Bond and stock market response to unexpected dividend changes

Journal of Empirical Finance 30 (2015) 1–15

Contents lists available at ScienceDirect

Journal of Empirical Finance journal homepage: www.elsevier.com/locate/jempfin

Bond and stock market response to unexpected dividend changes☆ Hui-Ju Tsai a,⁎, Yangru Wu b a b

Washington College, 300 Washington Avenue, Chestertown, MD 21620, United States Rutgers Business School-Newark and New Brunswick, Rutgers University, 1 Washington Park, Newark, NJ 07102, United States

a r t i c l e

i n f o

Article history: Received 18 February 2014 Received in revised form 27 October 2014 Accepted 3 November 2014 Available online 8 November 2014 JEL classification: G12 G14 G35

a b s t r a c t We use comprehensive transaction data from Trade Reporting and Compliance Engine to study the response in corporate bond market to dividend announcements and compare that with the response in stock market. We find that the information content/free cash flow effect dominates the wealth transfer effect in bond market. The relationship between the magnitude of dividend changes and future profitability is weak. However, the reaction in stock and bond markets on announcement dates can be informative about the earnings one year after announcements. Additionally, the reaction of speculative-grade bonds on announcement dates is more informative about future profitability than that of investment-grade bonds. © 2014 Elsevier B.V. All rights reserved.

Keywords: Dividend announcements Institutional-sized bond trades Earnings forecasts

1 . Introduction The announcement of unexpected dividend changes can have different impacts on financial markets. The information content hypothesis states that managers are more informed than outside stakeholders about company performance, and thus unexpected dividend changes are viewed by stakeholders as a signal sent by the managers about a firm's future profitability (see, e.g., Bhattacharya, 1979; Kalay, 1980; and Miller and Rock, 1985). An alternate, although not mutually exclusive, is Jensen's (1986) free cash flow hypothesis, which predicts that the distribution of dividends prevents managers from empire building and wasting resources in poor investment opportunities and therefore benefits both stockholders and bondholders. In contrast, the wealth transfer hypothesis states that the payment of dividends is a transfer of wealth from bondholders to stockholders and is perceived positively by stockholders, but negatively by bondholders. Therefore, these hypotheses suggest that the stock market will react positively (negatively) to the news of unexpected dividend increases (decreases), but their prediction of the reaction in bond market differs. While the information content and free cash flow hypotheses suggest a positive relationship between unexpected dividend changes and bond returns, the wealth transfer hypothesis suggests a reverse relationship. In the literature, the research on the impact of dividend announcements on stock market is extensive, but the studies on bond market response to unexpected dividend changes are either based on a small sample of debt issues or on monthly transaction quotes ☆ We thank the Editor and the anonymous referee for the helpful comments. Wu thanks the Whitcomb Center for Financial Services at the Rutgers Business School for the data and financial support. Part of this work was completed while Wu visited the Central University of Finance and Economics. We are responsible for any remaining errors. ⁎ Corresponding author. E-mail addresses: [email protected] (H.-J. Tsai), [email protected] (Y. Wu).

http://dx.doi.org/10.1016/j.jempfin.2014.11.001 0927-5398/© 2014 Elsevier B.V. All rights reserved.

2

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

from dealers (see, e.g., Dhillon and Johnson, 1994; Handjinicolaou and Kalay, 1984; Jayaraman and Shastri, 1988; and Woolridge, 1983). According to Bessembinder, Kahle, Maxwell, and Xu (2009), studies of corporate bond market based on monthly bond data are not well specified and are likely to incur type I and more significantly type II errors. Moreover, the evidence of the dominant effect of dividend changes on bond returns is inconclusive. Woolridge (1983) and Handjinicolaou and Kalay (1984) find that bond returns are positively correlated with unexpected dividend changes and thus claim that the information effect dominates the wealth transfer effect. By contrast, Dhillon and Johnson (1994) report that bond prices move in opposite directions to stock prices in response to large dividend changes, a phenomenon consistent with the wealth transfer hypothesis. Similarly, Jayaraman and Shastri (1988) find no evidence of significant change in bond returns around the announcements of specially designated dividends. In light of recent improvement in the transparency of corporate bond market, we use transaction data from Trade Reporting and Compliance Engine (TRACE) to study the relationship between unexpected dividend changes and bond returns around dividend announcements. Specifically, we examine corporate bond returns corresponding to 5571 dividend announcements. To the best of our knowledge, this is the first study that uses comprehensive transaction data to examine the reaction in bond market to dividend announcements. We find that both abnormal stock and premium bond returns on dividend announcement dates are positively associated with unexpected dividend changes. Our results imply that the information content/free cash flow effect dominates the wealth transfer effect in bond market. Furthermore, we examine the relationship between unexpected dividend changes and earnings. In the literature, the evidence of the relationship between dividend changes and future profitability is mixed. For instance, Nissim and Ziv (2001) find that dividend changes are positively related to future earnings changes, while Benartzi, Michaely, and Thaler (1997), DeAngelo, DeAngelo, and Skinner (1996), Grullon, Michaely, Benartzi, and Thaler (2005), and Penman (1983) indicate either a weak relationship, no relationship, or opposite relationship between dividend changes and future earnings. Unlike these studies that directly examine the relationship between dividend changes and future profitability, this paper identifies dividend announcements that result in significant abnormal stock or bond returns as predicted by the information content hypothesis because these dividend announcements are more likely to contain information of future profitability. We find the relationship between the magnitude of dividend changes and future profitability to be weak. However, if unexpected dividend increases (decreases) are accompanied by significantly positive (negative) abnormal stock or bond returns on the announcement dates, earnings after one year of announcements will improve (deteriorate). Additionally, the reaction of speculative-grade bonds on announcement dates is more informative about future profitability than that of investment-grade bonds. The relationship between dividend changes and profitability, however, becomes much weaker after two years of announcements, even after considering financial market reaction on announcement dates. The remainder of this paper is organized as follows. Section 2 reviews recent studies on the relationship between stock and corporate bond markets. In Section 3, we describe the data used in the study. Section 4 shows the abnormal stock and premium bond returns around dividend announcements. In Section 5, we examine the relationship between dividend changes and profitability. Section 5 concludes the paper. 2. Literature review According to Merton (1974), investing in corporate bonds can be viewed as having a long position in the riskless asset and a short position in a put option on the firm. On the other hand, buying stock is like purchasing a call option on the firm with exercise price equal to the face value of the corporate debt. When the bond matures, the stockholders decide if they want to exercise their call option by paying the face value of the debt and retain the ownership of the firm or let the option expired and bondholders be the residual claimants. From this point of view, both stock and bond values are positively related to firm value. Additionally, the volatility of firm value is positively related to equity value, but negatively related to bond value. Therefore, we would anticipate seeing that stock and bond returns are positively correlated when the market receives news about firm value but negatively correlated when information about the riskiness of the company arrives. A number of researchers have studied the co-movement of corporate bond and stock markets. For example, Schaefer and Strebulaev (2008) investigate the sensitivity of bond returns to equity returns (hedge ratio) and find that the empirical hedge ratio is similar to what is implied in structural models, such as the one in Merton (1974). Bao and Hou (2013) examine the hedge ratio of bonds issued by the same company with different time to maturity and find that the hedge ratio of de facto junior bonds (i.e., bonds with longer time to maturity) is higher than that of bonds that mature earlier. They also indicate that the sensitivity of corporate bond returns to equity returns increases with credit risk that is measured by book-to-market ratios or distance-to-default even after the control of credit ratings. Similarly, Huang and Shi (2013) examine the yield spread and conclude that the empirical sensitivity of change in credit spreads to equity returns is in line with what is implied in structural models. Thus, although there usually exists a discrepancy between credit spread implied in structural models and that observed in financial markets, current evidence seems to suggest that structural models are good in capturing the credit component of yield spread. Furthermore, several studies point out that illiquidity risk, which is not captured in structural models, plays an important role in determining credit spread and may help to explain part of the yield spread that cannot be justified with structural models. For example, Bao and Pan (2013) show that the volatilities of bond returns are much higher than what are implied in Merton's (1974) model and indicate that the excess volatilities can be explained by the idiosyncratic liquidity risk of corporate bonds. Additionally, Bao, Pan, and Wang (2011) find that the variation in yield spread of investment-grade bonds is significantly related to market-level illiquidity and the liquidity risk can explain the substantial part of yield spread at the individual bond level. Another way to examine the relationship between stock and corporate bond markets is to conduct event studies that compare the effects of various announcements on bond and stock returns. For instance, Wakeman (1978) and Hand, Holthausen, and Leftwich

