Vous êtes sur la page 1sur 33

Are Future Earnings Related to Past Earnings Performance and Dividend Changes?

Ebenezer Asem Faculty of Management University of Lethbridge Lethbridge, AB T1K 3M4 Canada Tel. 403.382.7142 ebenezer.asem@uleth.ca Aditya Kaul School of Business University of Alberta Edmonton, AB T6G 2R6 Canada Tel. 780.492.5027 akaul@ualberta.ca

Abstract Recent research suggests that market reactions to dividend changes are consistent with the hypothesis that dividend changes are managerial signals about the permanence of prior earnings changes. This view posits that increases in past earnings should be more permanent (transitory) for the firms that subsequently increase (reduce) their dividends than those that maintain their dividends. Likewise, declines in earnings should be more permanent (transitory) for the firms that subsequently decrease (increase) their dividends. Our evidence on the behavior of future earnings supports these predictions, except that increases in earnings are not more transitory for firms that reduce dividend. Overall, we find support for the permanence hypothesis. JEL classification: G30; G35 Keywords: Dividends payout; Earnings signaling theory; Event studies. Date: Feb., 2010.

We thank Gloria Tian, Vikas Mehrotra, and seminar participants at the University of Lethbridge for their very helpful comments. Maryna Skrynska provided useful research assistance. Ebenezer gratefully acknowledge the financial support from University of Lethbridge Research Fund. Any remaining mistakes are ours alone.

1
Electronic copy available at: http://ssrn.com/abstract=1782122

Are Future Earnings Related to Past Earnings Changes and Dividend Changes?

Evidence on the view that dividend changes are directly related to future earnings growth is mixed.1 More recent empirical studies suggest that dividend changes are actually managerial signals about the permanence of earnings changes that have already occurred; an idea alluded to by Lintner (1956) in his survey (for convenience, we refer to this as the permanence hypothesis). DeAngelo et al. (1992) document more persistent losses for firms that reduce dividends than those that maintain their dividends. Benartzi et al. (1997) report that increases in earnings are more permanent for the firms that subsequently increase their dividends than those that maintain theirs. Related studies investigate whether investors view dividend-change news as managerial private information about the permanence of prior earnings changes. Kane, Lee and Marcus (1984) find that investors give more credence to dividend changes that are in the same direction as prior earnings changes (corroboratory dividend changes). Koch and Sun (2004) reason that, if investors view dividend changes as signals about the permanence of prior earnings changes, dividend-change announcement returns should be related to the magnitude of the past earnings changes. For instance, a dividend increase that signals that a large increase in earnings is permanent should induce a larger positive market reaction than a dividend increase that signals that a small increase in earnings is permanent. Similar arguments apply to dividend reductions after earnings declines, suggesting confirmatory dividend-change announcement returns should be directly related to the magnitude of the past earnings changes.
Some of the studies that support this view include Brickley (1983), Healy and Palepu (1988), DeAngelo, DeAngelo and Skinner (1992), Aharony and Dotan (1994), and Nissim and Ziv (2001), while those that refute it include Watts (1973), Gonedes (1978), Penman (1983), Benartzi et al. (1997), and Grullon et al. (2005).
1

2
Electronic copy available at: http://ssrn.com/abstract=1782122

In contrast, contradictory dividend-change announcement returns should be inversely related to the magnitude of the prior earnings changes. Koch and Sun (2004) find support for these predictions. If dividend changes are managerial private information about the persistence of prior earnings changes, as the market seems to suggest, we can predict the future earnings patterns contingent on the sign of the prior earnings changes and dividend changes. We develop and test these future earnings implications of the permanence hypothesis, and our results support it. Very few studies condition the relation between dividend changes and changes in future earnings on past earnings performance. DeAngelo et al. (1992) and DeAngelo et al. (1996) examine NYSE-listed firms that sustain losses after long periods of continuous positive net income and Benartzi et al. (1997) focus on firms whose dividend increases confirm their past earnings increases. Our paper extends Benartzi et al. (1997) to include the complete menu presented by the permanence hypothesis: (i) earnings increases followed by dividend increases or reductions and (ii) earnings decreases followed by dividend increases or reductions.2 This has become important in light of Koch and Suns (2004) finding that market reactions to dividend-changing news are consistent with the view that dividend changes (not only confirmatory dividend increases) are managerial private information about the permanence of prior earnings changes. Thus, we test whether future earnings patterns are consistent with the permanence hypothesis as well as its supporting evidence from the markets reaction to dividend changes. Benartzi et al. (1997) suggest that dividend increases that follow earnings increases are managerial private signals that the higher earnings are permanent. If one extrapolates this to
While a limited number of data points for some of the cases made testing all four possible scenarios problematic in the past, we believe that current data are sufficient to test all four cases. This will allow a more robust conclusion to be reached on the permanence hypothesis.
2

