Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. *

Transcription

1 Asia-Pacific Journal of Financial Studies (2009) v38 n3 pp Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. * Mookwon Jung Kookmin University, Seoul, Korea Received 12 September 2007; Accepted 8 October 2008 Abstract This paper shows that long-term equity and operating performances that follow straight and convertible debt issuance in the U.S. are different, just as are announcement-period market reactions, supporting the notion that market investors initially underreact to a selective event of convertible debt issuance. Cross-sectional tests reveal that firms issuing convertible debt often suffer stock underperformance partly because of their deteriorating fundamentals subsequently to the issuance. Even when each convertible bond directly matches with a straight bond that has a similar offering date and issuer and bond characteristics, convertible debt issuers still significantly under-perform straight debt issuers, supporting that what matters for long-term equity and operating performances is the type of debt offering. Keywords: Convertible Debt; Straight Debt; Long-term Equity Performance; Long-term Operating Performance; Selective Event JEL Classification: G12, G14, G32 * This paper is part of my doctoral dissertation at the University of Houston. I am especially grateful to Sorin Sorescu for his encouragement and guidence. Also, I would like to thank D. Katherine Spiess at the FMA Conference and two anonymous referees for helpful comments ** Corresponding Author. Address: Kookmin University, College of Business Administration, Chongnung-dong, Songbuk-gu, Seoul, Korea, ; mjung@kookmin.ac.kr; Tel: ; Fax:

2 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. 1. Introduction One reason for issuing convertible debts is to obtain indirect financing when an equity issuance appears unattractive because of the adverse-selection problem under informational asymmetry. Stein (1992) argues that convertible debt can serve as an indirect mechanism for getting equity financing through the backdoor that mitigates the adverse-selection costs associated with direct equity issues. The backdoor equity explanation for the use of convertible debt emphasizes that convertible bonds have a call feature for early forced conversion. Myers and Majluf (1984) provide a rationale for this adverse-selection problem, arguing that managers, who know more than outside investors, use their superior information to benefit existing shareholders at the expense of new shareholders. Therefore, the announcement of an equity offering is regarded as less favorable than the announcement of a debt offering. Since convertible debt is a relative compromise between equity and straight debt, the announcement-window stock price reaction to convertible debt offerings is less negative compared to equity offerings and more negative compared to straight debt offerings. Recently, many studies regarding market responses to equity and debt offerings have shifted their focus from the announcement-window to long-term horizons. For instance, Spiess and Affleck-Graves (1999) report long-term negative stock performance following straight and convertible debt issuance; however, they do not relate long-term performance to the short-term reaction or to changes in the issuer s fundamentals. One possible explanation for the observed long-term stock underperformance is that the methodologies used to calculate long-term performance are misspecified. According to the efficient market hypothesis, all publicly available information about the prospects of a firm should be reflected immediately in the stock price; therefore, the abnormal performance for the long-term horizon beyond the narrow event window should be zero. Fama (1998) points out that the anomalies in the long-term drift studies could result from merely an incorrect asset pricing model or the use of biased methodologies. Thus Fama prefers to use a time-series model that specifies a method to measure abnormal stock performance. On the other hand, Loughran and Ritter (2000) cast doubt on the power of time-series regressions, arguing that the weighting scheme will differ depending on the hypothesis since time clustering is a result of 338

3 Asia-Pacific Journal of Financial Studies (2009) v38 n3 managerial action and behavioral timing in response to market investors misevaluations. The other possibility is that long-term underperformance could be explained in terms of a behavioral framework. Daniel et al. (1998) propose a model in which overconfident investors underreact to selective public events initiated by informed managers when the stock is overvalued. 1) Considering that a convertible debt offering is a selective event that is usually preceded by high stock performance, their model implies that subsequent long-term stock underperformance is a result of investors initial underreaction to the event intended to take into account the pre-event market overvaluation. However, investors do not show significant reaction to straight debt issuance because it is not associated with a large and significant pre-event market mispricing. Since this behavioral framework posits that long-term performance is of the same sign as the short-term reaction, the long-term negative stock performance following straight debt issuance (except perhaps among low-quality issuers) in Spiess and Affleck-Graves (1999) is somewhat puzzling because the short-term reaction to straight debt issuance is known to be insignificant. Using a larger sample of debt, I begin my study by revisiting the empirical evidence of abnormal stock performance before and after the issuance of convertible and straight debts to determine whether the stock price response conforms to efficient markets or to the behavioral model of DHS. Consistent with previous results, announcement-period stock prices are not significantly for straight debt issuers and are significantly negative for convertible debt issuers. However, I find that stock prices do not significantly underperform on average following the issuance of straight debt, while convertible debt issuers severely underperform their size-, book to market ratio-, and momentum-matched nonissuers. These results appear to be consistent with the empirical implication of Daniel et al. (1998) because both short-term reaction and long-term performance are negative for convertible debt issuance and are insignificant at conventional levels for straight debt issuance. I next examine operating performance subsequent to convertible and straight debt issuance by comparing them to long-term stock performance. Unlike straight debt 1) They define overconfident investors as ones who overestimate the precision of their private information signals. Selective events are the events that are correlated with pre-event stock market mispricing and that will forecast future stock price changes. Exchange offers, repurchases, and new issues are selective mostly owing to managers incentives to take into account pre-event market mispricing, while dividends and stock splits are selective primarily due to a signaling motive. 339