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

3

(1992) study the excess bond and stock returns associated with the announcements of credit rating changes (also see, e.g., Holthausen and Leftwich, 1986; and Weinstein, 1977). Although Wakeman (1978) finds no reaction in bond prices to rating changes, Hand, Holthausen, and Leftwich (1992) observe bond price effects for actual upgrade and downgrade announcements. Maxwell and Rao (2003), on the other hand, investigate the effects of spin-off announcements on stock and bond returns. They show that there exist significantly negative (positive) abnormal bond (stock) returns during the month of spin-off announcements, implying a wealth transfer from bondholders to stockholders. Consistent with the wealth transfer hypothesis, it is also found that firms are more likely to be downgraded after the spin-off. Another type of announcements that is of interest to both stockholders and bondholders is the announcement of stock repurchases. Dann (1981), for instance, studies the possible wealth transfer effect of tender offer repurchases but does not find supporting evidence. Maxwell and Stephens (2003), on the contrary, find evidence of both signaling and wealth redistribution effects associated with stock repurchases. They show that the loss to bondholders is positively related to the size of the repurchase and the financial risk of the firms. Also, they show that firms are more likely to be downgraded than upgraded after the announcements of stock purchases. The purpose of this study is to conduct an event study that examines the effect of unexpected dividend change announcements on the stock and bond markets. According to Merton (1974), if unexpected dividend changes are interpreted by the market as a signal sent by managers about future performance of the company, we should anticipate that both stockholders and bondholders react positively to the announcements of unexpected dividend increases, and negatively to unexpected dividend cuts (see, e.g., Bhattacharya, 1979; Kalay, 1980; and Miller and Rock, 1985). This line of thought is also supported by the empirical evidence that managers tend to maintain a stable dividend distribution policy and are reluctant to cut dividends. By contrast, the distribution of dividends increases financial leverage which in turn increases firm risk. Merton's (1974) option pricing theory predicts that firm risk is positively related to stock value but negatively to bond value. From this perspective, the announcements of unexpected dividend increases (decreases) should result in positive (negative) stock returns but negative (positive) bond returns. The effect of dividend distribution on the wealth of stock and bond holders can also be explained by agency theory. According to Jensen and Meckling (1976), there exist conflicts of interests between managers and stockholders in that managers can build up their empires or gain non-pecuniary benefits, while most of the price is paid by stockholders. Similarly, Jensen (1986) indicates that managers are more likely to engage in negative NPV projects that decrease firm value when there are more free cash flows under their control. Since the distribution of dividends reduces the amount of free cash flows in the company, Jensen's (1986) free cash flow theory suggests that the announcements of unexpected dividend increases should cause a positive reaction in stock and bond markets. On the other hand, due to the conflict of interests between stockholders and bondholders as indicated in Jensen and Meckling (1976), stockholders may have an incentive to distribute dividends and thus transfer wealth from bondholders to stockholders. From this perspective, bondholders will react negatively to the announcements of dividend increments, while stockholders will react positively. In sum, current theories have different predictions about the reaction in bond market to the announcement of unexpected dividend changes. These interpretations, however, need not to be mutually exclusive. A primary purpose of this study is to examine the net effect of unexpected dividend changes on bond returns and investigate which explanation is likely to dominate the others. Although in the literature several studies have examined bond market reaction to unexpected dividend changes, the evidence is inconclusive. Woolridge (1983) and Handjinicolaou and Kalay (1984), for instance, find that bond returns are positively correlated with unexpected dividend changes. However, consistent with the wealth transfer hypothesis, Dhillon and Johnson (1994) report that bond prices move in opposite directions to stock prices in response to large dividend changes. Jayaraman and Shastri (1988) find no evidence of significant change in bond returns around the announcements of specially designated dividends. Because these studies were conducted while the transparency in corporate bond market was limited, they were either based on a small number of bond issues or on monthly dealers' quotes. In this study, we use comprehensive transaction data from TRACE to examine the reaction in corporate bond market to dividend announcements and compare it with the reaction in stock market. This investigation allows us to better understand the impact of unexpected dividend changes on corporate bond market and shed light on the relationship between stock and corporate bond markets. 3. Data selection 3.1. Sample description We obtain dividend announcements data from the Center for Research in Securities Prices (CRSP) and identify all announcements of regular cash dividends made during the period from July 1, 2005 through December 31, 2012. We exclude special dividend announcements or dividend announcements with a distribution code that is different from their previous announcements. We also exclude the announcements made by the firms in either the finance or insurance industry and are left with a sample of 37,997 dividend announcements.1 As firms may make earnings and dividend announcements within a few days of one another, to eliminate the confounding effect from earnings announcements, we exclude the dividend announcements that are made within three days of earning announcements and are left with 29,580 events. We acquire bond transaction data from TRACE and then employ a set of screens to remove erroneous trades, including the trades that are canceled, corrected, or reversed. A screen test is implemented to ensure that no trade has more than 30% price reversal from around trades. In addition, only trades that have complete information of CUSIP, bond symbol, company name, execution time, price, 1 Firms are identified as in the finance or insurance industry if their NAICS code is 52 or if their SIC code is between 60 and 67 (excluding 65) when NAICS code is not available from the data.

4

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

and trading volume are considered. Due to the institutional and illiquid nature of corporate bond market, we restrict our sample to those announcements made by the firms that have institutional-sized bond trades, either on dividend announcement date or one day after that and are left with 5571 events.2 To determine unexpected dividend changes, a dividend expectation model is required. Following Handjinicolaou and Kalay (1984), we assume Et(Di,t+1) = Di,t, where Di,t is the amount of dividends paid by firm i at time t. That is, we assume that the unbiased estimate of dividends to be paid in the next period is equal to the amount of the previously paid dividends. The percentage change in D −D unexpected dividends for company i at time t is thus defined as i;t Di;t−1i;t−1. Using the financial data from CRSP and COMPUSTAT at the year of the announcement, we estimate the quarterly percentage change in dividends, market value, return on asset (ROA), and credit ratings of the firms that make the announcements. The result is provided in Table 1. Out of the 5571 events, 1067 are positive dividend announcements, 150 are negative announcements, and 4354 have no dividend changes. Consistent with previous studies (see, e.g., Aharony and Swary, 1980; and Yoon and Starks, 1995), most dividend announcements show no dividend changes, and there are more announcements that have dividend increases than those that have dividend decreases. The average percentage change of dividend increases is 32.72%, which is higher in absolute value than the −27.39% of dividend decreases. The median change of dividend increases is 10.71%, whereas the median change of dividend decreases is −15.52%. Consistent with prior studies (see, e.g., Grullon et al., 2005), firms with positive dividend announcements have higher market values and profitability than their counterparts that make neutral or negative dividend announcements. The average market value of firms that make positive dividend announcements is $32,031 million, compared to the value of $25,691 million for firms that make neutral dividend announcements, which in turn is higher than the value of $18,673 million for firms making negative announcements. In regard to profitability, the average ROA for firms that give positive dividend surprises is 6.64%, higher than the values of 4.98% and 3.97% for firms that make neutral and negative dividend announcements, respectively. It is not surprising that firms making positive dividend announcements also tend to have better credit ratings than those making neutral or negative announcements. About 34% of positive dividend announcements are made by investment-grade firms, while only 15% of negative announcements are made by investment-grade companies. There is also a higher percentage of firms that make negative announcements with no firm ratings. 3.2. Actively traded bonds According to Ronen and Zhou (2013), retail trades constitute about 65% of total bond trades in corporate bond market, but represent less than 2% of total trading volume. They indicate that it is important to differentiate between retail-sized and institutional-sized bond trades when examining the informational efficiency in corporate bond market (see also, e.g., Bessembinder, Kahle, Maxwell, and Xu, 2009). They show that the informational efficiency of the bond market is greatly improved when only institutional-sized bond trades are taken into account. To better discern the impact of dividend announcements on bond market, we construct bond returns based on all bond trades and institutional-sized bond trades, respectively. For each trading day, we define daily bond return (Rbi,t) as

b Ri;t

¼

  P i;t þ Ii;t þ C i;t − P i;t−1 þ Ii;t−1 P i;t−1 þ Ii;t−1

where Ci,t is the coupon paid by bond i at time t, and Pi,t and Ii,t denote volume weighted clean bond price and accrued interests of bond i at time t, respectively.3 For a bond that has no transactions on a trading day, we assume its volume weighted price is the same as its previously determined volume weighted price. Compared to the study of stock market, the examination of bond market reaction to dividend announcements presents researchers with some challenges. First of all, although most firms have one stock issue outstanding, they usually have multiple bond issues traded in the market. Secondly, these multiple bond issues may not have the same liquidity in the market. Ronen and Zhou (2013) find that the transactions usually concentrate on certain bonds in the market. In the literature, researchers use different ways to address the multiple-bond problem. We follow Asquith and Wizman (1990), DeFusco, Johnson, and Zorn (1990), and Dhillon and Johnson (1994) by selecting a representative bond to study bond market response to dividend announcements. Specifically, for each dividend announcement, we define the most actively traded bond as the one with the highest volume of institutional-sized bond trades on the announcement date or one day after if no institutional-sized bond trades occur on that date. We then study bond market response to unexpected dividend changes by examining the abnormal premium returns of the most actively traded bonds around dividend announcements. Selecting a representative bond that is actively traded by institutional investors allows us to reduce the potential measurement problem when the actively traded bond is far more liquid than the other bonds issued by the same company. In addition, the announcement of dividend distribution may have different impacts on various bonds. Holders of bonds with a longer time to maturity, for instance, may care more about the distribution of dividends than those whose bonds will mature soon. According to Ronen and Zhou (2013), the most actively traded bonds by institutional investors are more likely to be a long-term bond and most recently issued. Since the primary purpose of this study is to uncover the possible reaction, if any, from bondholders to the announcements 2 3

Following Bessembinder et al. (2009), we define a bond trade as an institutional-sized trade if its par value is $100,000 or above. Trades with par values of “+5MM” or “+1MM” in TRACE are assumed to have par values of $5,000,000 and $1,000,000, respectively.