3
Electronic copy available at: http://ssrn.com/abstract=1782122

dividend reductions that follow earnings increases, then these dividend reductions are signals that the higher earnings are not permanent. Consequently, after an increase in earnings, a dividend increase is a managerial signal that the higher earnings are permanent, while a dividend reduction suggests the higher earnings are not permanent, and an unchanged dividend conveys no information regarding earnings permanence. Accordingly, higher earnings should be more (less) permanent for the firms that subsequently increase (decrease) their dividends than their counterparts that maintain their dividends. Turning to firms with declines in prior earnings, the permanence hypothesis suggests that dividend reductions are managerial signals that the lower earnings are permanent. If dividend reductions after earnings declines are signals that the lower earnings are permanent, then dividend increases after declines in earnings are managerial private information that the lower earnings are temporary. As a result, lower earnings should be more (less) permanent for the firms that reduce (increase) their dividends than those that maintain theirs. Thus, the permanence hypothesis indicates that the relation between dividend changes and future earnings is governed by the sign of the prior earnings change. For instance, a dividend increase after an increase in earnings signals that the higher earnings is permanent, while the same dividend increase after a decline in earnings signals that the lower earnings is temporary, indicating growth in future earnings. To test these hypotheses, we classify all dividend payers by the sign of their earnings changes and by their dividend changes, and study the behavior of future earnings these groups. Our results show that confirmatory dividend changes are signals that the prior earnings changes are permanent. In particular, as in Benartzi et al. (1997), we find that increases in prior earnings are less likely to reverse for firms that subsequently increase dividends than those that

maintain their dividends. In addition, we find that declines in prior earnings persist more for the firms that subsequently reduce their dividends than those that leave their dividends unchanged.3 Turning to firms with contradictory dividend changes, our results show that declines in earnings tend to reverse more among firms that subsequently increase their dividends than those that maintain theirs as the permanence hypothesis predicts. This suggests that managers do not only increase dividends based on increases in past earnings, they also increase dividends in anticipation of future earnings increases, especially when their expectation is contradicted by the current earnings performance. This is consistent with the models of Bhattacharya (1979) and Miller and Rock (1985), which suggest that managers change dividends to convey information about future earnings in the presence of information asymmetry. In contrast to the permanence hypothesis, we do not find that increases in earnings reverse more firms that subsequently reduce dividends than those that maintain their dividends. It is well-established that managers dislike dividend reductions due to the associated negative market reactions. Given the reluctance to reduce dividends and the fact that prior earnings increased, it does not appear that contradictory dividend reductions are motivated expectations that the higher earnings will reverse. Thus, unlike contradictory dividend increases, where it is conceivable that anticipation of reversals in declined earnings can drive increases in dividends, it difficult to understand why dividend-reduction-averse managers would reduce dividends in anticipation of reversals in increases in current earnings. The relatively strong performance of the firms with contradictory dividend reductions in the prior and subsequent quarters led us to study the capital expenditure patterns of our sample

This is in line with DeAngelo et al. (1992) finding that losses are more permanent for firms that reduce their dividends than those that maintain theirs.

firms. The results show that firms with contradictory dividend increases increase their capital expenditures the most. This suggests that dividend reductions following increases in earnings are, at least partly, motivated by cash needs to boost investment. Thus, with the exception of contradictory dividend increases, the evidence from future earnings behavior is exactly as the permanence hypothesis predicts. Overall, therefore, we find support the hypothesis that dividend changes are managerial signals about the persistence in past earnings changes. Our results make two important contributions to the empirical literature on the information content of dividend changes. First, our paper extends Benartzis et al. (1997) findings that confirmatory dividend increases are signals that the higher earnings will not reverse in future. Specifically, we find that confirmatory dividend reductions are signals that the lower earnings will not reverse, and contradictory dividend increases are signals that the lower earnings will reverse. Second, our result shows that dividend changes do not only help investors update their assessments of the persistence in past earnings changes (Koch and Sun (2004)), but they also predict future earnings patterns conditional on the sign of the prior earnings change. That is, the future earnings patterns are consistent with investors interpretations of dividend changes given the sign of the prior earnings change. The rest of the paper is organized as follows: Section I summaries the related literature and develops the testable hypotheses; Section II presents the data and initial evidence; Section III discusses the results; Section IV explores the robustness; and Section V concludes.

