Convertible Bond Calls

Transcription

1 The Difference Between Out-Of-The-Money and In-The-Money Convertible Bond Calls Ken L. Bechmann a Department of Finance Copenhagen Business School Current version: January 22, 2001 Key words: Convertible bond calls; Announcement effects; Long-term stock returns JEL Classification: G14; G32 a Solbjerg Plads 3, DK-2000 Frederiksberg, Denmark Phone: Fax: kb.fi@cbs.dk The author owes thanks to Giovanni Barone-Adesi, Bruce D. Grundy, Bjarke Jensen, Peter Løchte Jørgensen, Ken Nyholm, Kristian Rydqvist, Ilya Strebulaev, and the participants in the 2000 EFMA Conference and the 2000 EFA Conference for comments and suggestions. Sanjoy Ghosh is acknowledged for help with accessing the data and Sophia Barkat and Malene Bechmann for excellent research assistance. The author acknowledges financial support from the University of Aarhus Research Foundation. The most recent version of this paper is available at:

2 The Difference Between Out-Of-The-Money and In-The-Money Convertible Bond Calls JEL Classification: G14; G32 Abstract A large amount of literature provides strong evidence for a negative announcement effect in connection with in-the-money convertible bond calls. In contrast to this, only one study have considered out-of-the-money convertible bond calls where the announcement effect was found to be positive. The two different stock market reactions to convertible bond calls raise two primary questions regarding the firms making the calls. First, why are so many firms willing to make in-the-money calls when it is only out-of-the-money calls that are associated with a positive announcement effect? Second, what can explain the two different stock market reactions? In order to answer these questions, a more thorough examination and comparison of outof-the-money and in-the-money calls is required. This paper is the first to make such a study by examining a large data-set including both types of calls. The study documents several important differences between the two types. First, the study confirms the earlier results with respect to the announcement effect. Second, the in-the-money calls seem to be larger and to occur earlier relative to the maturity date of the bonds. Finally, the return on the stocks of in-the-money calls are higher than both the return on the value weighted market index and the return on stocks for out-of-the-money calls in a period of 400 days before the calls. However, after the call the situation is reversed. During a 670-day period after the call, the stocks of out-of-the-money calls have higher returns than the stocks of in-the-money calls. 1

3 1 Introduction Many studies have examined calls of convertible bonds and found evidence for a stock market reaction to the announcement of such calls. 1 However, with one exception these studies have only considered in-the-money calls, i.e. calls where the value that will be obtained upon conversion is larger than the call payment offered by the firm. For such calls, the existing literature provides strong evidence for a negative announcement effect. Different explanations have been given for this announcement effect: bad news about the firms prospectives, short selling induced price pressure, underwriting of the calls, loss of tax shields etc. In contrast to these results, the study in Cowan, Nayar, and Singh (1993) of out-of-the-money calls documented a positive and significant announcement effect. Cowan, Nayar, and Singh (1993) explained the announcement effect as being primarily due to good news revealed by the call. The two different stock market reactions raise two primary questions regarding the firms making out-of-the-money and in-the-money calls. First, why are so many firms willing to make in-the-money calls when it is only out-of-the-money calls that are associated with a positive announcement effect? Second, what can explain the two different stock market reactions? In order to answer these questions, a more thorough examination and comparison of out-ofthe-money and in-the-money calls is required. This paper is the first to make such a study by examining the differences between these two types of calls with respect to their characteristics, their announcement effects, and their long-term stock performance before and after the call. This will be done using the largest data-set of convertible bond calls examined in the literature so far and by examining some aspects of the calls that have not been considered earlier. The rest of the paper is organized as follows. Section 2 briefly describes some details related to convertible bond calls and defines out-of-the-money and in-the-money calls more formally. Section 3 describes the different explanations for the stock market reaction to convertible bond calls and the related literature for both out-of-the-money and in-the-money calls. Section 4 describes the methodology and the data-set used in this paper. Section 5 presents the empirical results for the announcement effects while section 6 considers the long-term stock returns. Section 7 discusses further research and concludes the paper. 1 The existing literature is described in section 3. 2

4 2 Convertible bond calls Convertible bonds are hybrid securities having elements of both ordinary bonds and common stock. On one hand convertible bonds are interest paying and have a promised payment at maturity just like ordinary bonds. On the other hand, convertible bonds have a stock component because the holder of a convertible bond has an option to exchange the bonds for new common stock in the firm. The option to convert is an American style option that matures at the same time as the bond. The terms under which the bond can be converted into stock are given in form of either the conversion price or the conversion ratio. The conversion price denotes the price in $ face value that will be paid per share received upon a conversion while the conversion ratio denotes the number of shares that will be received per face value of the bond. Both terms are given in the bond indenture and are adjusted for stock splits and stock dividends. In addition to the conversion option, the majority of convertible bonds is also what is known as callable. When a bond is callable, it means that it is possible for the issuing firm to redeem the bonds prematurely. 2 The firm redeems the bonds by mailing a notice of redemption (the call announcement) to each bondholder in which the firm offers to buy each bond for the socalled call payment. The call payment, also known as the effective call price, is the sum of the interest accrued since the last interest payment and a term calculated based on a pre-specified value known as the call price. The call price denotes the percent of the face value that has to be offered to bondholders in addition to the accrued interest and is normally a number higher than 100% decreasing over time towards 100%. If the bond is called, it will still be possible for bondholders to convert into stock. The announcement of the call starts the conversion period (or notice period) in which each bondholder must decide whether to convert into stock or to accept the call payment. The length of the conversion period varies but is on average around 35 calender days. The call announcement will state the last date where conversion is possible and the call date (redemption date), where the unconverted bonds will be redeemed and the corresponding call payments made. The call date is often the same date as the last conversion date but can also be a later date. Out-of-the-money calls are calls where the conversion value, i.e. the value of the shares that will be received upon a conversion is below the call payment at the time of the call meaning that 2 However, many convertible bonds are non-callable for a period, typically the first three years after they are issued or until the stock price has reached a certain level. 3