4 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. issuers, convertible debt issuers severely underperform their matching nonissuer in operating performance, and the abnormal operating performance is positively and contemporaneously correlated with abnormal stock performance. This implies that stock underperformance following convertible debt issuance can be partly explained by subsequently deteriorating operating performance, suggesting that market investors underreact to convertible debt issuance that connotes not only managers motives to take advantage of a transitory market overvaluation but also a negative signal of less favorable future operating performance. Finally and more importantly, to provide a better comparison of performance after controlling for potential differences in initial market conditions and issuer and bond characteristics, I introduce a direct comparison method in which each sample convertible debt issuer is directly matched against a control straight debt issuer that has a similar offering date, adjusted for momentum, size, book-to-market ratio, pre-event beta, and initial bond rating. Still, a clear difference between long-term equity and operating performances is observed and this implies that low returns following substantial runups, as in the case of a seasoned equity offering discussed in Loughran and Ritter (1995), owe to the type of debt issuance: convertible or straight. The rest of this paper is organized as follows: Section 1 reviews the literature regarding stock price reactions to both convertible and straight debt offerings. Section 2 describes the sample selection and descriptive statistics in which debt issuers are grouped by their initial bond ratings. Section 3 reexamines abnormal stock performance around convertible and straight debt issuance, using size, book-to-market ratio, and momentum as benchmarks. Section 4 reports long-term operating performance subsequent to convertible and straight debt offerings. Section 5 introduces a direct comparison of long-term stock performance, using various benchmarks including offering date and initial bond rating. Section 6 shows a direct comparison of long-term operating performance. Section 7 concludes with a summary. 2. Review of the Literature 2.1 Adverse-Selection and Announcement-Period Market Reactions Negative stock price reactions around announcement dates of convertible debt issuance can be explained in the context of models using adverse selection. Because 340

5 Asia-Pacific Journal of Financial Studies (2009) v38 n3 investors feel that inside managers use their superior information against new shareholders, the offering of equity or equity-related securities (rather than straight bonds) sends a negative signal to market investors (Myers and Majluf, 1984). 2) As an alternative view, the matching of financial and real investment options is suggested by Myers (1999). He shows that capital expenditures and new long-term financing start to increase in the year of the forced conversion of the convertible debt. Mikkelson and Partch (1986) document that announcements of equity and convertible debt offerings convey that the firm is overvalued. Many studies that examine announcement-period reactions surrounding offerings of equity and convertible debt provide support for these hypotheses. Mikkelson and Partch (1986) and Eckbo (1986) report significantly negative event-window returns to convertible debt announcements of -1.39% and %, respectively. They also report negative but insignificant abnormal returns to the announcement of straight debt offerings. In addition, Asquith and Mullins (1986) report a two-day announcement period return of -3.22% to equity offering announcements. I also report a 3-day announcement-period return of -1.46% for convertible debt issuers, which is statistically significant, and -0.20% for straight debt issuers, which is not significant. These results suggest that market reactions to convertible and straight debt issuance are different. However, not much empirical attention has been paid to the relation between the initial reactions and subsequent long-term stock performance. 2.2 Long-term Abnormal Stock and Operating Performances Using buy-and-hold and cumulative abnormal returns as measures of long-term stock performance, previous studies have examined long-term stock price reactions following convertible and straight debt issuance. Jung, Kim, and Stulz (1996) report a 2% abnormal performance over the five-year period following a straight debt offering. Lee and Loughran (1998), and Lewis, Rogalski, and Seward (1999) report, respec- 2) The adverse selection explanation does not hold for other countries. Kang and Stutz (1996) report a 3-day abnormal return of 2.79% to the announcement of convertible debt offerings in Japan. The possible explanation for the positive abnormal returns is that the convertible debt issue in Japan is not preceded by abnormally high stock performance and that a firm s convertible debt issue conveys to the market that the firm becomes independent from banks. In Korea, the stock market response to convertible debt offerings is mixed depending upon the sample period. 341

6 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. tively, abnormal returns of -30.4% and % over the five-year period that follows a convertible debt offering. Spiess and Affleck-Graves (1999) show that mean (median) differences in five-year holding period returns between sample firms and control firms are % (-19.8%) for convertible debt issuers and -14.3% (-18.7%) for straight debt issuers. Using a larger sample of 604 independent convertible debt and 1,301 straight debt issues that are controlled for equity, convertible debt offerings and other repeated issues, and adjusting for momentum, size, and book-to-market ratio, I report a % underperformance for convertible debt issuers and 1.56% for straight debt issuers over the five-year horizon. The use of control firms of similar momentum as a benchmark is motivated by the idea that when sample firms experience abnormal preevent performance, matching procedures that ignore the pre-event performance may yield misspecified test statistics (Lyon, Barber, and Tsai, 1999). The small and insignificant number in the case of straight debt issuers contrasts with the result from Spiess and Affleck-Graves (1999) but is close to Jung, Kim, and Stulz (1996). Eckbo, Masulis and Norli (2000) also argue that the high magnitude of negative abnormal performance following straight debt issuance is unreasonable because straight debt issuance has a less adverse-selection problem and is less likely to reflect opportunistic timing by managers. Unlike prior studies that mostly focus on the phenomenon of long-term stock performance, I further examine long-term operating performance and show that severely deteriorating operating performance following convertible debt issuance could account for some part of subsequent long-term stock underperformance. In addition, holding constant offering dates and issuer and bond characteristics, I compare convertible debt issuers directly with straight debt issuers and show that their performances are clearly different and that the type of debt issued is important in determining post-event equity and operating performances. 2.3 Market Efficiency and Model Misspecification When viewed within the context of the efficient market hypothesis, many documented cases of long-term abnormal stock performance are difficult to understand. Addressing the apparent long-term abnormal performance documented in many studies, Fama (1998) raises concerns about model misspecification in calculating long- 342

