the issuance of these instruments by analyzing how contingent convertible debt issuances may affect the volatility of a bank s equity returns over a one-year post-issuance time period. Forty-six issuances by public companies in the time span were analyzed in order to test how a stock s returns reacted to an issuance. In terms of the overall standard deviation studies, using different return frequencies yielded varying results. While the lower frequency (daily and weekly) volatility studies yielded positive results approximately 80% (80.0% and 83.3%, respectively) of the time after being adjusted for market capitalization restrictions, the monthly study yielded only a 60.0% success rate after market value adjustment. Segmentation by asset value produced similar results, varying approximately ± 3% from the result of the market capitalization segmentation study. An analysis of each firm s underlying idiosyncratic volatility was then used to control market effect. Similar to the standard deviation studies, time periods of one year before and after

36 28 announcement were used to assess change in the volatility of a firm s residual returns. This test produced results similar to that of the monthly standard deviation study, and actually yielded a success rate of 60.0% after market capitalization adjustment. Asset-based segmentation decreased results by approximately 4.5%, less than a no segmentation result of 56.5%. While my initial hypothesis claimed that contingent convertible debt issuances should, as a whole, decrease a bank s stock risk and the volatility of a bank s equity returns, my results were inconclusive in answering my initial question of whether contingent convertible debt issuances have reduced the risk perception of issuing banks by investors. While the 80% success rate may look promising, I believe it is premature to make a conclusion based on only six years of issuance data. Additionally, these results may be randomly assorted as there seems to be no trends or patterns that indicate why a certain bank s results may be more successful than others, and therefore, results may be insignificant. After conducting this study, I believe underlying changes in risk may be better analyzed from a balance sheet perspective, and a summary of suggested further study can be found in Chapter 6 of the thesis.

37 29 Chapter 6 Suggested Further Study Contingent convertible debt issuances are still a relatively new financial instrument, and as a result, quantitative studies that have been conducted are scarce and generally, highly theoretical. My study was another attempt to try to find hard quantitative evidence that proves that these instruments are serving the purpose for which they were invented. My results, however, do not fully support my initial hypothesis and further research is needed. After conducting this study, I have devised a few questions have that could be topics of further study. Link Between Stock Return Volatility and CoCo Issuance The evidence found in my contingent convertible debt study may not fully explain changes in stock volatility. While the purpose of contingent convertible debt issuance on a high level is to decrease overall risk exposure for an issuing bank, this does not necessarily always translate to how a firm s stock reacts. I conjecture that this can happen for a number of reasons. First, CoCos are a relatively new instrument, and many investors may not either know what these instruments are or how they work, or possibly both. As a result, this may not trigger any reaction from these investors to change their behavior toward the stock after a CoCo issuance. Second, even if investors do understand how these instruments work and believe that CoCos will reduce a bank s risk, this may not translate to a change in the stock s return volatility. However, it may affect outright performance on a higher level. By this, I mean that if investors perceive the

38 30 company is less risky, they may feel safer investing their money in the company, and if a large enough population possesses this sentiment, this could boost returns. If this is in fact true, then it may be beneficial to look at abnormal returns after the issuance, rather than the underlying volatility of these returns. This type of abnormal returns analysis could also apply to the faction of investors who feel that CoCo issuance can actually have an adverse effect on a company s risk profile. Contingent Convertible Debt and Market Shocks The original event that drew me to contingent convertible debt issuances in the first place was the downward spiral of Deutsche Bank s CoCo bonds in early In this case, a negative news event that had the potential to have significant implication on Deutsche Bank s bottom line caused many investors to start selling the stock. This scared CoCo investors, as they feared that Deutsche Bank may skip an interest payment on the instruments, and this led to a huge selloff in the company s CoCos. Additionally, this instilled more fear in equity investors, as a negative price spiral could lead to conversion of CoCos, which could lead to overall earnings per share dilution. Yields skyrocketed as the stock price continued to fall. As I watched this event, I was intrigued by how wildly the company s stock was acting during the period, which eventually led me to conduct my study. However, I did not make the connection at that point that what triggered the stock price spiral was a mix of news and CoCos working together. What I now realize would be an interesting study would be to investigate whether the existence of CoCo bonds on a bank s balance sheet actually increase financial difficulty in times of distress. However, in order to conduct this study, a market trigger or shock

39 would need to occur to the company first. Unfortunately, there are extremely limited cases to 31 investigate in order to conduct such a study. It may take years for enough market events to occur in order to conduct such a study. Benchmark Control Selection In this particular study, the STOXX Europe 600 Banks was used as the control. While this benchmark is a solid tool for replicating the European banking sector, it may not have been the best choice for the experiment conducted. A majority of the companies in the study are constituents of the benchmark, and as a result, have influence over how the overall benchmark moves and performs, possibly skewing results. In further studies, it could be beneficial to identify which companies in the STOXX Europe 600 Banks have not issued CoCos, and then create a basket of these companies to use as the control. Alternatively, one could use a different benchmark. However, based off my own research, I feel that a majority of benchmarks will pose the same problem just discussed. Sample Expansion The sample in this study consisted strictly of companies that trade on exchanges that use the Euro. However, there are approximately 70 other public companies that have also issued CoCos outside the Eurozone. Combined with the other suggestions posed above, expanding the sample could possibly yield better results.

40 Financial Statement Analysis 32 In this study, how investors perceive the risk of a company s stock after a contingent convertible debt issuance was analyzed. However, a problem previously discussed, is that the link between issuance and returns may not be strong. As a result, it may be better to analyze CoCos effectiveness from a balance sheet and income statement perspective. An example of this type of analysis would be to analyze how a bank s z-score changes after issuance. This may give a better perspective on how the bank s underlying risk actually changes with CoCo issuance. Breakdown by CoCo Features As explained in the literature review section of this paper, CoCos are versatile instruments and possess a number of unique features. Further study can be conducted based on these features. An individual could focus his/her study on conversion trigger level, type of conversion (equity vs. write-down), amount issued as a ratio of overall assets/liabilities, and more.