5 it will not be optimal to convert into new shares (the conversion option is out-of-the-money). 3 Therefore, an out-of-the-money call is not expected to lead to new shares being issued but the bonds are expected to be redeemed by cash being paid to the bondholders. In contrast, in-the-money calls (or conversion forcing calls) are calls where the conversion value is higher than the call payment at the time of the call. Therefore, in the case of in-themoney calls the bonds are expected to be converted into new shares in the firm unless there is a stock price decline during the conversion period that makes the conversion value smaller than the call payment. This also explains why these calls are known as conversion forcing calls. 3 The stock market reaction to convertible bond calls 3.1 Out-of-the-money calls As mentioned in the introduction, to the best of our knowledge only one paper has looked at out-of-the-money calls. Cowan, Nayar, and Singh (1993) examined out-of-the-money calls using a data-set of 26 calls. For this data-set they found that stock prices and the price of the called convertible bond react significantly and positively to the announcement. In addition, Cowan, Nayar, and Singh suggested some possible explanations for the positive announcement effect and discussed some arguments that instead would have predicted a negative announcement effect. These explanations and a few additional arguments are presented and discussed in the following. The arguments that predict a negative announcement effect are as follows. First, by calling the bond, the firm will lose the tax shield attached to the bond. This could cause a decrease in the firm value and thereby also in the value of common stock. Second, the call comes too early relative to the optimal call policy derived in for example Brennan and Schwartz (1977) and Ingersoll (1977). If, at the time of the call, the call payment paid to the bond holders is above the price of the convertible bond, the call will transfer wealth to the bondholders. 4 Finally, as argued by Ross (1977), Jensen (1986), and others, debt can in general have a positive effect on firm value. Therefore, one could expect a negative announcement effect when a convertible bond 3 More precisely, the conversion value at a certain point in time is the current stock price times the conversion ratio. 4 As mentioned, the call payment is normally more than 100% of the face value of the bond. Therefore, if the call is unexpected it is possible that the call payment is higher than the price at which the bonds trades before the call. 4

6 issue is called. On the other hand, there are also arguments for why the announcement effect of out-of-themoney calls should be positive. First, the call can be good news because it shows that the firm has access to enough cash to redeem the bonds. 5 Second, the firm may call the debt in order to get rid of some restrictive covenants attached to the bond issue. 6 Third, the call may be good news when the firm is simply financing the call using for example a new convertible bond issue, issued at lower costs. Fourth, if bondholders are hedging the equity risk associated with the convertible bonds by short selling the underlying stock as argued in Bechmann (1999), this suggests a new explanation for the positive announcement effect: An out-of-the-money call will make the bondholders buy shares in the market in order to close out their short positions. This will create an upward price pressure leading to the positive announcement effect. Finally, based on asymmetric information, it is possible that the call announcement reveals good news about the firm to the stock market. This will for example be the case if the management of the firm has positive private information about the firm and this causes the management to call the bonds. One reason for a management with positive information to call the convertible bonds is that the management this way avoid the dilution of the shares outstanding that will follow from a later conversion of the bonds into new shares. 7 Another reason could be that the management wants to avoid calling as soon as the bonds become in-the-money because the call in this case could be taken as bad-news as argued by Harris and Raviv (1985) cf. below. 3.2 In-the-money calls The first study to examine calls of convertible bonds was Mikkelson (1981). Mikkelson (1981) found a significant negative stock market reaction to announcements of in-the-money calls of convertible bonds but not for convertible preferred stocks. 8 Based on this, Mikkelson argued 5 The Free Cash Flow theory of Jensen (1986) would strengthen this in the cases where the firm in this way pays out free cash. 6 Vu (1986) suggested this as the reason for why firms call non-convertible debt. 7 In several of the out-of-the-money calls, the avoid-dilution argument is given as the reason for the call. For example, on February 8, 1989 MCI Communications Corp. called a $400 million issue. The call payment was $1, while the conversion value was only $964.12, i.e. the call was out-of-the-money. As reason for the call, the Chief Financial Officer of MCI said that the redemption will eliminate the potential dilutive effect of conversion of the issue to common stock and will strengthen MCI s balance sheet. 8 Convertible preferred stocks are preferred stocks that just like convertible bonds are convertible into common stocks. Preferred stocks are stocks with a certain level of promised dividends. The promised dividends should be 5