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

5

Table 1 Description of sample. This table provides descriptive statistics of quarterly percentage change in dividends, market value, return on asset (ROA), and credit ratings of the firms making dividend announcements. The data is based on the year of the announcement. The descriptive statistics are provided separately for firms making positive, neutral, and negative dividend announcements. Sample size

Quarterly percentage change in dividends (%)

Market value ($ million)

ROA (%)

Firm ratings (% of total) Investment-grade

BB or B

C or D

Not rated

6.64 6.25 5.23

34

41

1

24

25,691 10,704 42,786

4.98 4.88 7.10

24

53

2

22

18,673 13,188 22,851

3.97 3.93 6.93

15

43

7

35

Positive announcements Average 1067 Median Standard deviation

32.72 10.71 127.74

32,031 16,159 47,851

Neutral announcements Average 4354 Median Standard deviation

0.00 0.00 0.00

Negative announcements Average 150 Median Standard deviation

−27.39 −15.52 −29.35

of dividend distribution instead of measuring the change in bondholders' wealth as a whole, we study the abnormal returns of the most actively traded bonds that are likely to be affected by dividend announcements. Table 2 displays the descriptive statistics of the most actively traded bonds in our final sample. For comparison purposes, the results are provided separately for bonds traded around positive, neutral, and negative dividend announcements. The average offering amount of bonds that are actively traded around positive dividend announcements is $664 million, similar to the amount of $636 and $708 million of bonds traded around neutral and negative announcements, respectively. There is not much difference in age or time to maturity among bonds traded around different dividend announcements. Bonds that are actively traded around positive dividend announcements have an average age of 3.20 years and a remaining time to maturity of 11.35 years, while bonds traded around negative announcements have an average age of 3.56 years and a remaining time to maturity of 12.27 years. In terms of coupon rates, bonds traded around positive announcements have an average rate of 5.46%, lower than the rates of 5.67% and 6.29% for bonds traded around neutral and negative announcements, respectively. Additionally, bonds traded around positive dividend announcements have better credit ratings than bonds traded around negative announcements, and bonds traded around neutral dividend announcements have credit ratings in between. For instance, about 37% of bonds traded around positive announcements are rated at A or above, while only 17% of bonds traded around negative announcements have ratings at A or above. Similarly, 12% of bonds traded around positive announcements are speculative-grade, but about 24% of bonds traded around negative announcements are rated at BB or below. This is consistent with the result shown in Table 1, which indicates that the firms that make positive dividend announcements are more likely to have better financial performance and credit ratings than those that make neutral or negative dividend announcements.

Table 2 Descriptive statistics of actively traded bonds around dividend announcements. This table provides descriptive statistics of offering amount, coupon rate, age, remaining time to maturity, and credit ratings of bonds that are actively traded around dividend announcements. For each dividend announcement, an actively traded bond is identified as the one with the highest trading volume of institutional-sized trades at the time of dividend announcements. The statistics are based on the year of the announcement and provided separately for bonds traded around positive, neutral, and negative dividend announcements. Sample size

Positive announcements Average Median Standard deviation Neutral announcements Average Median Standard deviation Negative announcements Average Median Standard deviation

Offering amount ($ thousand)

Coupon rate (%)

Age

Remaining time to maturity

Bond ratings (% of total) A or above

BBB

BB or B

C or D

Not rated

1067

663,572 500,000 589,232

5.46 5.75 1.99

3.20 2.40 2.96

11.35 7.91 10.37

37

45

12

0

6

4354

635,738 500,000 576,330

5.67 5.90 2.04

3.09 2.30 2.86

10.01 7.26 9.18

30

40

23

1

6

150

708,496 500,000 604,180

6.29 6.28 1.33

3.56 2.83 3.04

12.27 7.38 11.37

17

58

23

1

1

6

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

4. Methodology and empirical results 4.1. Methodology To examine stock and bond market response to unexpected dividend changes, we follow the methodology in Handjinicolaou and Kalay (1984) to compute the abnormal stock and premium bond returns around dividend announcement dates. The expected stock and premium bond returns are estimated with the mean of stock and premium bond returns during the period from t = − 60 to t = − 16, where t = 0 is the dividend announcement date. The same time interval is also used to estimate the standard deviation of stock and premium bond returns. We define premium bond return as the difference between bond returns and five-year Treasury note returns. In order to better estimate mean and standard deviation of premium bond returns, we exclude dividend announcements from our analysis if their actively traded bonds have less than eight returns during the period from t = − 60 to t = − 16. This screening leaves us with a final sample of 5143 (5246) events based on institutional-sized bond trades (all bond trades). We then use the estimated mean and standard deviation of stock and premium bond returns to standardize the daily stock and premium bond returns around dividend announcements from t = −8 to t = 8.4 The standardized stock and premium bond returns have a student-t distribution with a mean of 0 and degrees of freedom equal to ns − 2 and nb − 2 respectively, where ns and nb are the number of days with stock and bond returns during the period from t = −60 to t = −16. Following Handjinicolaou and Kalay (1984), we use the standardized stock and bond returns to form equally weighted stock and bond portfolios. According to the central limit theorem, the return of the equally weighted stock (bond) portfolio is approximately normally distributed with mean 0 and variance 1/n, where n is the number of stocks (bonds) in the portfolio.

4.2. Abnormal stock returns around dividend announcements Because the market may have an asymmetric response to positive/negative dividend announcements and the abnormal premium bond returns may vary with credit ratings, our sample is categorized into six groups: 1) Invement_Positive, 2) Investment_Negative, 3) Investment_Neutral, 4) Speculative_Positive, 5) Speculative_Negative, and 6) Speculative_Neutral, where Positive/Negative/ Neutral refers to the direction of unexpected dividend changes and Investment/Speculative is defined based on the rating of the most actively traded bonds around dividend announcements.5 To see if there exists a systematic difference in unexpected dividend changes among these groups, we report the average, median and standard deviation of percentage dividend changes for the Positive/Negative groups and the result is displayed in Table 3.6 It shows that the percentage dividend change is higher in magnitude in the speculative group than in the investment group. For dividend increases, the average (median) percentage change is 24 (10) percent for the investment group, compared to the value of 78 (17) percent for the speculative group. Similarly, for dividend decreases, the average (median) percentage change is − 23 (− 9) percent for the investment group, compared to the value of − 43 (− 41) percent for the speculative group. The standard deviation of percentage change in dividends is also higher in the speculative group than in the investment group. We then estimate the abnormal stock portfolio returns around dividend announcements from t = −8 to t = 8. The results for the investment and speculative groups are provided in Tables 4 and 5 respectively. Consistent with previous research (see, e.g., Aharony and Swary, 1980; Asquith and Mullins, 1983; Eades, 1982; Kalay and Loewenstein, 1985; Kwan, 1981; Petit, 1972; Woolridge, 1983; and Yoon and Starks, 1995), abnormal stock returns are positively correlated with unexpected dividend changes.7 The mean abnormal return of speculative stock portfolio around dividend announcements is higher in magnitude than that of investment stock portfolio. The abnormal return of the Investment_Positive stock portfolio on the announcement date is 0.28%, whereas the abnormal return of the Speculative_Positive portfolio is 0.55%. Similarly, the abnormal return of the Investment_Negative stock portfolio on the announcement date is −0.33%, while the abnormal return of the Speculative_Negative portfolio is −0.91%. For the portfolios with unexpected dividend changes, the abnormal returns of all except the Speculative_Negative stock portfolio are significant with expected sign on the announcement dates. As to the announcements without dividend changes, the abnormal returns of the investment and speculative portfolios are both close to 0 and not statistically significant. The last two columns in each panel in Tables 4 and 5 show the number of stocks with positive and negative abnormal returns in the portfolio. Not surprisingly, we see more stocks with positive (negative) abnormal returns within the first few days after the announcements of unexpected dividend increases (decreases). Interestingly, the Speculative_Positive stock portfolio has significant abnormal returns with expected sign on t = − 1. The number of stocks with positive abnormal returns is also increasing in the Speculative_Positive portfolio on the same day. It seems to suggest that the stock market reacts before the announcements are made.

4 In a study not presented here, we estimate the standardized stock and premium bond returns around dividend announcements from t = −10 to t = 10 and obtain similar patterns. The results are available from the authors upon request. 5 Bonds rated at BBB or above are investment-grade. 266 (277) bonds that have no credit ratings are not included in the computation of bond portfolio returns based on institutional-sized (all) bond trades. Of these 266 (277) bonds, 46 (50) correspond to positive dividend changes, 2 (2) to negative dividend changes, and 218 (225) to neutral dividend changes. 6 We do not provide statistics for the Neutral groups because by definition, their dividend changes are zero. 7 Our estimates of abnormal stock returns are comparable but somewhat smaller in magnitude than those found in Aharony and Swary (1980) and Yoon and Starks (1995). This may be due to different samples or the approach used to estimate the expected returns.

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

7

Table 3 Percentage changes in dividends. This table shows the percentage changes in dividends for four groups of events: 1) Investment_Positive, 2) Investment_Negative, 3) Speculative_Positive, and 4) Speculative_Negative, where Positive/Negative refers to the direction of unexpected dividend changes and Investment/Speculative is defined based on the ratings of the most actively traded bonds at the time of dividend announcements.