7 Asia-Pacific Journal of Financial Studies (2009) v38 n3 term abnormal returns. He argues that all models of expected returns are incomplete descriptions of the systematic variation of expected returns across firms. As the horizon in an event study becomes larger, it is more likely that misspecified asset pricing would produce abnormal returns, due to the effect of compounding. Eckbo, Masulis and Norli (2000), using a model based on macroeconomic factors, report that issuer undeperformance following offerings of convertible debt, straight debt, and equity reflects lower expected returns due to lower post-event exposure to unexpected inflation and default risk. Brav, Grezy and Gompers (2000) also find that equity issuer returns are not a distinct anomaly and covary with non-issuing firm returns. Mitchell and Stafford (2000) address the positive cross-correlation of abnormal returns due to the time and industry clustering of major corporate events, including seasoned equity offerings. After adjusting for positive cross-correlations of individual returns, they find much smaller test statistics for the long-term abnormal returns. However, tests that weigh firms equally may not be always better than tests that weigh each time period equally. Since managers voluntary actions to take advantage of transitory market misevaluations tend to occur simultaneously in certain time periods, the subsequent abnormal returns may not be detected if all periods are equally weighted (Loughran and Ritter, 2000). In view of the potential drawbacks associated with each long-term performance measure, I report the results using both cumulative and buy-and-hold abnormal returns, as well as a calendar time portfolio methodology. 2.4 Behavioral Model Alternatively, long-term abnormal performance may imply that market investors behave irrationally in their initial responses or that they do not properly recognize managers information. It is common that behavioral models associate long-term negative drifts with high pre-event stock performance and investors tendency to be unduly optimistic regarding recent stock performance. Daniel et al. (1998) propose that if investors are overconfident, selective events that are initiated when the stock is overvalued (undervalued) by the market will on average be associated with negative (positive) announcement-period abnormal price changes and will on average be followed by negative (positive) post-announcement abnormal price changes. Their model predicts that the average post-event returns are of the same sign as the average event-date returns and that securities issuance related to adverse selection will 343

8 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. tend to occur when prior runup as a measure of mispricing is high. The role of momentum traders relying on recent stock performance and investors tendencies to put too much weight on recent good and bad performance are also discussed in order to explain the phenomenon of long-term drifts (Hong and Stein, 1999 and Brav and Heaton, 2000). However, Loughran and Ritter (1995) show that low returns following substantial runups are related more to whether or not a firm has issued equity than to a simple manifestation of long-term mean reversion, which is consistent with the model of Daniel et al. (1998) who emphasize the role of selective events in forecasting future stock price changes. I report 24.15% and 6.32% of one-year pre-event abnormal stock returns for convertible and straight debt issuers, respectively. Such behaviors as the abnormally high pre-event stock performance, the negative announcement period reaction, and the significant long-term underperformance in firms issuing convertible debt suggest that overconfident market investors might not rationally react to the selective event of convertible debt issuance in the case where pre-event stock performance is abnormally high. In this paper, a convertible debt offering is regarded as selective not only owing to managers motives to take advantage of a transitory market overvaluation but also owing to a signaling role that conveys less confidence in future operating performance. On the other hand, neither the announcement-period nor the long-term abnormal returns for straight debt issuers are significant, and their post-event abnormal operating performance is not as negative as that for convertible debt issuers. This implies that straight debt issuance is neither initiated to take advantage of misvaluation nor related to a negative signal. Furthermore, comparing two types of debt issuers directly, I show that a straight debt offering is not a selective event even though the issue is preceded by prior high stock performance. 3. Sample Selection and Descriptive Statistics Convertible and straight bond data are obtained from the Fixed Income Securities Data (Lehman Brothers Bond Database), provided by LJS Global Information Services Incorporated. 3) The bonds issued by foreign companies have been excluded from 3) These data were obtained from the Fixed Income Research Program at the University of Houston, Bauer College of Business. LJS Global Information services was formerly known as FITCH IBCA Information, Inc. 344