7 that the announcement effect is due to a loss of the tax shield which is only associated with convertible bonds. However, several studies have later found that the announcement effect also exists for convertible preferred stock and that the loss of the tax shield does not seem to be the reason for the announcement effect. 9 Instead, Harris and Raviv (1985) provided a signaling model which suggests that the calls reveal negative information to the stock market about the firm and that this is the cause of the negative announcement effect. The idea behind the signaling model in Harris and Raviv (1985) is that only managers with negative private information about the future will call a convertible bond issue. Managers with positive private information will delay the call because they expect that the bondholders themselves will find it optimal to convert into stock in the future. Ofer and Natarajan (1987) claimed to find empirical evidence for this signaling model because the stock prices seem to perform relatively poorly in the time following the call. However, several studies like Campbell, Ederington, and Vankudre (1991) and Cowan, Nayar, and Singh (1990) subsequently showed that the findings in Ofer and Natarajan (1987) suffer from a selection bias. This selection bias will be further discussed later in this paper. Furthermore, Byrd (1992) and Byrd and Moore (1996) showed that there are no evidence for the bad-news explanation in the way financial analysts react to the announcements. However, some evidence for the bad-news explanation is provided by Datta and Iskandar- Datta (1996). Datta and Iskandar-Datta (1996) showed that the calls lead to a significant loss in total firm value. Furthermore, the calls lead to a wealth transfer from stockholders to holders of straight debt in the firm because of a reduction in leverage. Therefore, the negative announcement effect is argued to be caused by both bad-news and a wealth transfer. The findings and explanations given in Singh, Cowan, and Nayar (1991) are also related to the argument that the call reveals new information to the stock market. Singh, Cowan, and Nayar (1991) showed that the call announcement effect for in-the-money calls depend on whether the call was underwritten or not. 10 They found a significant announcement effect only for the underwritten calls and argued that this is caused by either an agency conflict between the underwriter and the calling firm or that underwriters are used primarily by firms with unfavorable paid before dividends can be paid to common stocks. 9 See Mais, Moore, and Rogers (1989), Byrd (1992), Cowan, Nayar, and Singh (1992), and Byrd and Moore (1996) for studies of calls of convertible preferred stocks. 10 A underwritten call is a call where an investment bank guarantees that all bonds are converted into new shares. 6

8 information. The findings and arguments in Mazzeo and Moore (1992) suggest a reason for the announcement effect which differ from the bad-news explanation. Mazzeo and Moore (1992) documented that stock prices recover during the conversion period following the announcement of the call and argued therefore that the announcement effect is due to price pressure caused by investors selling the shares obtained upon conversion. Bechmann (1999) also provided support for the price pressure explanation but argued instead that the price pressure is caused by bondholders who hedge the equity risk associated with the convertible bonds by short selling stock. The pattern in the short selling of the underlying stock is documented to be related to the call of the convertible bond. The following will discuss some of the different explanations for the out-of-the-money and in-the-money calls in the light of the empirical evidence. 4 Methodology and data-set Section 4.1 describes the methodology used to examine the announcement effect in section 5 and to examine the long-term stock returns in section 6. Thereafter, section 4.2 presents the data-set used in this paper including a range of summary statistics. 4.1 The methodology In order to examine the effect of an announcement, the stock returns must be adjusted for the general market movements. However, market adjusted returns can be calculated in several different ways and it can always be discussed which method would be more appropriate to use in the different cases. 11 In addition, if an estimation period is required in order to derive the parameters used to calculate the normal returns, this will also raise the question of which estimation period to use. As shown in Cowan, Nayar, and Singh (1990) and Campbell, Ederington, and Vankudre (1991), the estimated parameters in a market model will depend on whether the estimation 11 Different methods are described in for example Fama et al. (1969), Dimson and Marsh (1986), Agrawal, Jaffe, and Mandelker (1992), and Campbell, Lo, and MacKinlay (1997). 7

9 period is before or after the call announcement. 12 For this reason and because we wish to say something about the long-term stock performance, it is desirable that the method used to obtain the market adjusted returns does not depend on an estimation period. Furthermore, because daily expected returns are close to zero the way the adjusted returns are obtained does not have a big effect on the results when examining stock returns over only very short time horizons. 13 So when examining the announcement effect (here over two days) in section 5 we will just use a very simple market adjustment of returns, where the adjusted stock return, AR i,t, for stock i during period t is given by AR i,t = R i,t R m,t. Here R i,t denotes the return for stock i during period t while R m,t denotes the return on a market index. The value weighted market index (VWRETD) obtained from the CRSP-file will be used as the market index. Given that the returns are aligned in event time such that the announcement of the call happens at day t = 0, the announcement effect for firm i is calculated as the sum of the adjusted return for day 0 and day 1, i.e. CAR i,ann CAR i,0:1 = AR i,0 + AR i,1. The reason for including both the return for day 0 and day 1 is that it is possible that the announcement of the call in the news wires happens late in day 0 such that the stock market effect only will be seen at day 1. The average announcement effect for N firms in a group of calls is then calculated as CAR ANN = 1 N N CAR i,ann. (1) i=1 In order to examine the announcement effect we will perform several tests. First, a standard t-test is used to test if there is a significant announcement effect for the two types of calls. Second, a sign test as described in Campbell, Lo, and MacKinlay (1997, p. 172) is used to test if the announcement effect is significant based on a nonparametric test. Third, a standard t-test for the difference in the means between two samples is used to test if there is a significant difference in the announcement effect between out-of-the-money and in-the-money calls. 12 The use of an estimation period before the call creates a downward bias in the stock returns after the call. This is the reason why the findings in Ofer and Natarajan (1987) are biased downwards as mentioned earlier. 13 This is discussed in Fama (1997) in connection with the use of event studies to examine long-term stock performance. 8