Average Median Standard deviation Sample size

Investment_Positive

Investment_Negative

Speculative_Positive

Speculative_Negative

0.24 0.10 0.85 876

−0.23 −0.09 0.27 112

0.78 0.17 1.91 132

−0.43 −0.41 0.32 36

4.3. Abnormal premium bond returns around dividend announcements We estimate abnormal premium bond returns based on institutional-sized bond trades. Tables 6 and 7 display the abnormal premium returns around dividend announcements from t = −8 to t = 8 of the investment and speculative groups, respectively. The results show that, consistent with the information content/free cash flow hypothesis, the abnormal premium returns are generally positively correlated with unexpected dividend changes. For investment-grade bonds, the abnormal portfolio returns are significantly positive around the announcements of unexpected dividend increases, while the abnormal returns are not significant for unexpected dividend decreases. The mean excess premium return of the Investment_Positive portfolio on the announcement date is 0.08%, and for the Investment_Negative portfolio the return is −0.08% but not significant. For speculative-grade bonds, the mean excess premium return around the announcements of unexpected dividend changes is significant with the expected sign as predicted by information content hypothesis. The mean excess premium return of the Speculative_Positive portfolio on the announcement date is 0.11%, while the return of the Speculative_Negative portfolio is − 0.34%. As to the portfolios corresponding to no dividend changes, the mean excess premium returns of the investment and speculative groups are very small in magnitude (0.01% and −0.04%, respectively), albeit statistically significant. Similar to what is found in the stock portfolio, the mean excess premium bond return on announcement dates is larger in magnitude for speculative-grade bonds than for investment-grade bonds. This may be due to the fact that the percentage change in dividends is higher in magnitude for the speculative group than for the investment group in our sample. Alternatively, it may be explained by the way through which bond derives its value. Companies issuing investment-grade bonds are less likely to default, and thus the values of investment-grade bonds are more related to general interest rates than to company values. On the other hand, for speculative-grade bonds, when a company performs poorly, stockholders can exercise their options to put the firm back to bondholders and bondholders can only receive whatever is left in the firm (see Merton, 1974). As a result, the return of speculative-grade bonds is more related to company values (see, e.g., Downing, Underwood, and Xing, 2009; Hotchkiss and Ronen, 2002; and Kwan, 1996). If unexpected dividend changes are perceived by bondholders as an indicator of changes in company values, we would expect to see that speculative-grade bonds are more responsive to unexpected dividend changes than investment-grade bonds.

Table 4 Abnormal returns of stock portfolios in the investment group around dividend announcements. This table shows the abnormal returns of equally weighted stock portfolios in the investment group around dividend announcements. The announcement is made when t = 0, whereas t = −k and t = k denote the kth business day before and after the announcement date, respectively. The test statistic of the mean excess return is provided in the column next to it. No. positive (No. negative) shows the number of stocks with positive (negative) abnormal returns in the portfolio. Dividend increase

Dividend decrease

No dividend change

Event day (t)

Mean excess return (%)

Test statistic

No. positive

No. negative

Mean excess return (%)

Test statistic

No. positive

No. negative

Mean excess return (%)

Test statistic

No. positive

No. negative

−8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8

0.07 −0.16 −0.04 0.09 0.07 −0.01 0.00 −0.06 0.28 0.24 0.07 0.00 0.05 −0.02 −0.03 −0.08 −0.02

−0.10 −2.95 0.49 1.85 0.04 0.38 0.53 −0.78 4.22 4.35 1.58 −0.02 1.24 −0.51 −0.65 −1.82 −0.86

406 370 402 428 410 406 405 384 449 447 426 389 412 411 395 399 419

407 443 411 385 403 407 408 429 364 366 387 424 401 402 418 414 394

−0.28 −0.27 0.18 1.29 1.10 −0.28 0.12 −0.14 −0.33 −0.15 −0.21 −0.19 −0.34 0.24 −0.21 0.44 0.00

−0.65 −0.94 0.91 2.54 1.37 −1.52 −1.19 −1.07 −2.67 0.35 −0.88 0.12 −0.97 1.00 −0.88 0.66 −0.94

46 53 53 64 55 48 49 50 50 55 52 48 51 55 45 56 52

60 53 53 42 51 58 57 56 56 51 54 58 55 51 61 50 54

0.04 −0.01 0.01 −0.05 −0.02 −0.04 0.03 −0.04 0.00 0.04 0.04 −0.05 0.06 0.04 −0.01 −0.03 0.01

1.28 −0.88 −1.22 −2.47 −1.20 −1.78 0.95 −0.96 −0.43 1.30 0.75 −2.41 0.73 −0.37 −1.26 −2.77 −0.48

1393 1407 1380 1359 1406 1382 1406 1391 1394 1438 1417 1369 1422 1432 1415 1368 1394

1437 1423 1450 1471 1424 1448 1424 1439 1433 1392 1413 1461 1408 1398 1415 1462 1436

8

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

Table 5 Abnormal returns of stock portfolios in the speculative group around dividend announcements. This table shows the abnormal returns of equally weighted stock portfolios in the speculative group around dividend announcements. The announcement is made when t = 0, whereas t = −k and t = k denote the kth business day before and after the announcement date, respectively. The test statistic of the mean excess return is provided in the column next to it. No. positive (No. negative) shows the number of stocks with positive (negative) abnormal returns in the portfolio. Dividend increase

Dividend decrease

No dividend change

Event day (t)

Mean excess return (%)

Test No. No. Mean excess statistic positive negative return (%)

Test No. No. Mean excess statistic positive negative return (%)

Test No. No. statistic positive negative

−8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8

0.14 0.05 0.42 −0.27 0.13 0.05 −0.27 0.37 0.55 0.56 −0.14 0.00 −0.07 0.08 0.09 0.05 0.01

1.06 0.52 2.53 −0.95 0.94 −0.60 −0.60 2.26 3.02 4.15 0.56 −0.09 −0.39 1.56 −0.90 −0.36 0.23

2.06 0.11 −0.96 1.80 0.99 1.24 −0.37 0.46 −0.63 −0.81 −0.31 −0.23 1.38 0.86 −0.50 0.20 0.01

1.35 −1.39 −0.01 1.04 0.10 −0.80 −0.81 −2.62 −0.35 0.85 −0.77 −1.95 1.13 −0.19 1.44 −1.09 0.25

65 62 67 56 64 52 62 68 71 67 63 54 63 60 58 64 57

58 61 56 67 59 71 61 55 52 56 60 69 60 63 65 59 66

0.61 0.67 −0.52 1.58 1.05 1.00 −0.03 0.68 −0.91 −0.63 −0.21 −0.09 0.45 0.62 0.16 −0.33 −0.09

19 18 21 19 20 21 19 17 17 16 19 14 21 19 19 23 18

16 17 14 16 15 14 16 18 18 19 16 21 14 16 16 12 17

0.23 −0.08 −0.06 0.10 0.01 0.09 −0.01 −0.13 0.01 0.11 −0.03 −0.19 0.04 −0.01 0.06 0.00 0.02

496 451 473 488 484 459 461 439 474 492 459 471 478 478 461 456 456

474 519 497 482 486 511 509 531 495 478 511 499 492 492 509 514 514

For comparison purposes, we also compute the abnormal premium bond returns based on all bond trades. The results of the investment and speculative group are reported in Tables 8 and 9, respectively. Similar to what we find from the institutional-sized bond trades, the mean excess premium bond returns based on all bond trades are positively correlated with unexpected dividend changes. The abnormal premium return of the Investment_Positive portfolio on the announcement date is 0.07%, whereas the abnormal premium return of the Speculative_Positive portfolio is 0.12%, which is statistically significant at the 10% significance level. Similarly, the abnormal premium return of the Investment_Negative portfolio on the announcement date is − 0.07%, while the abnormal premium return of the Speculative_Negative portfolio is − 0.35%. Therefore, consistent with the results found in institutional-sized bond trades, the information effect of dividend announcements dominates the wealth transfer effect.

Table 6 Abnormal returns of bond portfolios based on institutional-sized bond trades in the investment group around dividend announcements. This table shows the mean excess premium returns of equally weighted investment-grade bond portfolios around dividend announcements. The announcement is made when t = 0, whereas t = −k and t = k denote the kth business day before and after the announcement date, respectively. The test statistic of the mean excess premium return is provided in the column next to it. No. positive (No. negative) shows the number of bonds with positive (negative) excess premium returns in the portfolio. Dividend increase

Dividend decrease

No dividend change

Event day (t)

Mean excess premium return (%)

Test statistic

No. positive

No. negative

Mean excess premium return (%)

Test statistic

No. positive

No. negative

Mean excess premium return (%)

Test statistic

No. positive

No. negative

−8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8

0.01 −0.01 −0.01 0.03 0.02 0.01 −0.04 −0.02 0.08 0.03 0.00 −0.02 0.01 −0.01 0.02 0.02 0.02

−0.60 0.11 0.03 0.12 1.47 −0.61 −2.72 −1.28 3.18 1.13 −1.04 −0.07 0.01 −0.15 1.25 0.04 1.25

408 397 409 419 407 388 398 381 429 418 392 375 415 410 405 415 408

405 416 404 394 406 425 415 432 384 395 421 438 398 403 408 398 405

−0.17 −0.22 −0.04 0.18 0.10 0.03 −0.02 −0.28 −0.08 0.08 −0.08 0.03 0.01 −0.01 0.12 −0.06 0.06

−1.64 −1.05 0.90 0.50 −0.27 −0.05 0.23 −3.05 −0.15 0.60 −1.94 −0.17 −0.13 −0.97 −0.41 −0.86 0.03