9 Asia-Pacific Journal of Financial Studies (2009) v38 n3 the analysis since activities in a foreign stock may contaminate the effects observed in U.S. markets. Asset-backed securities and warrants are also excluded. Firms that are not on the CRSP NYSE/AMEX or NASDAQ database as of the offer date are also left out. Out of 794 samples, I include convertible debt offers if the issuing firm has not conducted a seasoned equity offering and a convertible debt offering for the five years before and after the offering date. Out of 5,930 samples, I include straight debt offers if the issuing firm has not conducted a seasoned equity offering, a convertible debt offering and a straight debt offering for the five years before and after the offering date. A total of 604 independent convertible bond and 1,301 straight bond issues over the period are included in the final samples. 4) To measure firms operating performance, return on assets and return on equity in the calendar years (up to 1998) surrounding the offering year are obtained from COMPUSTAT. Table 1 presents descriptive statistics for convertible bond and straight bond issues. In Panel A, observations are grouped according to their initial bond ratings that are measured as the average rating assigned by the three leading rating agencies, Moody s, S&P, and FITCH. It is interesting to note that the distribution of the credit rating categories is quite different across the two samples. The majority of the convertible bonds are rated in the range of BBB+ to B-, while the majority of the straight bonds are rated between AAA and BB-. This distribution is consistent with the empirical implications of Stein (1990) who argues that when the costs of financial distress are high, medium-quality firms will tend to issue convertible debt. The table also presents the yield to maturity (YTM) at the time of issue. The average YTM of the convertible bonds is 7.27% while that of the straight bonds is 9.99%. The average maturity of the convertible bonds is years that is a little longer than years of the straight bonds. Yet, the YTM of convertible bonds looks much lower despite their ratings, even considering a little longer maturity of the convertible bonds. This implies that the conversion option provides clear value to market investors. In Panel B, the observations grouped by offering years show that the majority of convertible bond issues took place in the 1980s and 1990s. 4) The Lehman Brothers Bond Database that is also available from WRDS (Wharton Research Data Services) demised in For the screening purpose, 1,929 samples of seasoned equity offerings from 1986 to 1995 are used. Using a larger total of 5,930 straight bond and 794 convertible bond issues does not affect the significance level of the results in this paper. 345

10 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. Table 1. Descriptive Statistics Convertible and straight debt issues are grouped according to their initial bond ratings (Panel A) and offering years (Panel B). Maturity (year) is the time to maturity of the bond. YTM is the yield to maturity of the bond. Size ($ mil) indicates the market value of the sample firm as of the end of the offering month. Panel A: Maturity and YTM Convertible debt Straight debt Rating category No obs Maturity YTM (%) No obs Maturity YTM (%) AAA to A BBB+ to BB B+ to B CCC+ to D Non-rated Total , Panel B: Offering years Convertible debt Straight debt Offering year No obs Size ($ mil.) No obs Size ($ mil.) , , , ,358 Total , Abnormal Stock Performance around Debt Issuance Table 2 reports abnormal stock performances following convertible debt (Panel A) and straight debt (Panel B) issuance for pre-event, announcement, and post-event periods, grouped by initial bond ratings. Indicated is the number of observations whose one-year post-event abnormal returns are available from the CRSP NYSE/ AMEX or NASDAQ database, which is also equivalent to the number of sample debt issuers in Table

11 Asia-Pacific Journal of Financial Studies (2009) v38 n3 Table 2. Abnormal Stock Performance around Debt Issuance Rating category No obs Pre-event Announcement Post-event 1-year 3-day 1-year 3-year 5-year Panel A: Convertible debt Cumulative abnormal returns (%) AAA to A (4.63) *** (-2.22) ** (-0.15) (1.39) (1.10) BBB+ to BB (11.61) *** (-2.64) *** (1.04) (-0.65) (0.17) B+ to B (10.27) *** (-5.56) *** (-3.25) *** (-4.01) *** (-2.92) *** CCC+ to D (3.60) *** (-1.65) * (-2.42) ** (-2.45) *** (-3.98) *** Non-rated (5.64) *** (-4.20) *** (-0.69) (-1.38) (-1.43) Total (15.55) *** (-7.68) *** (-2.33) ** (-3.56) *** (-3.17) *** Mean monthly abnormal returns (%) FF intercept (-2.94) *** (-5.22) *** (-4.62) *** Adj FF intercept (-1.29) (-1.96) ** (-2.09) ** No month Panel B: Straight debt Cumulative abnormal returns (%) AAA to A (4.86) *** (-0.23) (2.58) *** (2.40) ** (3.75) *** BBB+ to BB (4.18) *** (-1.50) (3.01) *** (1.23) (1.18) B+ to B (2.51) ** (-1.34) (0.55) (0.06) (-0.66) CCC+ to D (2.61) *** (-1.06) (-1.36) (-1.94) * (-2.82) *** Non-rated (1.51) (0.48) (-0.22) (-0.81) (-0.33) Total 1, (7.02) *** (-1.08) (2.10) ** (0.46) (0.45) Mean monthly abnormal returns (%) FF intercept (-0.96) (3.04) *** (-2.23) ** Adj FF intercept (1.22) (1.22) (-2.24) * No month

12 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. 4.1 Pre-event Abnormal Performance Some corporate events, including equity and convertible debt offerings, are preceded by abnormally positive stock performance. This justifies the need to adequately control for the momentum effect when measuring long-term abnormal returns. The first column of each panel in Table 2 shows mean cumulative abnormal returns (CARs) for the one-year horizon prior to the announcement month, measured by subtracting the sum of the monthly returns on an equally weighted market index from the sum of the monthly returns of the corresponding debt issuer. The CARs are 24.15% for convertible debt issuers and 6.32% for straight debt issuers, both of which are significant, although the magnitude for straight debt issuers is not large. This implies that convertible debt issuance is preceded by abnormally high stock performance. Jung, Kim, and Stulz (1996) report -1.63% (-3.26%) and 13.95% (15.07%) of mean (median) cumulative abnormal returns for an 11-month period prior to straight debt and equity issuance, respectively. Spiess and Affleck-Graves (1999) report mean buy-and-hold abnormal returns of 187% and 74% for a five-year horizon prior to convertible and straight debt offerings, respectively. High pre-event stock performance raises the question of whether negative postevent performance may actually result from a bias in the statistical methodology employed. Lyon, Barber, and Tsai (1999) show that when compared to firms with low pre-event returns, firms with high six-month pre-event returns tend to underperform their benchmarks for three- and five-year horizons. Therefore, they recommend matching sample firms to firms with similar pre-event abnormal performance to avoid misspecification, which highlights the importance of controlling for pre-event momentum in this current study. 4.2 Announcement-Period Market Reaction The second column of each panel in Table 2 shows short-term reactions to the announcement of convertible and straight debt issuance. Cumulative abnormal returns for the announcement period are calculated over the 3-day (t-1, t+1) window surrounding the announcement date t, using the market model of Brown and Warner (1985). The benchmark estimation period is (t-100, t-6). The 3-day announcementwindow return for convertible debt issuers is -1.46%, which is statistically significant, while the abnormal return for straight debt issuers is -0.20% and insignificant. These 348