10 Regarding the long-term stock returns, there is unfortunately no standard and agreed-upon method to examine these. In particular, the problem as to which benchmark should be used and the way the adjusted stock returns should be obtained and measured has been the topic for much discussion. However, the purpose of this paper is not to examine if stocks corresponding to one or both types of the calls under- or overperform relative to similar stocks. Instead, the purpose is to examine if there is a difference between the stock returns before and after the call for firms making out-of-the-money calls and firms making in-the-money calls. Therefore, the following will restrict the analysis to different ways of measuring stock returns relative to the value weighted market index (VWRETD) on the CRSP-file. First, for different time periods relative to the call announcement day we estimate the relation between the return for stock i and the market index using the market model. More precisely, for each stock i and for each of the different time periods we run the following ordinary least squares regression R i,t = α i + β i R m,t + ɛ i,t. (2) The results from this regression are used to examine the difference in the long-term stock returns between the two types of calls but also to examine if there is a change over time in the returns. The pattern in the returns over time is also derived by calculating the cumulative average raw return and the cumulative average adjusted return. For each day t in the period considered, the cumulative average adjusted return from day t 0 to day t is calculated as CAR t0 :t = 1 N = 1 N N i=1 N i=1 CAR i,t0 :t t s=t 0 AR i,s. Similarly, the cumulative average raw return, CR t0 :t is defined as CAR t0 :t but with the raw return, R i,s instead of the adjusted return AR i,s. Finally, because the CAR-approach implicitly assumes portfolio rebalancing, it is often argued that this approach may not be suitable for long-term studies. 14 Instead, a similar method based on wealth relatives is suggested. Therefore, we will also examine the long-term stock performance 14 Refer for example to Cowan and Sergeant (1997) and Lyon, Barber, and Tsai (1999). 9

11 using wealth relatives. For firm i at time t the wealth relative is the wealth, W i,t, obtained from investing $1 in the firm s stock at an earlier point in time relative to the wealth, W m,t that would have been obtained by investing $1 in the market portfolio at the same time. This means that if we let t 0 be the starting point for the investment, we have Thereby, we can write the wealth relative as t W i,t = (1 + R i,s ) s=t 0 t W m,t = (1 + R m,s ). s=t 0 W i,t W m,t = = t s=t 0 (1 + R i,s ) t s=t 0 (1 + R m,s ) t s=t 0 (1 + R i,s) (1 + R m,s ). (3) The time pattern in the wealth relatives can then be examined by calculating and plotting the average just as for the cumulative average returns. 4.2 The data-set The data-set of convertible bond calls is obtained by an extensive search for complete calls announced on the Dow Jones News Wires in the period from 1986 to found by this search, we select those that fulfill the following selection criteria: From the calls The calling firm should have the shares traded on either NYSE, ASE, or NASDAQ. The bond called should only be convertible into stock in the calling firm. 16 Information about conversion terms and face value outstanding should be available either from the call announcement or from Moody s or Standard & Poor s bond guides. 15 A complete call is a call where the whole convertible bond issue is called. It also happens that firms make a partial call where only a fraction of an issue is called. 16 Sometimes convertible bonds are convertible into stocks of other firms, warrants, preferred stocks, or into combinations of these. 10

12 This leads to a total of 539 calls distributed over time as shown in table I. Table I shows that there is no severe clustering of observations in certain years even though there clearly are some years with fewer observations than in other years. In addition, the table also shows that the fraction of out-of-the-money calls varies from year to year but there does not seem to be any clear trend in the use of out-of-the-money calls relative to in-the-money calls. Year Total number of calls Fraction out-of-the-money (%) Table I: Distribution through time of the sample of 539 convertible bond calls in the period from 1986 to 1996 and the fraction of these calls that are out-of-the-money. As mentioned above, information about conversion terms and the size of the called issue is obtained from either the call announcement or from Moody s or Standard & Poor s bond guides. Information about stock returns, share prices, and the number of shares outstanding is obtained from the CRSP-file Summary statistics Summary statistics for the data-set divided into out-of-the-money calls and in-the-money calls are given in table II. From the table we make several observations regarding convertible bond calls and the differences between the two types of call. As mentioned earlier, we have that slightly more than 20% of the sample is out-of-the-money calls. This study is the first to present a number on the relative use of out-of-the-money calls. It is therefore difficult to say whether 20% is surprisingly low or high. However, based on the amount of research looking at in-the-money calls relative to the research looking at out-of-themoney calls, the 20% may be surprisingly high. One factor that can explain if this fraction is higher than expected based on earlier research is that nearly all other studies of convertible bond calls require that the calls are announced in the Wall Street Journal. This is not the case for 11