44 55 55 58 55 53 54 48 54 59 51 51 55 45 49 47 59

62 51 51 48 51 53 52 58 52 47 55 55 51 61 57 59 47

0.01 0.00 0.01 −0.01 −0.02 −0.03 −0.01 −0.01 0.01 −0.01 0.02 −0.01 0.02 0.01 0.08 −0.04 −0.02

0.17 −0.37 0.72 −0.40 −1.80 −2.53 −2.41 −0.40 2.09 −0.56 1.05 −2.34 1.11 0.76 7.21 −3.80 −0.71

1379 1415 1404 1419 1365 1346 1378 1419 1455 1424 1378 1336 1427 1419 1420 1341 1395

1451 1415 1,426 1411 1465 1484 1452 1411 1372 1,406 1452 1494 1403 1411 1410 1489 1435

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

9

Table 7 Abnormal returns of bond portfolios based on institutional-sized bond trades in the speculative group around dividend announcements. This table shows the mean excess premium returns of equally weighted speculative-grade bond portfolios around dividend announcements. The announcement is made when t = 0, whereas t = −k and t = k denote the kth business day before and after the announcement date, respectively. The test statistic of the mean excess premium return is provided in the column next to it. No. positive (No. negative) shows the number of bonds with positive (negative) excess premium returns in the portfolio. Dividend increase

Dividend decrease

No dividend change

Event day (t)

Mean excess premium return (%)

Test statistic

No. positive

No. negative

Mean excess premium return (%)

Test statistic

No. positive

No. negative

Mean excess premium return (%)

Test statistic

No. positive

No. negative

−8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8

0.02 −0.16 0.10 0.02 −0.01 −0.05 −0.02 0.02 0.11 0.10 −0.11 −0.03 −0.08 0.06 −0.06 0.17 −0.05

0.06 −1.64 1.70 −0.19 −0.88 −1.71 −1.17 −0.54 3.38 2.51 0.68 −0.69 −0.63 1.85 −0.41 3.88 −0.94

61 61 62 64 62 51 53 64 73 65 70 61 61 67 58 73 61

62 62 61 59 61 72 70 59 50 58 53 62 62 56 65 50 62

0.29 0.19 0.01 −0.04 0.27 0.11 0.16 0.04 −0.34 −0.37 0.29 0.16 0.12 −0.32 0.06 0.15 0.01

2.29 1.63 −0.38 −0.76 2.02 1.30 0.32 0.30 −3.34 −0.77 1.50 0.45 1.12 −0.59 −0.55 0.94 −0.25

25 23 24 19 24 19 17 16 18 17 21 25 21 20 21 26 20

10 12 11 16 11 16 18 19 17 18 14 10 14 15 14 9 15

0.00 0.01 0.00 0.05 0.05 −0.01 −0.03 −0.04 −0.04 0.01 0.01 −0.02 −0.01 −0.05 −0.04 −0.09 0.02

−1.34 0.53 0.95 2.23 2.55 −1.11 −0.93 −3.38 −1.86 0.71 0.56 −1.22 1.27 −3.32 −2.64 −2.60 0.10

469 499 466 493 500 472 470 445 521 478 466 468 488 478 458 454 485

501 471 504 477 470 498 500 525 448 492 504 502 482 492 512 516 485

4.4. The relationship between dividend changes and profitability The results in Section 4 show that the information content/free cash flow effect of dividend announcements dominates the wealth transfer effect in bond market. According to the information content hypothesis, unexpected dividend changes are positively correlated with abnormal stock and bond returns because these changes convey information about future profitability (see, e.g., Miller and Modigliani, 1961; and Nissim and Ziv, 2001). However, previous studies that examine the relationship between dividend changes and future profitability have obtained mixed empirical results. For instance, Nissim and Ziv (2001) find that dividend changes are positively related to future earnings changes, supporting the information content hypothesis. By contrast, Benartzi et al. (1997), DeAngelo et al. (1996), Grullon et al. (2005), and Penman (1983) find either a weak relationship, no relationship, or opposite relationship between dividend changes and future profitability. Table 8 Abnormal returns of bond portfolios based on all bond trades in the investment group around dividend announcements. This table shows the mean excess premium returns of equally weighted investment-grade bond portfolios around dividend announcements. The announcement is made when t = 0, whereas t = −k and t = k denote the kth business day before and after the announcement date, respectively. The test statistic of the mean excess premium return is provided in the column next to it. No. positive (No. negative) shows the number of bonds with positive (negative) excess premium returns in the portfolio. Dividend increase

Dividend decrease

No dividend change

Event day (t)

Mean excess premium return (%)

Test statistic

No. positive

No. negative

Mean excess premium return (%)

Test statistic

No. positive

No. negative

Mean excess premium return (%)

Test statistic

No. positive

No. negative

−8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8

0.01 −0.02 −0.01 0.03 0.01 0.01 −0.04 −0.02 0.07 0.03 0.00 −0.02 0.01 −0.01 0.02 0.03 0.02

−0.85 −0.38 −0.67 0.29 1.42 −1.00 −2.33 −0.99 2.21 0.97 −1.09 −0.30 0.04 −0.05 0.15 0.30 0.71

407 395 409 423 407 395 403 391 427 421 401 381 423 410 414 413 407

417 429 415 401 417 429 421 433 397 403 423 443 401 414 410 411 417

−0.17 −0.23 −0.04 0.17 0.09 0.00 −0.03 −0.31 −0.07 0.06 −0.10 0.01 0.01 0.01 0.12 −0.08 0.06

−1.71 −1.12 1.04 0.78 −0.05 −0.32 −0.02 −2.68 1.03 0.30 −1.68 −0.54 −0.38 −0.74 −0.25 −1.02 −0.17

44 58 57 62 59 55 55 51 56 58 51 49 58 49 57 50 63

66 52 53 48 51 55 55 59 54 52 59 61 52 61 53 60 47

0.01 0.00 0.00 −0.01 −0.03 −0.04 −0.02 −0.01 0.01 −0.01 0.02 −0.01 0.02 0.01 0.07 −0.04 −0.02

0.12 −0.74 0.52 −1.00 −1.26 −2.77 −1.86 −0.66 −0.12 −0.03 0.44 −2.05 0.47 0.43 7.18 −4.26 −0.25

1407 1424 1416 1443 1386 1360 1373 1,432 1488 1442 1394 1349 1449 1437 1440 1356 1413

1484 1467 1475 1448 1505 1531 1518 1459 1400 1,449 1497 1542 1442 1454 1451 1535 1478

10

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

Table 9 Abnormal returns of bond portfolios based on all bond trades in the speculative group around dividend announcements. This table shows the mean excess premium returns of equally weighted speculative-grade bond portfolios around dividend announcements. The announcement is made when t = 0, whereas t = −k and t = k denote the kth business day before and after the announcement date, respectively. The test statistic of the mean excess premium return is provided in the column next to it. No. positive (No. negative) shows the number of bonds with positive (negative) excess premium returns in the portfolio. Dividend increase

Dividend decrease

No dividend change

Event day (t)

Mean excess premium return (%)

Test statistic

No. positive

No. negative

Mean excess premium return (%)

Test statistic

No. positive

No. negative

Mean excess premium return (%)

Test statistic

No. positive

No. negative

−8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8

0.02 −0.15 0.10 0.03 0.00 −0.05 −0.02 0.03 0.12 0.11 −0.11 −0.02 −0.08 0.05 −0.05 0.17 −0.05

−0.05 −1.09 1.75 0.09 −0.43 −1.30 −0.91 0.14 1.81 1.43 0.00 −0.36 −0.32 1.23 −0.95 3.25 −0.80

64 64 63 65 66 51 54 65 76 67 72 65 65 70 60 75 63

61 61 62 60 59 74 71 60 49 58 53 60 60 55 65 50 62

0.29 0.18 0.00 −0.05 0.26 0.11 0.15 0.03 −0.35 −0.38 0.28 0.15 0.11 −0.32 0.05 0.14 0.00

1.87 1.43 −0.17 −0.58 1.86 1.16 0.06 0.22 −2.88 −0.52 1.04 0.44 1.10 −0.43 −0.17 0.77 −0.27

25 25 25 20 24 17 16 15 17 16 21 23 21 20 20 24 20

10 10 10 15 11 18 19 20 18 19 14 12 14 15 15 11 15

0.00 −0.01 0.00 0.04 0.05 −0.01 −0.04 −0.05 −0.05 0.00 0.00 −0.03 −0.02 −0.06 −0.05 −0.10 0.01

−1.00 1.02 0.20 1.15 2.19 −0.64 −1.32 −2.60 −1.11 −0.25 0.54 −1.21 1.24 −2.39 −2.43 −1.93 0.76

473 492 470 491 497 471 474 447 522 478 470 468 481 485 466 460 481

511 492 514 493 487 513 510 537 461 506 514 516 503 499 518 524 503

Looking at the individual stock or bond return within the portfolio, however, we find that not all stocks and bonds have significant abnormal returns on announcement dates.8 If the abnormal stock or premium bond return is resulted from the information contained in unexpected dividend changes, we would expect that future profitability may be better predicted by those dividend changes that result in significant abnormal stock or premium bond returns on announcement dates. Different from previous studies that directly investigate the relationship between dividend changes and future profitability, we examine the relationship between dividend changes and future earnings, conditional on those announcements that result in significant market reactions. For comparison purposes, we firstly run the following regression without differentiating dividend announcements: ðEτ −Eτ−1 Þ=B−1 ¼ β0 þ β1 %div0 þ ετ ;