13 Asia-Pacific Journal of Financial Studies (2009) v38 n3 results are consistent with the previous findings of Mikkelson and Partch (1986) and Eckbo (1986). One possible explanation for the negative stock price reactions around announcement dates of convertible debt issuance can be provided by the adverseselection framework in which market investors feel that the firm is overvalued. 4.3 Post-event Abnormal Performance Buy-and-hold abnormal returns (BHARs), as a long-term performance measure, provide a better representation of the overall investment experience but may overstate abnormal returns by compounding the returns occurring at the beginning of the postevent period over longer horizons (Fama, 1998) and do not adequately control for crosssectional correlation among individual firms, leading to higher test statistics (Mitchell and Stafford, 2000). CARs are also vulnerable to the problem of cross-sectional dependence among event firms but are less likely to yield overstated returns. In this section, I report post-event abnormal performance using CARs and BHARs as well as the Fama- French (1992) three-factor model and a calendar time methodology Long-term CARs The last three columns of each panel of Table 2 show post-event stock performance. The post-event CARs are calculated over one-, three-, and five-year horizons following convertible and straight debt offerings by subtracting the sum of monthly returns of the matching nonissuer from the sum of monthly returns of the corresponding debt issuer. First, I identify the candidate firms whose market value of equity is between 70% and 130% of the market value of the sample firm as of the end of the offering month. Then, I choose the candidate firms whose book-to-market ratio is between 70% and 130% of the book-to-market ratio of the sample firm. From these candidates, I select the firm whose one-year pre-event stock performance is closest to that of the sample firm. There is only one case in which a firm is used twice in the same calendar year when matched to convertible debt issuers, while there are two cases for straight debt issuers. Matching nonissuers are also screened to be independent of equity, convertible debt offerings and other repeated issues for the five-year horizon before and after issuance. In contrast to the pre-event results, the mean post-event CARs for the one-, three-, and five-year horizons following convertible debt offerings are -5.00%, %, and %, respectively. The measures for the three- and five-year horizons are strongly 349

14 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. significant at the better than one-percent level. By contrast, the mean post-event CARs for the one-, three-, and five-year horizons following straight debt offerings are 2.85%, 1.21%, and 1.56%, respectively, while only the one-year CAR is significant. 5) Note that straight debt issuers with initial bond ratings in the [AAA to A-] category actually realize positive abnormal returns, while straight debt issuers with initial bond ratings in the [CCC+ to D] category realize negative returns. In addition, for both convertible and straight debt issuers, long-term stock performance appears to become more negative as bond quality declines. The % mean CARs for the five-year horizon following convertible debt offerings is comparable to the abnormal performances of -30.4%, %, and % (-19.8% median) for the same horizon obtained in Lee and Loughran (1998), Lewis, Rogalski, and Seward (1999), and Spiess and Affleck-Graves (1999), respectively. By contrast, the insignificant five-year mean CAR of 1.56% for straight debt issuers is comparable only to the 2% (median -18.6%) measure documented in Jung, Kim, and Stulz (1996) but differs significantly from the -14.3% (median -18.7%) measure reported in Spiess and Affleck-Graves (1999). 6) Fama-French Analysis CARs could suffer from a potential problem of cross-sectional dependence among sample firms since events could be subject to time clustering. To overcome this problem, I construct calendar time portfolios of raw returns of debt issuers during the period from 1965 to 1999 and measure the abnormal performance using the Fama- French three-factor model (1992). The return of each calendar month is calculated as the mean return of an equally-weighted portfolio of all sample firms having convertible or straight debt offerings during the one-, three-, or five-year periods preceding the month of inclusion in the portfolio. The excess returns used for the Fama-French three-factor regression are calculated by subtracting the Treasury Bill monthly return from the CRSP equal weighted index return and from the monthly return of convertible or straight debt issuers. Fama-French (FF) intercepts at the bottom part of each panel are estimated using the ordinary least square regression and indicates 5) Mean post-event BHARs are -4.7%, -14.9%, and -18.2% for convertible debt issuers, while they are 3.6%, 2.7%, and 4.5% for straight debt issuers (not reported in Table 2). 6) Jung, Kim, and Stulz (1996) and Lewis, Rogalski, and Seward (1999) use size-matched firms as benchmarks to calculate CARs. Lee and Loughran (1998) and Spiess and Affleck-Graves (1999) use firms in similar size and book-to-market equity ratio to calculate BHARs. 350