13 Out-of-the-money In-the-money Difference (OTM) (ITM) ITM OTM Number of observations Fraction of total sample (%) Fraction of the subsample being: underwritten calls (%) 0 22 related to takeovers (%) 9 7 Fraction of the subsample listed on NYSE (%) ASE (%) NASDAQ (%) Mean Median Mean Median t-test,p-value Face value of called issue ($ millions) , 6% Conv. value/call payment , < 0.1% New shares/shares outstanding (%) , 0.5% Total call payment/value of equity (%) , 16% Years left to maturity , < 0.1% Table II: Descriptive statistics for the 539 convertible bond calls in the period from 1986 to The calls are divided into two sub-samples dependent on whether the calls are out-of-the-money or in-the-money. The Difference presents a standard t-test for a difference in means between the two sub-samples. The face value of called issue is the face value outstanding of the called convertible issue in million dollars. Conv. value/call payment is the conversion value divided by the call payment. The conversion value is based on the stock price two days prior to the call while the call payment is the call price plus accrued interest. New shares/shares outstanding is the number of new shares that would be issued if all the bonds were converted into new shares divided by the number of shares outstanding before the call. Total call payment/value of equity is the total amount of money that should be paid in order to redeem all the bonds for cash divided by the total value of the equity before the call. Years left to maturity is the number of years left to maturity of the bond at the time of the call. this study where the calls are just required to be announced on the news wires. Hence, if there is a tendency that only in-the-money calls are mentioned in the Wall Street Journal this would explain if the fraction of out-of-the-money calls is higher in the present data-set than observed in other data-sets of convertible bond calls. As expected the table also shows that there are no underwritten out-of-the-money calls while there are 95 underwritten in-the-money calls corresponding to 22% of these calls. Regarding the stock exchange on which the calling firm is listed there is a tendency to a higher fraction of NYSE firms in the sample of out-of-the-money calls. Table II also presents more detailed information about the called convertible bond issue. First, we observe that both types of calls on average are quite big calls, that upon full conversion 12

14 would increase the number of shares outstanding with 10 14%. Second, we also observe that the total call payment corresponds to around 7 11% of the total value of equity. Third, we find that the out-of-the-money calls on average are 24% out-of-the-money at the time of the call while on average the in-the-money calls are 47% in-the-money. Finally, we observe that the bonds are called several years before maturity. The table also presents some interesting results with respect to the differences between the two types of calls. The test for a difference in means between the two sub-samples provides mixed results regarding the size of the called issue. If the face value of the called issue is used as size measure there are indications of out-of-the-money calls being larger than the in-the-money calls with a p-value of 6%. The same is the case when total call payment/value of equity is used as size measure but now the p-value is only 16%. However, these results are probably due to a few very large and influential calls in the sample of out-of-the-money calls because the median in both cases is larger for the in-the-money calls. Similarly, if the new shares/shares outstanding is used as size measure instead, we get that the in-the-money calls are larger than the out-of-the-money calls with a p-value of 0.5%. Therefore, the conclusion about the size of the calls is that except for a few very big calls in the sample of out-of-the-money calls, the in-the-money calls are larger than the out-of-the-money calls. The years left to maturity also reveals a difference between the two types of calls. The inthe-money calls have more years left to maturity meaning that these calls occur earlier in the life of the convertible bond. The findings that the in-the-money calls seem to be larger and occur earlier relative to the maturity of the bonds may be a surprise because convertible bonds are out-of-the-money when they are issued. However, there are at least three possible explanations for why this is the case. First, it may be the case that the bonds in out-of-the-money calls have earlier been in-the-money and that some bondholders converted when this was the case. Second, the firms making out-of-the-money calls may earlier have called fractions of the issues. Third, it may simply be that firms in general only are able to redeem for cash the smaller issues. 5 Announcement effect This section will examine the stock market reaction to the announcement of the calls. When examining the stock market reaction, we have to make sure that a potential announcement ef- 13