ð1Þ

for τ = 1 and 2, where Eτ denotes Earnings before Interest, Taxes, Depreciation, and Amortization (EBITDA) in year τ, B−1 is the book value of assets in year τ = − 1, %div0 denotes the quarterly percentage change in dividends, and τ = 0 is the year of dividend announcement.9 For monthly, semi-annual, and annual dividend changes, we convert the percentage dividend change into quarterly change in the regression analysis. Then we consider the possibility that the announcements that result in significantly abnormal stock or bond returns are more likely to contain information about firm value by running the following regressions: ðEτ −Eτ−1 Þ=B−1 ¼ β0 þ β1 %div0 þ β2 SD  SInfo þ β3 SD  BInfo þ ωτ ;

ð2Þ

ðEτ −Eτ−1 Þ=B−1 ¼ β0 þ β1 %div0 þ β2 %div0  SInfo þ β3 %div0  BInfo þ ϖτ

ð3Þ

ðEτ −Eτ−1 Þ=B−1 ¼ β0 þ β1 %div0 þ β2 %div0  SInfo þ β3 %div0  BInfo þ β4 SD  SInfo þ β5 SD  BInfo þ ψτ

ð4Þ

and ðEτ −Eτ−1 Þ=B−1 ¼ β0 þ β1 SD þ β2 SD  SInfo þ β3 SD  BInfo þ ξτ ;

ð5Þ

for τ = 1 and 2, where SD takes the value of 1 (−1) for unexpected dividend increases (decreases), and SInfo (BInfo) is a dummy variable that takes the value of 1 when the abnormal stock (bond) return on the announcement date is significantly positive for dividend increases or significantly negative for dividend decreases. 8 The standardized stock and premium bond returns both have a student-t distribution with a mean of 0 and degrees of freedom equal to ns − 2 and nb − 2 respectively, where ns and nb are the number of days with stock and bond returns during the period from t = −60 to t = −16. 9 Following Nissim and Ziv (2001) and Watts (1973), we assign each dividend announcement to year τ if it is declared in the second, third, or fourth physical quarter of that year. Dividends declared in the first physical quarter of year τ + 1 is also assigned to year τ.

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

11

Table 10 Earnings predictability of dividend changes: full sample. This table reports results of the following regressions: ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ ετ ; ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 SD  SInfo þ β 3 SD  BInfo þ ωτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 %div0  SInfo þ β 3 %div0  BInfo þ ϖτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 %div0  SInfo þ β 3 %div0  BInfo þ β4 SD  SInfo þ β 5 SD  BInfo þ ψτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 SD þ β 2 SD  SInfo þ β 3 SD  BInfo þ ξτ for τ = 1 and 2, where τ = 0 is the year of dividend announcement, %div0 denotes the percentage changes in dividends, SD takes the value of 1 (−1) for unexpected dividend increases (decreases), and SInfo (BInfo) is a dummy variable that takes the value of 1 when the abnormal stock (bond) return on the announcement date is significantly positive for dividend increases or significantly negative for dividend decreases. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively. All t-statistics are computed using White heteroscedastic-consistent variance estimates and are shown in the parentheses.

Intercept %div

(1) τ = 1

(2) τ = 2

(3) τ = 1

(4) τ = 2

(5) τ = 1

(6) τ = 2

(7) τ = 1

(8) τ = 2

(9) τ = 1

(10) τ = 2

0.0061*** (4.60) 0.0004 (0.11)

0.0088*** (4.90) −0.0037 (−0.46)

0.0059*** (4.35) −0.0005 (−0.14)

0.0088*** (4.89) −0.0035 (−0.43)

0.0062*** (4.61) −0.0004 (−0.08)

0.0088*** (4.96) −0.0016 (−0.20)

0.0059*** (4.35) −0.0002 (−0.05)

0.0086*** (4.80) −0.0015 (−0.19)

0.0055*** (3.88)

0.0089*** (4.66)

0.0027 (0.82)

−0.0014 (−0.35)

SD −0.0012 (−0.25) 0.0031 (0.66) −0.0125** (−2.06) 0.0062 (0.50) 3936

4624

3936

−0.0032 (−0.65) 0.0009 (0.19) 0.0155** (2.35) 0.0117*** (2.62) 4624

0.0003

0.0000

0.0009

0.0006

%div *SInfo %div *BInfo SD *SInfo

4628

3940

0.0143** (2.29) 0.0121*** (2.70) 4624

0.0000

0.0001

0.0006

SD *BInfo Sample size R square

−0.0305** (−2.06) −0.0808 (−1.18)

−0.0160 (−1.27) −0.1369 (−1.52) −0.0104 (−1.57) 0.0301* (1.84) 3936

0.0119* (1.79) 0.0098* (1.89) 4624

−0.0120* (−1.78) 0.0068 (0.53) 3936

0.0090

0.0007

0.0002

We firstly run regressions based on all dividend announcements including events without dividend changes and the results are provided in Table 10. For comparison purposes, we also conduct an analysis solely based on announcements with unexpected dividend changes (i.e., positive and negative dividend announcements and exclude those without dividend changes) and obtain similar conclusions. The regression results are reported in Appendix Table A.10 In the analysis, we compute test statistics using White heteroscedastic-consistent variance estimates and they are provided within the parentheses underneath the estimated coefficients. Columns 1 and 2 in Table 10 show the results when we do not differentiate dividend announcements. The coefficient of %div0 is equal to 0.0004 and −0.0037 for τ = 1 and 2, respectively, and is insignificant in either case. Consistent with Benartzi et al. (1997) and Nissim and Ziv (2001), the coefficient of %div0 is not significant, suggesting that unexpected dividend changes are not indicative of future profits. Columns 3 through 10 in Table 10 report the regression results when we consider the fact that not all announcements result in significant market reactions. While the coefficient of the %div0 variable alone is not significant, the coefficients of SD × SInfo and SD × BInfo are significantly positive in the regression when τ = 1. In addition, the coefficients of SD × SInfo are higher in value than those of SD × BInfo when τ = 1. For instance, the estimated coefficients in Eq. (4) reported in Column (7) suggest that on average the profits (relative to book value of assets) will be improved by 1.55 (1.17) percent after one year of positive dividend announcements if the announcements result in significantly abnormal stock (bond) returns. Interestingly, the coefficients of %div0 × SInfo and %div0 × BInfo are in general not statistically significant. Thus, although unexpected dividend increases (decreases) are more likely to be followed by better (worse) earnings after one year of announcements if the announcements are accompanied by significantly positive (negative) abnormal stock or bond returns, future profitability are not significantly related to the magnitude of dividend changes. The relationship between dividend changes and future profitability becomes much weaker after two years of announcements. Interestingly, for τ = 2, the coefficient of SD × SInfo has a negative sign and is statistically significant in two out of three cases, suggesting a reversion in earnings after two years of announcements. We then perform a similar analysis separately for the investment and speculative groups, where the rating is based on actively traded bonds and report respective results in Panels A and B of Table 11. Again, we also conduct an analysis solely based on events with unexpected dividend changes. The patterns are similar to what we find when all events are included and the results are provided 10

We also conduct an analysis when BInfo is identified using all bond trades and reach similar conclusions. The regression results are available upon request.

12

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

Table 11 Earnings predictability of dividend changes: investment vs. speculative. This table reports results of the following regressions for the investment and speculative groups, respectively ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ ετ ; ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 SD  SInfo þ β 3 SD  BInfo þ ωτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 %div0  SInfo þ β 3 %div0  BInfo þ ϖτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 %div0  SInfo þ β 3 %div0  BInfo þ β4 SD  SInfo þ β 5 SD  BInfo þ ψτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 SD þ β 2 SD  SInfo þ β 3 SD  BInfo þ ξτ ; for τ = 1 and 2, where τ = 0 is the year of dividend announcement, %div0 denotes the percentage changes in dividends, SD takes the value of 1 (−1) for unexpected dividend increases (decreases), and SInfo (BInfo) is a dummy variable that takes the value of 1 when the abnormal stock (bond) return on the announcement date is significantly positive for dividend increases or significantly negative for dividend decreases. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively. All t-statistics are computed using White heteroscedastic-consistent variance estimates and are shown in the parentheses. (1) τ = 1 Panel A: Investment-grade Intercept 0.0075*** (8.15) %div 0.0018 (0.35) SD

(2) τ = 2

(3) τ = 1

(4) τ = 2

(5) τ = 1

(6) τ = 2

(7) τ = 1

(8) τ = 2

(9) τ = 1

(10) τ = 2

0.0079*** (6.66) −0.0082 (−0.71)

0.0073*** (7.83) 0.0010 (0.20)

0.0079*** (6.71) −0.0081 (−0.70)

0.0075*** (8.10) 0.0020 (0.32)

0.0080*** (6.84) −0.0056 (−0.52)

0.0073*** (7.74) 0.0021 (0.34)

0.0078*** (6.66) −0.0055 (−0.51)

0.0065*** (6.40)

0.0084*** (6.88)

0.0050** (2.38)

−0.0045* (−1.72)