15 Asia-Pacific Journal of Financial Studies (2009) v38 n3 the mean monthly abnormal returns. The regression intercepts for the convertible debt samples in Panel A are all negative and significant at the one-percent level over the horizons. The intercept of -0.80% for the five-year horizon means that convertible debt issuers underperform an equally weighted market index by 80 basis points per month over the five-year horizon, after controlling for the three Fama-French risk factors. The number is approximately equivalent to -48% of a five-year abnormal performance and is comparable to the % CARs (-18.2% BHARs) for the five-year horizon following the convertible debt issuance. The intercept for straight debt issuers, shown in Panel B, is -0.23% for the fiveyear horizon, but its magnitude is small. Using 207 convertible bonds and 133 straight bonds, Spiess and Affleck-Graves (1999) report intercepts of -31 and -29 basis points for the five-year horizon, respectively. The bottom row of each panel reports the adjusted Fama-French (Adj FF) intercepts that are proposed by Mitchell and Stafford (2000). The motivation is based on the idea that the traditional three-factor model may not completely explain the crosssection of stock returns since measured abnormal performance is also likely to exist in firms with similar characteristics. To calculate the adjusted FF intercepts, I estimate two separate Fama-French three-factor models for a portfolio of sample issuers and nonissuers (matched on size, book-to-market ratio, and momentum) and then subtract the intercept for the nonissuers from the intercept for the sample issuers. In the case of convertible debt issuers, the adjusted intercepts are -38 and -36 basis points for the three- and five-year horizons, respectively. They are both significant at the five-percent level although their magnitudes are a little smaller than the FF intercepts. The adjusted intercepts for straight debt issuers are -13 and -23 basis points for the three- and five-year horizons, respectively. The one for the five-year horizon is significant at the ten-percent level, but its magnitudes are smaller than that of convertible debt issuers. The number of months indicates the total number of months in the time series under investigation Analysis and Discussion Abnormal returns for the short-term and long-term horizons are negative and significant for convertible debt issuers, while those of straight debt issuers are not significant, on average. The overall results seem to conform to the behavioral framework of Daniel et al. (1998) because the sign of the long-term price reaction is the same as 351

16 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. the sign of the short-term price reaction. Therefore, the long-term underperformance following convertible debt issuance could possibly be interpreted as a result of the short-term underreaction to the selective corporate event that is triggered when the pre-event abnormal returns (24.15%) are high. On the other hand, long-term abnormal returns following straight debt offerings are insignificant, in line with the relatively lower pre-event abnormal returns (6.32%) and insignificant announcement period returns. 5. Long-term Operating Performance following Debt Issuance As a possible reason for post-event abnormal stock performance, the change in issuer s risk is also examined, but I could not find any significant risk changes between pre-event and post-event periods, nor any significant cross-sectional relation between post-event risk changes and long-term abnormal performance following convertible debt issuance. 7) The betas estimated from the market model for pre-event two-year and post-event three-year periods are not significantly different for convertible debt issuers either. Using the Fama-French three-factor model, Loughran and Ritter (1995) find that issuing firms have even slightly higher betas than non-issuing firms and conclude that underperformance subsequent to seasoned equity offerings is not merely a manifestation of differences in betas between issuing and non-issuing firms. Therefore, this section focuses on whether both of convertible debt and straight debt issuers experience any subsequent change in fundamentals and then examines whether this change is somehow related to post-event abnormal stock performance. 5.1 Long-term Operating Performance ROA (return on asset) and ROE (return on equity) to measure firms operating performance are obtained from COMPUSTAT. ROA and ROE are defined as the ratio (%) of income before extraordinary items to the total asset and to the book value of stockholders equity, respectively. Operating performance in this study refers to the 7) To estimate the firm-specific Fama and French regression for the [m-36, m+36] window, where m is the month of debt offering, I run the following regression: R i,t -R f,t = α i + D t α Δi + β i (R m,t -R f,t ) + s i SMB t + h i HML t + β Δi D t (R m,t -R f,t ) + s Δi D t SMB t + h Δi D t HML t + e i,t In this model, D t is a dummy variable equal to one (zero if otherwise) if the month is in the post-event period. 352

17 Asia-Pacific Journal of Financial Studies (2009) v38 n3 changes in ROA and ROE over the horizons of interest. Since ROA and ROE are already expressed in percentage terms and their values could be negative or very small, the use of either another percentage change or numerical differences of these values may sometimes cause a biased comparison, leading to exaggerations. Therefore, I first examine if ROA and ROE of both debt issuers and their matching nonissuers are increased or decreased after issuance, i.e., whether operating performance is positive or negative, without considering the change in magnitude. I then calculate how much the ROA and ROE have changed in real terms and test a cross-sectional relation between abnormal operating and stock performances. Table 3. Long-term Operating Performance (dummy variable) following Debt Issuance The mean changes in ROA and ROE for convertible debt and straight debt issuers from the year before issuance to the years after issuance are shown with those for their matching nonissuers, controlled for pre-event stock performance, firm size, and book-to-market ratio. As longterm operating performance, ROA change is defined as +1 (-1) if ROA in the one, three and five calendar years after issuance is increased (decreased) compared to ROA in the one calendar year before issuance. The difference in the ROA changes between debt issuer and matching nonissuers is shown as Adjusted. ROE change is defined in a similar manner. ROA and ROE are defined as the ratio (%) of income before extraordinary items to the total asset and to the book value of stockholders equity, respectively. ROA change Yr-1 to Yr +1 Yr-1 to Yr +3 Yr-1 to Yr +5 ROE change No obs ROA change ROE change No obs ROA change ROE change Convertible debt issuers Matching nonissuers Adjusted No obs Straight debt issuers Matching nonissuers Adjusted In Table 3, the changes in ROA and ROE for both debt issuers from the year before issuance to the three and five years after issuance are shown along with those for their matching nonissuers, adjusted for pre-event stock performance, firm size, and book-to-market ratio. As operating performance, ROA change is defined as +1 (-1) if the ROA in the three and five calendar years after issuance is increased (decreased) compared to the ROA in the one calendar year before issuance. The difference in the ROA changes between both debt issuers and their matching nonissuers is shown as adjusted. ROE change is defined in a similar manner. 353