15 fect is not caused by other news announced about the firm at the same time. Therefore, when examining the announcement effect we will require that there is no other news about the calling firm in the Wall Street Journal in a period from one day before the call announcement until two days after. 17 Similarly, we will also remove firms that call for example warrants, convertible preferreds, or straight debt at the same time. 18 This reduces the number of out-of-the-money calls from 112 to 85 and the number of in-the-money calls from 427 to 361. Put differently, for the out-of-the-money calls we have that 24% of the calls are associated with other news while only 15% of the in-the-money calls are associated with other news. This indicates that out-ofthe-money calls more often are associated with other news or other financial restructuring than in-the-money calls. 5.1 The announcement of out-of-the-money and in-the-money calls First, we will examine if there is a significant announcement effect for the two types of calls and given this, we will examine whether the two types of calls differ with respect to the announcement effects. Table III shows the results from calculating the announcement effect as defined in equation (1) while figure 1 shows the distribution of the announcement effects for the two types of calls. From table III it follows that there is a positive and significant announcement effect of 1.74% for out-of-the-money calls while the announcement effect for in-the-money calls is a negative and significant 1.47%. The two announcement effects are significant at the 1% level both according to the t-test and according to the sign test. For the out-of-the-money calls 74% of the announcement effects are positive while for the in-the-money calls only 26% of the calls are positive. Furthermore, table III also shows that the difference in the announcement effect between the two types of calls is significant at the 1% level. From figure 1 it follows that the significance of the announcement effect for the two types of calls and the significance of the difference in the announcement effects are not due to outliers. Based on the results above, we conclude that the stock market reaction to convertible bond 17 The reason for this is that the announcement period used is the day the call is announced at the news wires and the following day. 18 For firms that call several convertible bonds at the same time, these bonds are aggregated in order for example to obtain the total number of new shares to be issued upon a full conversion of the bonds. 14

16 Mean Median Pos/Neg OTM 1.74%*** 1.28% 63/22*** ITM 1.47%*** 1.39% 95/266*** Difference 3.21%*** Table III: The mean announcement effect and the number of positive/negative announcement effects for the out-of-the-money and in-the-money convertible bond calls that have no other news announced in the Wall Street Journal in the period from one day before the call to two days after. The test for significance of the mean of the announcement effect is based on a standard t-test. The test for significance of the number of positive relative to the number of negative announcement effects is based on the sign test as described in Campbell, Lo, and MacKinlay (1997, p. 172). The test for difference in the announcement effect between out-of-the-money and in-the-money calls is a standard t-test for the difference in means between two samples. *** denotes significance on the 1% level. calls is fundamentally different from out-of-the-money calls to in-the-money calls: For out-of-themoney calls it is positive, for in-the-money calls it is negative. Therefore, these results confirm the results in Cowan, Nayar, and Singh (1992) regarding out-of-the-money calls and all the literature on in-the-money calls. In order to provide more insight about the cause of the announcement effect for the two types of calls, the following will examine if there is a relation between some of the characteristics of the convertible bonds and the associated announcement effect. In addition, as seen from table II, there are some differences in the characteristics of the two subsamples of calls. Therefore, the following will also examine if some of these differences can explain the difference in the announcement effect. We will first examine if the listing of the firms stock can explain the stock market reaction. Subsection 5.2 will then examine if the characteristics of the bonds are important in explaining the announcement effect Stock exchange difference There are several reasons why the stock exchange where the stock is listed may be relevant for the size of the announcement effect. First, as it follows from Mazzeo and Moore (1992) and Bechmann (1999) the announcement effect may be related to the depth of the market which may differ between the three stock exchanges. The same holds for other market microstructure 15

17 18 70 Number of out-of-the-money calls Out-of-the-money In-the-money Number of in-the-money calls % -14% -13% -12% -11% -10% -9% -8% -7% -6% -5% -4% -3% -2% -1% 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12% 13% 10 0 The announcement effect Figure 1: The distributions of the announcement effect for the two types of calls. In order to be able to compare the two distributions, the histogram of the announcement effects for the out-of-the-money calls are plotted using the left axis, while the histogram of the announcement effects for the in-the-money calls are plotted using the right axis. aspects such as the bid-ask spread or restrictions on short selling. 19 Furthermore, it is also possible that the reaction to information revealed by the calls differs between the stock exchanges. Finally, it is also relevant to examine the effect of the stock exchanges because most of the earlier literature restricts the data-set to firms listed on NYSE and ASE. In order to examine the importance of the stock exchange, we run the regression: CAR i,ann = γ 0 + γ 1 ASE i + γ 2 NASDAQ i + ɛ i, (4) where ASE i is a dummy variable equal to one if observation i is listed on the American Stock Exchange at the time of the call, while NASDAQ i similarly is equal to one if the observation is 19 For example, Christie and Schultz (1994) and Huang and Stoll (1996) documented that the spreads are wider on NASDAQ than on the NYSE. 16

18 listed on NASDAQ. Table IV shows the results from this regression for both the out-of-the-money and the in-the-money calls. Intercept ASE NASDAQ ˆγ 0 ˆγ 1 ˆγ 2 OTM (3.88, < 0.1%) (0.20, 84%) (1.90, 6.1%) ITM ( 8.12, < 0.1%) ( 0.07, 94%) (0.30, 76%) Table IV: Results from the regression given in equation (4). ASE is a dummy variable which is one if the stock is listed on the American Stock Exchange. Similarly, NASDAQ is one if the stock is listed on NASDAQ. The brackets give the t-value and the corresponding p-value in a standard t-test for significance of the variable. From table IV it follows that there is only very weak evidence for that the stock exchange has an effect on the announcement effect. Only for the out-of-the-money calls is there some evidence for the stock exchange being relevant for the announcement effect. For the out-of-the-money calls, the NASDAQ dummy enters positively with a p-value of 6.1%. However, a further look at the data reveals that a call with an announcement effect of 13.9% has the stock listed on the NASDAQ. If the same regression is run without this observation, the NASDAQ dummy becomes insignificant with a p-value of 35%. The regression in equation (4) has also been run using the whole data-set and in different extended versions. However, just as table IV these do not provide any evidence that the stock exchange is relevant for the sign or the size of the announcement effect. 5.2 Characteristics of the called convertible bond According to several explanations for the announcement effects, one should expect that some of the characteristics of the called convertible bonds are relevant for the size of the announcement effect. Therefore, the following will examine the relation between several characteristics and the announcement effect. 17