−0.0016 (−0.24) 0.0038 (0.18) −0.0105 (−1.58) 0.0071 (0.47) 2940

3432

2940

−0.0058 (−0.90) −0.0151 (−0.73) 0.0169** (2.41) 0.0090** (2.09) 3432

%div *SInfo %div *BInfo SD *SInfo

0.0116* (1.72) 0.0028 (0.63) 3432

−0.0081 (−1.20) 0.0095 (0.61) 2940

3435

2943

0.0001

0.0011

0.0016

0.0016

0.0001

0.0042

0.0019

0.0072

0.0033

0.0016

0.0167*** (2.71) 0.0036 (0.25)

−0.0022 (0.41) −0.0024 (−0.59)

0.0170*** (2.73) 0.0041 (0.27)

−0.0019 (−0.36) −0.0041 (−0.56)

0.0168*** (2.72) 0.0045 (0.30)

−0.0021 (−0.40) −0.0040 (−0.55)

0.0168*** (2.67) 0.0045 (0.30)

−0.0011 (−0.22)

0.0159*** (2.66)

−0.0163** (−0.77)

0.0157 (0.65)

Panel B: Speculative-grade Intercept −0.0020 (−0.38) %div −0.0007 (−0.18) SD %div*SInfo

0.0076 (0.90) 0.0073 (0.99)

SD*SInfo

983

819

0.0129 (0.76) 0.0341** (2.26) 982

0.0000

0.0000

0.0006

SD*BInfo

−0.0530* (−1.73) −0.0194 (−0.52)

−0.0221 (−1.46) 0.0005 (0.04) 818

982

818

0.0027 (0.29) 0.0038 (0.51) 0.0121 (0.59) 0.0316** (2.07) 982

0.0002

0.0002

0.0002

0.0006

%div*BInfo

Sample size R square

−0.0070 (−0.56) −0.1490 (−1.36) −0.0107 (−1.53) 0.0334* (1.68) 2940

0.0154** (2.25) 0.0067 (1.59) 3432

SD *BInfo Sample size R square

−0.0209 (−1.39) −0.0893 (−1.08)

−0.0331 (−0.82) −0.0370 (−0.72) −0.0112 (−0.59) 0.0080 (0.42) 818

0.0268 (1.00) 0.0472* (1.87) 982

−0.0359 (−1.29) −0.0137 (−0.50) 818

0.0002

0.0016

0.0010

in Appendix Table B. For the investment group, the coefficient of %div0 is not significant, implying that the magnitude of dividend changes is not directly related to future profitability. The coefficients of SD × SInfo are significantly positive when τ = 1 in all regressions. Unlike the results shown in Table 10, the coefficients of SD × BInfo are not significant for the investment group, suggesting a weaker relationship between abnormal bond returns and future profitability. This pattern is not surprising since investment-grade bonds have low default risk, and their returns are more related to general interest rates than to the value of the underlying company (see, e.g., Kwan, 1996). On the other hand, stock returns are mainly driven by firm values. If stockholders perceive the unexpected dividend changes containing information about future performance, we would anticipate seeing a positive relationship between abnormal stock returns and future profitability. When τ = 2, similar to what we find previously, the coefficients of SD × SInfo are negative, suggesting some reversion in earnings. However, for the investment group, the evidence is not statistically significant. Table 11 Panel B shows the results for the speculative group. Because speculative-grade bonds have higher default risk, their returns are more driven by firm value than by interest rates. This is also supported by the evidence in previous studies that the

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

13

sensitivity of bond returns to equity returns increases when the credit ratings of corporate bonds decline (see, e.g., Bao and Hou, 2013). Consistently, the coefficient of SD × BInfo is significantly positive when τ = 1, implying that the reaction of speculativegrade bonds on dividend announcements is informative about firm profitability one year after announcements. The coefficient of SD × SInfo is positive, but not significant in the regression. This may be attributed to the smaller sample size in the speculative group. Similar to what we find in the investment group, the relationship between dividend changes and future profitability becomes very weak after two years of dividend announcements. In sum, we find a weak or no relationship between the magnitude of dividend changes and future profitability. When considering those announcements that result in significant market reactions, we find that unexpected dividend increases (decreases) are more likely to be followed by better (worse) earnings one year after announcements if the announcements are accompanied by significantly positive (negative) abnormal stock or bond returns. Additionally, we find that the reaction of speculative-grade bonds on announcement dates is more informative about future earnings than that of investment-grade bonds. The relationship between dividend changes and firm profitability becomes very weak after two years of announcements.

5. Conclusions We use comprehensive transaction data from Trade Reporting and Compliance Engine (TRACE) to study the response in corporate bond market to dividend announcements and compare that with the response in stock market. We find that the information content/ free cash flow effect dominates the wealth transfer effect; the mean excess premium bond returns on announcement dates are positively related to unexpected dividend changes. As for stock, consistent with previous studies, stock returns are positively correlated with unexpected dividend changes. We investigate the relationship between dividend changes and future profitability while considering that not all dividend announcements result in significant market reactions. We show that the magnitude of dividend changes is weakly related to future profitability. However, when the announcements are accompanied by significant market responses, unexpected dividend increases (decreases) are followed by better (worse) earnings one year after the announcements. In addition, the reaction of speculative-grade bonds is more informative about future profitability than that of investment-grade bonds. After two years of announcements, however, the relationship between dividend changes and future profitability becomes much weaker. Table A Earnings predictability of dividend changes: excluding events with no dividend changes. This table reports results of the following regressions based on events with unexpected dividend changes, excluding events with no dividend changes: ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ ετ ; ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 SD  SInfo þ β 3 SD  BInfo þ ωτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 %div0  SInfo þ β 3 %div0  BInfo þ ϖτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 %div0  SInfo þ β 3 %div0  BInfo þ β4 SD  SInfo þ β 5 SD  BInfo þ ψτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 SD þ β 2 SD  SInfo þ β 3 SD  BInfo þ ξτ ; for τ = 1 and 2, where τ = 0 is the year of dividend announcement, %div0 denotes the percentage changes in dividends, SD takes the value of 1 (−1) for unexpected dividend increases (decreases), and SInfo (BInfo) is a dummy variable that takes the value of 1 when the abnormal stock (bond) return on the announcement date is significantly positive for dividend increases or significantly negative for dividend decreases. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively. All t-statistics are computed using White heteroscedastic-consistent variance estimates and are shown in the parentheses.

Intercept %div

(1) τ = 1

(2) τ = 2

(3) τ = 1

(4) τ = 2

(5) τ = 1

(6) τ = 2

(7) τ = 1

(8) τ = 2

(9) τ = 1

(10) τ = 2

0.0083*** (3.05) −0.0003 (−0.10)

0.0091*** (2.84) −0.0039 (−0.49)

0.0072** (2.48) −0.0008 (−0.26)

0.0093*** (2.78) −0.0038 (−0.47)

0.0084*** (3.11) −0.0012 (−0.26)

0.0095*** (2.97) −0.0019 (−0.25)

0.0072** (2.50) −0.0007 (−0.16)

0.0086** (2.56) −0.0014 (−0.19)

0.0046 (1.50)

0.0109*** (3.32)

0.0033 (1.01)

−0.0030 (−0.84)

SD

−0.0128** (−2.01) 0.0058 (0.46) 878

1030

878

−0.0027 (−0.58) 0.0014 (0.30) 0.0146** (2.14) 0.0107** (2.17) 1030

0.0015

0.0002

0.0049

0.0024

%div *SInfo

0.0016 (0.32) 0.0036 (0.78)

%div *BInfo SD *SInfo

1030

878

0.0135** (2.07) 0.0112** (2.25) 1030

0.0000

0.0005

0.0023

SD *BInfo Sample size R square

−0.0307** (−2.04) −0.0813 (−1.18)

−0.0160 (−1.28) −0.1369 (−1.52) −0.0103 (−1.50) 0.0301* (1.81) 878

0.0119* (1.78) 0.0098* (1.89) 1030

−0.0119* (−1.76) 0.0067 (0.53) 878

0.0087

0.0028

0.0014

14

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

Table B Earnings predictability of dividend changes: investment vs. speculative. This table reports results of the following regressions based on events with unexpected dividend changes, excluding events with no dividend changes. Panels A and B show the results for the investment and speculative groups, respectively. ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ ετ ; ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 SD  SInfo þ β 3 SD  BInfo þ ωτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 %div0  SInfo þ β 3 %div0  BInfo þ ϖτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 %div0 þ β 2 %div0  SInfo þ β 3 %div0  BInfo þ β4 SD  SInfo þ β 5 SD  BInfo þ ψτ ðEτ −Eτ−1 Þ=B−1 ¼ β 0 þ β1 SD þ β 2 SD  SInfo þ β 3 SD  BInfo þ ξτ ; for τ = 1 and 2, where τ = 0 is the year of dividend announcement, %div0 denotes the percentage changes in dividends, SD takes the value of 1 (−1) for unexpected dividend increases (decreases), and SInfo (BInfo) is a dummy variable that takes the value of 1 when the abnormal stock (bond) return on the announcement date is significantly positive for dividend increases or significantly negative for dividend decreases. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively. All t-statistics are computed using White heteroscedastic-consistent variance estimates and are shown in the parentheses. (1) τ = 1 Panel A: Investment-grade Intercept 0.0105*** (5.62) %div 0.0006 (0.11) SD