18 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. Both ROA and ROE following straight and convertible debt issuance decrease, but convertible debt issuers experience more severe deterioration than straight debt issuers. The number for convertible debt issuers in the third column of Table 3 means that the number of firms whose ROA in the fifth year after an offering is lower than the ROA at the end of fiscal year preceding the offering is 108 ( ). The more negative adjusted number in convertible debt issuers implies that operating performance is negative more frequently in firms issuing convertible debt than in firms issuing straight debt, even after adjusting for matching nonissuers. Table 4. Long-term Operating Performance (real values) following Debt Issuance ROA and ROE for convertible debt (Panel A) and straight debt (Panel B) issuers in the year before issuance and in the three and five years after issuance are shown with those for matching nonissuers controlled for pre-event stock performance, firm size, and book-to-market ratio. As abnormal operating performance, abnormal ROA change is defined as (ROAi,t ROAi,-1) (ROAm,t ROAm,-1), where ROAi,t and ROAm,t are the return on assets for sample debt issuers and matching nonissuers, respectively, t calendar years after issuance. Abnormal ROE change is defined in a similar manner. Panel A: Convertible debt issuer Yr-1 Yr+3 Yr+5 Yr-1 Yr+3 Yr+5 Yr-1 to Yr+3 Yr-1 to Yr+5 Convertible debt issuers Matching nonissuers Abnormal changes ROA(%) ROE(%) No obs Panel B: Straight debt issuers Yr-1 Yr+3 Yr+5 Yr-1 Yr+3 Yr+5 Yr-1 to Yr+3 Yr-1 to Yr+5 Straight debt issuers Matching nonissuers Abnormal changes ROA(%) ROE(%) No obs The first two columns of Table 4 show the actual ROA and ROE for both debt issuers and matching nonissuers in the year before and the three and five years after issuance. As the abnormal operating performance, the abnormal ROA change in the third column is defined as (ROAi,t ROAi,-1) (ROAm,t ROAm,-1). ROAi,t and ROAm,t are the returns on assets for sample issuers and matching nonissuers, respectively, t calendar years after issuance. Since the actual values include some extreme values, the 354

19 Asia-Pacific Journal of Financial Studies (2009) v38 n3 median values are used hereafter for discussion. It is notable that the median ROA/ROE following the convertible debt issuance sharply declines over the three- and five-year horizons. The median ROA (ROE) for convertible debt issuers is 4.21% (11.15%) at the end of the year before issuance and falls to 2.03% (8.83%) five years after issuance. However, the ROA (ROE) for the matching nonissuers has been almost stable from 4.36% (12.67%) to 3.63% (11.71%). This result is consistent with the findings of Lee and Loughran (1998), who report 5.0% of median ROA in the year prior to a convertible debt offering and 2.2% four years after the offering. For the straight debt issuers, the median ROA (ROE) in the year before issuance is 3.24% (12.69%) and drops to 2.50% (12.27%) five years after issuance, while the ROA (ROE) for the matching nonissuers has been changed from 4.93% (13.49%) to 4.04% (12.41%). The median abnormal changes in ROA (ROE) for convertible debt issuers are -0.76% (-1.64%) and -1.34% (-3.08%) in the three and five years after issuance, respectively. These numbers are still more negative compared to -0.12% (-0.70%) and 0.50% (0.97%) for straight debt issuers. Loughran and Ritter (1997) also report a deteriorating operating performance of firms conducting seasoned equity offerings but do not test a cross-sectional relation between operating and stock performances. 5.2 The Relationship between Abnormal Long-term Stock and Operating Performances Table 5 reports the ordinary least square regressions in which the long-term abnormal stock performance following debt issuance is regressed on the contemporaneous abnormal operating performance. This tests whether the subsequent stock performance is also a result of the change in the issuer s fundamentals. The post-event CARs for three- and five-year horizons, which are illustrated in Table 2, are used as the long-term abnormal stock performance. The median abnormal ROA changes for the post three- and five-year periods, which are illustrated in Table 4, are used as the abnormal operating performance. The use of the abnormal operating performance as an independent variable is desirable because the post-event abnormal stock performance as a dependent variable is already adjusted for matching firms. To prevent the results from being dominated by a small number of outliers, the regressions are performed after excluding the top and bottom one percent of the observations. 355