19 Bond rating effect The first relation that will be examined is the relation between the rating of the convertible bond and the announcement effect. This relation has not been examined before for calls of convertible bonds but there are several reasons why one could expect to find a relation. For example, if the announcement effect is due to new information revealed by the call, it is natural to reason that the announcement effect should depend on the rating of the convertible bonds. 20 In order to examine the relation, we run a regression of the announcement effect on different dummy variables measuring the rating of the convertible bonds, i.e. we run the regression CAR i,ann = γ 0 + γ 1 NR i + γ 2 InvGr i + ɛ i. (5) Here NR i is one if there is no rating available for bond i, while InvGr i is one if bond i is considered investment grade, i.e. the rating is BBB or above according to the latest S&P rating of the bond before the call. The different explanations for the announcement effect can predict different signs on the variables in regression (5). For example, for the out-of-the-money calls the good-news hypothesis can be argued to predict a negative sign on both the No Rating variable and the Investment Grade variable. This is because the possible good-news revealed by the call must be most valuable for the firms with the lowest ranking. Similarly, the bad-news hypothesis for the in-the-money calls predict a positive coefficient on both the No Rating and the Investment Grade variable because the bad news revealed by the call will be most severe for the firms with the lowest ranking. The results from the regression in equation (5) for the two types of calls are given in table V. From table V we obtain only weak evidence for a relation between the bond rating and the announcement effect. The only significant relation is for No Rating variable for the in-the-money calls. There the relation between the announcement effect and No Rating is positive as predicted for example by the bad-news hypothesis. However, as this is the only significant relation it is difficult to draw any conclusions based on the result in table V. 20 Other studies have examined if there is a relation between an announcement effect and bond ratings. For example, Shleifer (1986) examined the relation between bond ratings and the announcement effect associated with the inclusion of new stocks into the S&P 500 list. Related to convertible bonds, Mikkelson and Partch (1986) found that the announcement effect related to the issuance of convertible bonds depends on bond ratings. 18

20 Intercept No Rating Investment Grade NR InvGr ˆγ 0 ˆγ 1 ˆγ 2 OTM (2.55, 1.2%) (0.36, 71%) ( 0.29, 84%) ITM ( 7.77, < 0.1%) (2.01, 4.5%) (0.60, 55%) Table V: Results from the regression given in equation (5). NR is a dummy which is one if there is no rating available for the convertible bond, while InvGr is one if the convertible bond is considered investment grade, i.e. the rating is BBB or above. The brackets gives the t-value and the corresponding p-value in a standard t-test for significance of the variable. Effect of other characteristics of the call The following will examine the relation between the announcement effect and other characteristics of the call. According to several of the explanations for the announcement effect, one should expect that the size of the called convertible issue is relevant for the announcement effect. According to the good-news hypothesis for out-of-the-money calls, one should expect a positive relation between the announcement effect and the size of the call, while the bad-news hypothesis for in-the-money calls predicts a negative relation between the announcement effect and the size of the call. The same holds for the avoid-dilution hypothesis for out-of-the-money calls and the price pressure hypothesis for both types of calls. With respect to the size of the call, we have a least two possible measures. One measure is the number of shares that the called issue is convertible into divided by the number of shares outstanding. We will denote this measure SizeM easure1. Another measure is the total call payment offered to the bondholders divided by the total value of equity outstanding. This measure will be denoted SizeM easure2. Table III has already documented that the announcement effect depend on whether the conversion right is out-of-the-money or in-the-money. It is therefore natural to examine if the extent to which the conversion right is out-of-the-money or in-the-money is relevant for the announcement effect. The following will do this by including a variable that measures the moneyness of the conversion right in a regression trying to explain the announcement effect. More precisely, 19