(2) τ = 2

(3) τ = 1

(4) τ = 2

(5) τ = 1

(6) τ = 2

(7) τ = 1

(8) τ = 2

(9) τ = 1

(10) τ = 2

0.0071*** (2.64) −0.0077 (−0.65)

0.0097*** (5.00) 0.0002 (0.03)

0.0071*** (2.69) −0.0076 (−0.64)

0.0105*** (5.52) 0.0008 (0.12)

0.0077*** (2.96) −0.0054 (−0.49)

0.0095*** (4.78) 0.0012 (0.19)

0.0067** (2.57) −0.0048 (−0.44)

0.0052 (1.64)

0.0111*** (3.27)

0.0060* (1.84)

−0.0067* (−1.89)

−0.0010 (−0.15) 0.0001 (0.00) −0.0100 (−1.51) 0.0076 (0.50) 730

856

730

−0.0050 (−0.76) −0.0142 (−0.68) 0.0153** (2.14) 0.0072 (1.58) 856

%div *SInfo %div *BInfo SD *SInfo

0.0116* (1.71) 0.0029 (0.65) 856

−0.0080 (−1.18) 0.0093 (0.60) 730

856

730

0.0001

0.0033

0.0047

0.0054

0.0001

0.0148

0.0055

0.0263

0.0096

0.0057

0.0226 (1.13) 0.0020 (0.16)

−0.0107 (−0.56) −0.0011 (−0.42)

0.0232 (1.10) 0.0025 (0.19)

−0.0084 (−0.47) −0.0024 (−0.50)

0.0220 (1.09) 0.0030 (0.23)

−0.0105 (−0.55) −0.0019 (−0.40)

0.0227 (0.97) 0.0027 (0.22)

−0.0026 (−0.21)

0.0143 (1.05)

−0.0153 (−0.97)

0.0167 (0.95)

Panel B: Speculative-grade Intercept −0.0089 (−0.49) %div 0.0006 (0.25) SD %div *SInfo

0.0076 (0.90) 0.0062 (1.12)

SD *SInfo

137

116

0.0174 (0.83) 0.0369** (2.14) 137

0.0000

0.0001

0.0030

SD *BInfo

−0.0493 (−1.56) −0.0152 (−0.42)

−0.0242 (−1.31) −0.0030 (−0.16) 116

137

116

0.0005 (0.07) 0.0020 (0.38) 0.0175 (0.72) 0.0354** (2.04) 137

0.0008

0.0005

0.0006

0.0030

%div *BInfo

Sample size R square

−0.0076 (−0.61) −0.1497 (−1.37) −0.0099 (−1.38) 0.0343* (1.71) 730

0.0140** (2.03) 0.0049 (1.13) 856

SD *BInfo Sample size R square

−0.0208 (−1.41) −0.0890 (−1.07)

−0.0143 (−0.19) −0.0157 (−0.18) −0.0194 (−0.52) 0.0005 (0.01) 116

0.0268 (0.99) 0.0470* (1.93) 137

−0.0363 (−1.38) −0.0138 (−0.50) 116

0.0008

0.0058

0.0037

Appendix A

References Aharony, Joseph, Swary, Itzhak, 1980. Quarterly dividend and earnings announcements and stockholders' returns: an empirical analysis. J. Financ. 35, 1–12. Asquith, Paul, Mullins Jr., David W., 1983. The impact of initiating dividend payments on shareholders' wealth. J. Bus. 56, 77–96. Asquith, Paul, Wizman, Thierry A., 1990. Event risk, covenants, and bondholder returns in leveraged buyouts. J. Financ. Econ. 27, 195–213. Bao, Jack, Hou, Kewei, 2013. Comovement of corporate bonds and equities. Working Paper. Bao, Jack, Pan, Jun, 2013. Bond illiquidity and excess volatility. Rev. Financ. Stud. 26, 3068–3103.

H.-J. Tsai, Y. Wu / Journal of Empirical Finance 30 (2015) 1–15

15

Bao, Jack, Pan, Jun, Wang, Jiang, 2011. The illiquidity of corporate bonds. J. Financ. 66, 911–946. Benartzi, Shlomo, Michaely, Roni, Thaler, Richard, 1997. Do changes in dividends signal the future or the past? J. Financ. 52, 1007–1034. Bessembinder, Hendrik, Kahle, Kathleen, Maxwell, William, Danielle, Xu., 2009. Measuring abnormal bond performance. Rev. Financ. Stud. 22, 4219–4258. Bhattacharya, Sudipto, 1979. Imperfect information, dividend policy and ‘the bird in the hand fallacy’. Bell J. Econ. 10, 259–270. Dann, Larry, 1981. Common stock repurchases: an analysis of returns to bondholders and stockholders. J. Financ. Econ. 9, 113–138. DeAngelo, Harry, DeAngelo, Linda, Skinner, Douglas J., 1996. Reversal of fortune: dividend signaling and the disappearance of sustained earnings growth. J. Financ. Econ. 40, 341–371. DeFusco, Richard, Johnson, Robert, Zorn, Thomas, 1990. The effect of executive stock option plans on stockholders and bondholders. J. Financ. 45, 617–627. Dhillon, Upinder, Johnson, Herb, 1994. The effect of dividend changes on stock and bond prices. J. Financ. 39, 281–289. Downing, Chris, Underwood, Shane, Xing, Yuhang, 2009. An intraday analysis of the relative informational efficiency of stocks and bonds. J. Financ. Quant. Anal. 44, 1081–1102. Eades, Kenneth, 1982. Empirical evidence on dividends as a signal of firm value. J. Financ. Quant. Anal. 17, 471–500. Grullon, Gustavo, Michaely, Roni, Benartzi, Shlomo, Thaler, Richard, 2005. Dividend changes do not signal changes in future profitability. J. Bus. 78, 1659–1682. Hand, John, Holthausen, Robert, Leftwich, Richard, 1992. The effect of bond rating agency announcements on bond and stock prices. J. Financ. 47, 733–752. Handjinicolaou, George, Kalay, Avner, 1984. Wealth redistributions or changes in firm value: an analysis of returns to bondholders and stockholders around dividend announcements. J. Financ. Econ. 13, 35–63. Holthausen, Robert, Leftwich, Richard, 1986. The effect of common bond rating changes on common stock prices. J. Financ. Econ. 17, 57–89. Hotchkiss, Edith, Ronen, Tavy, 2002. The informational efficiency of the corporate bond market: an intraday analysis. Rev. Financ. Stud. 15, 1325–1354. Huang, Jingzhi, Shi, Zhan, 2013. A resolution to the equity-credit market integration puzzle. Working Paper. Jayaraman, Narayanan, Shastri, Kuldeep, 1988. The valuation impact of specially designated dividends. J. Financ. Quant. Anal. 23, 301–312. Jensen, Michael, 1986. The agency costs of free cash flow: corporate finance and takeovers. Am. Econ. Rev. 76, 323–329. Jensen, Michael, Meckling, William H., 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. J. Financ. Econ. 3, 305–360. Kalay, Avner, 1980. Signalling, information content, and the reluctance to cut dividends. J. Financ. Quant. Anal. 15, 855–869. Kalay, Avner, Loewenstein, Uri, 1985. Predictable events and excess returns: the case of dividend announcements. J. Financ. Econ. 14, 423–449. Kwan, Clarence C.Y., 1981. Efficient market tests of the informational content of dividend announcements: critique and extensions. J. Financ. Quant. Anal. 16, 193–206. Kwan, Simon, 1996. Firm-specific information and the correlation between individual stocks and bonds. J. Financ. Econ. 40, 63–80. Maxwell, William, Rao, Ramesh, 2003. Do spin-offs expropriate wealth from bondholders? J. Financ. 58, 2087–2108. Maxwell, William, Stephens, Clifford, 2003. The wealth effects of repurchases on bondholders. J. Financ. 58, 895–920. Merton, Robert C., 1974. On the pricing of corporate debt: the risk structure of interest rates. J. Financ. 29, 449–470. Miller, Merton, Modigliani, Franco, 1961. Dividend policy, growth and the valuation of shares. J. Bus. 34, 411–433. Miller, Merton, Rock, Kevin, 1985. Dividend policy under asymmetric information. J. Financ. 40, 1031–1051. Nissim, Doron, Ziv, Amir, 2001. Dividend changes and future profitability. J. Financ. 56, 2111–2133. Penman, Stephen H., 1983. The predictive content of earnings forecasts and dividends. J. Financ. 38, 1181–1199. Petit, Richardson R., 1972. Dividend announcements, security performance, and capital market efficiency. J. Financ. 27, 993–1007. Ronen, Tavy, Zhou, Xing, 2013. Trade and information in the corporate bond market. J. Financ. Mark. 16, 61–103. Schaefer, Stephen, Strebulaev, Ilya A., 2008. Structural models of credit risk are useful: evidence from hedge ratios on corporate bonds. J. Financ. Econ. 90, 1–19. Wakeman, L. Macdonald, 1978. Bond rating agencies and the capital markets. Working Paper. Watts, Ross, 1973. The information content of dividends. J. Bus. 46, 191–211. Weinstein, Mark, 1977. The effect of a rating change announcement on bond price. J. Financ. Econ. 5, 329–350. Woolridge, J. Randall, 1983. Dividend changes and security prices. J. Financ. 38, 1607–1615. Yoon, Pyung Sig, Starks, Laura, 1995. Signaling, investment opportunities, and dividend announcements, 1995. Rev. Financ. Stud. 8, 995–1018.