20 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. Table 5. The Relationship between Long-term Abnormal Stock and Operating Performances Shown are the ordinary least square regressions in which long-term abnormal stock performance is regressed on long-term abnormal operating performance, initial bond ratings, and their interaction variable. As long-term abnormal stock performance, post-event cumulative abnormal returns are calculated for the three-, and five-year horizons following convertible and straight debt offerings by subtracting the sum of monthly returns of the matching nonissuer from the sum of monthly returns of the corresponding debt issuer. The matching nonissuers are selected based on pre-event stock performance, firm size, and book-to-market ratio. As long-term abnormal operating performance, abnormal ROA change is defined as (ROAi,t ROAi,-1) (ROAm,t ROAm,-1), where ROAi,t and ROAm,t are the return on assets for sample debt issuers and matching nonissuers, respectively, t calendar years after issuance. The initial bond rating is a cardinal number that is assigned to each letter-based rating. Number one (twentyfive) is assigned to AAA (DDD). t-statistics are in parentheses. ***, **, and * denote significantly different from zero at the 1%, 5%, and 10% levels, respectively. Post-event cumulative abnormal returns (%) Convertible debt issuers Straight debt issuers 3-year 5-year 3-year 5-year Intercept (-4.24) *** (-2.44) ** (0.19) (-0.30) Abnormal ROA change (6.17) *** (5.50) *** (7.69) *** (4.87) *** No obs Intercept (2.77) *** (3.86) *** (3.12) *** (4.90) *** Initial bond rating (-4.10) *** (-5.11) *** (-3.07) *** (-5.15) *** No obs , Intercept (2.32) ** (2.59) *** (1.72) * (2.93) *** Abnormal ROA change (5.19) *** (2.01) ** (5.80) *** (2.70) *** Initial bond rating (-3.29) *** (-3.10) *** (-2.03) ** (-3.41) *** ROA change*bond rating (-0.97) (-0.84) (1.97) ** (-0.03) No obs Adj-R

21 Asia-Pacific Journal of Financial Studies (2009) v38 n3 In the upper part of the table, the coefficients of the abnormal ROA change for both straight and convertible debt issuers are all positive and significant at the onepercent level. The coefficient for the five-year horizon following convertible debt issuance implies that the one percent decrease (increase) of the abnormal change in ROA is related with a 2.92% decrease (increase) in the abnormal stock returns. The intercept terms for convertible debt issuers are also large and significantly negative, suggesting that some factor other than the operating performance, possibly the severe pre-event overvaluation, could also affect the stock underperformance. The coefficient for straight debt issuers is for the same period and is not much different from that of convertible debt issuers, implying that the changes in post-event abnormal stock performance per unit change in operating performance are almost the same regardless of debt type. This is not inconsistent with the earlier conclusion since the less decrease in ROA for straight debt issuers can be reflected in the less decrease in the stock price, but shows that the stock price is related to a firm s fundamentals to a certain degree. However, the intercept terms for straight debt issuers are small in an absolute value and insignificant, contrasting with those for convertible debt issuers. These results suggest that the relatively lesser negative operating performance and no large mispricing for straight debt issuers lead to the lack of underperformance on average, while more negative operating performance and large pre-event stock misvaluation could account for the underperformance of convertible debt issuers. In the middle part of the table, an initial bond rating at the time of issue is used as an independent variable, serving as a proxy of firm quality. The initial bond rating used in this table is a cardinal number assigned to each letter-based rating. Number one (twenty-five) is assigned to AAA (DDD). Fenn (2000) also uses the similar numeric ratings despite the fact that cardinal measure of bond rating assumes each unit change in ratings to have the same effect on long-term stock performance. The initial bond ratings of debt issuers are not adjusted for those of nonissuers since the data for nonissuers are not available. However, in a later section in which convertible debt issuers are directly compared with straight debt issuers, the difference in their bond ratings is used as an independent variable. The coefficients of the initial bond rating are all negative and significant across both types of bonds over the three- and five-year horizons. The higher (lower) the cardinal number (quality of bond), the lower the abnormal returns for both debt issuers, implying that misvaluation is more extreme for lower quality bonds. Lee and Loughran (1998) and Spiess and Affleck- 357

22 Long-term Equity and Operating Performances following Straight and Convertible Debt Issuance in the U.S. Graves (1999) also report more negative long-term abnormal returns for speculativegrade debt offerings compared to investment-grade offerings. In the bottom part of the table, the abnormal ROA change, initial bond rating, and their interaction variable are used as independent variables. The coefficients for the abnormal ROA change are still significant at least at the five-percent level, even after controlling for bond quality. The result stays consistent when using the abnormal ROE change as an independent variable and a debt-to-equity ratio change as an additional independent variable. The change of debt-to-equity ratio is defined as the difference between a debt-to-equity ratio at the end of the offering year and a debt-toequity ratio at the end of the previous calendar year. The coefficients of debt-toequity ratio are not significant across the type of debt while the significance level of the other coefficients remain unchanged (not reported). Figure 1. Long-term Abnormal Stock Performance grouped by Long-term Abnormal Operating Performance (Indirect match) As long-term abnormal stock performance, post-event cumulative abnormal returns (CARs) are calculated for the three- and five-year horizons following convertible debt (CB) and straight debt (SB) offerings by subtracting the sum of monthly returns of the matching nonissuer from the sum of monthly returns of the corresponding debt issuer. The matching nonissuers are selected based on pre-event stock performance, firm size and book-to-market ratio. As long-term abnormal operating performance, abnormal ROA change is defined as (ROAi,t ROAi,-1) (ROAm,t ROAm,-1), where ROAi,t and ROAm,t are the return on assets for sample debt issuers and matching nonissuers, respectively, t calendar years after issuance. The sample group is defined as Category 1 (2) if abnormal ROA change < (>) zero. CARs vs. Abnormal ROA change CARs(%) yr CB (150, 110) 5yr CB (122, 84) 3yr SB (186, 177) 5yr SB (113, 136) Category 1 Category 2 358