21 the moneyness is defined as the conversion value divided by the call payment. Again, the different explanations for the announcement effect can predict different relations between the size of the announcement effect and the moneyness. First, for the out-of-the-money calls there is one main reason why a positive relation is expected between the moneyness and the announcement effect, and this reason is related to the optimal call policy as described in section 3.1. If the firm calls too early meaning that the moneyness is much lower than one, the firm will transfer money from the stockholders to the holders of the convertible bonds. Therefore, the closer the moneyness is to one, the higher the announcement effect is expected to be. For the in-the-money calls, both the bad-news hypothesis and the price pressure hypothesis (or maybe more precisely the hedging explanation in Bechmann (1999)) predict a positive relation between the moneyness and the announcement effect. Convertible bond calls can happen as part of a merger, a takeover, or an acquisition. Therefore, several studies of convertible bond calls suggest that these calls should be removed from the sample. In this paper, these calls have only been removed if there is other news announced about the firm at the same time as the call. Instead, this paper will examine if the announcement effect in general is different for these calls. This is done by including a dummy variable, T O, in the regression of the announcement. The dummy variable is one if the call is part of a merger, a takeover, or an acquisition. Finally, as argued in Singh, Cowan, and Nayar (1991), the announcement effect for in-themoney calls depend on whether the call is underwritten or not. Therefore, the following will also include a dummy variable, which is one if the call is underwritten. In order to examine how the announcement effect is related to the variables described above, we run the following regression: CAR i,ann = γ 0 + γ 1 SizeMeasure1 i + γ 2 SizeMeasure2 i + γ 3 Moneyness i + γ 4 T O i + γ 5 Underwritten i + ɛ i (6) The results from running this regression for the two types of calls are given in table VI. As it follows from the table the two types of calls are quite different. Therefore, we have chosen only to present the results from the regression using the two sub-samples. Table VI provides us with several interesting results about the announcement effect for convertible bond calls. We start by observing that the results for the in-the-money calls are consistent with the findings in the existing literature on these calls. Next, we take the findings one by one. 20

22 Intercept Size Size Moneyness Takeover Under- R 2 Measure 1 Measure 2 written F,p ˆγ 0 ˆγ 1 ˆγ 2 ˆγ 3 ˆγ 4 ˆγ 5 OTM % (2.6, 1%) (2.2, 3%) 4.7,3% % (2.8, 1%) (1.9, 7%) 3.5,7% % ( 0.5, 61%) (2.1, 4%) 4.3,4% % (5.4, < 1%) ( 1.0, 31%) 1.0,31% % ( 0.4, 67%) (1.6, 11%) (1.5, 13%) 3.5,3% % ( 0.1, 91%) (0.5, 65%) ( 0.2, 87%) (1.0, 31%) ( 0.6, 55%) 1.8,13% ITM % ( 4.9, 0%) ( 2.3, 2%) 5.2,2% % ( 5.0, 0%) ( 2.6, 1%) 6.6,1% % ( 6.0, 0%) (2.6, 1%) 6.5,1% % ( 10.4, 0%) (0.3, 77%) 0.1,77% % ( 7.6, 0%) ( 2.6, 1%) 6.7,1% % ( 4.3, 0%) ( 1.9, 6%) (2.2, 3%) 5.2,0.6% % ( 3.6, 0%) ( 0.1, 90%) ( 0.3, 75%) (1.7, 9%) ( 0.4, 70%) ( 1.6, 11%) 2.6,2% Table VI: Results from the regression given in equation (6) for the out-of-the-money and the in-the-money calls. Size Measure 1 is the number of shares that the called bond issue is convertible into divided by the number of shares outstanding before the call. Size Measure 2 is the total call payment offered to the bondholders divided by the total value of equity outstanding before the call. Moneyness is the conversion value divided by the call payment. Takeover is a dummy variable which is one if the call happens in connection with a takeover, a merger, or an acquisition. Underwritten is a dummy variable which is one if the call is underwritten. The brackets give the t-value and the corresponding p-value in a standard t-test for significance of the individual variables. A p-value of 0% denotes that the p-value is less than 0.1%. R 2 is the coefficient of determination for the individual regressions. F,p denotes the F-test for significance of the regression and the corresponding p-value. 21

23 First, from the individual regressions it follows for out-of-the-money calls that there is a positive relation between the two size measures and the announcement effect while the relation is negative for the in-the-money calls. In addition, these relations are significant at the 1 to 3% level except for Size Measure 2 for out-of-the-money calls where the level of significance is 7%. As expected from several of the explanations for the announcement effect, this shows that the larger the size of the call the larger is the stock market reaction. Second, the relation between the moneyness and the announcement effect is significant and positive for both types of calls. This means that a higher moneyness for out-of-the-money calls will lead to a more positive announcement effect while a higher moneyness for in-the-money calls leads to a less negative announcement effect. Third, it follows from the table that there is no evidence that calls related to takeovers etc. are associated with a different stock market reaction. Fourth, the announcement effect for underwritten calls is significantly less than for non-underwritten calls, but the announcement effect for non-underwritten calls is still significantly negative. Finally, in the joint regression on Size Measure 1 and Moneyness or on all variables we observe that the variables becomes less significant. Especially in the regression including all variables, all the variables, that were highly significant in the individual regressions, are now insignificant. This is clearly a problem of multicollinearity because by the fact that several of the variables are correlated. One example is of course the two size measures, but other variables are also known to be correlated. The moneyness and the underwriting dummy are correlated as shown in Singh, Cowan, and Nayar (1991). Bechmann (1999) also discussed some correlations among the other variables. 6 The long-term stock performance Having examined the stock returns in a two day interval around the announcement of the call, it is interesting to examine if there also is a difference between the two types of calls with respect to the long-term stock returns before and after the call. The following will do this using the different methods described in section 4.1. The first results on the long-term stock returns are based on the market model regression given in equation (2). Table VII lists the results from running this regression for the two types 22