I. Related Literature and Testable Hypotheses Lintners (1956) influential study of dividend policy suggests that firms increase dividends when management believes that earnings have permanently increased. Subsequently, Miller and Modigliani (1961), Bhattacharya (1979), Miller and Rock (1985), and John and Williams (1985) theorize that dividend changes can also convey information about future cash flow when markets are incomplete or when there is information asymmetry. These models point to a direct relation between dividend changes and future earnings. To empirically investigate this prediction, Watts (1973) regresses next years earnings on this years dividend changes and obtains positive but insignificant coefficient estimates. Subsequent studies by Gonedes (1978), Penman (1983), Jensen et al. (2010) report similar results. Also, Healy and Palepu (1988) report that the earnings of firms that omit dividends do not decline in the following years, but rather increase. These studies, therefore, do not support the notion that dividend changes are directly related to future earnings changes. In line with a direct relation, Healy and Palepu (1988) report that earnings of dividend initiators increase in the two years after the initiations. Brickley (1983) also report the earnings increase after large dividend increases (of more than 20%). Aharony and Dotan (1994) document earnings increases in the four quarters after dividend increases. After using several different measures of earnings, Benartzi et al. (1997) they find that dividend changes are not directly related to future earnings. Nissim and Ziv (2001) counter that dividend changes are directly related to future earnings changes after controlling for linear mean revisions in earnings. Grullon et al. (2005) suggest that dividend changes are not directly linked to future earnings changes after adjusting for nonlinear patterns in earnings. Collectively, the evidence in support of a direct link between dividend changes and future earnings changes is weak.

Not finding a direct relation between dividend changes and future earnings changes, Benartzi et al. (1997) test Lintners (1956) finding that dividend increases signal permanence of past earnings increases. Specifically, they analyze firms with similar past earnings increases and find that the higher earnings are less likely to reverse for those that increase their subsequent dividends than those that maintain theirs.4 Thus, the authors conclude that corroboratory dividend increases are signals that the increases in earnings are permanent. Koch and Sun (2004) study whether the market views dividend changes as managerial information about the permanence of prior earnings changes. They argue that, if the market interprets dividend changes as such, corroboratory dividend-change announcement returns will be directly related to prior earnings changes while contradictory dividend-change announcement returns will be inversely related to prior earnings changes. With the exception of contradictory dividend increases, Koch and Sun find that investors interpret dividend changes as providing managerial information about the permanence of earnings changes that have already occurred. Are these investors interpretations correct in the face of future earnings changes? We study the evolution of earnings contingent of past earnings performance and dividend changes to address this question. Both the permanence hypothesis and its supporting evidence from the markets reactions to dividend changes (Koch and Sun (2004)) have clear implications for future earnings pattern. Specifically, they both suggest that increases in earnings should exhibit more permanence among firms that subsequently increase dividends but exhibit more reversal among those that subsequently reduce their dividends compared to the firms that leave their dividends
Other studies that condition the relation between dividend changes and future earnings on past performance include DeAngelo et al. (1992) who studied NYSE-listed dividend payers that reported positive earnings in each of the ten years prior to annual losses and find that the losses tend to be more permanent for those that decrease their dividends than those that maintain theirs, while DeAngelo et al. (1996) find that dividend increases by such firms do not convey any information about the permanence of their current earnings problems.
4

unchanged. Likewise, following declines in earnings, firms that reduce (increase) dividends should display more persistence (reversal) in their lower earnings than those that maintain their dividends. Thus, if Y0,T denotes the change in earnings from the current period to future period T, then following increases in earnings, Y0,T should on average be higher for the firms that subsequently increase dividends than those that maintain their dividends, and these firms should, in turn, have higher average Y0,T than those that reduce their dividends. Similarly, after declines in earnings, Y0,T should be higher (lower) on average for the firms that subsequently increase (reduce) their dividends than those that maintain their dividends. Accordingly, we test the following hypotheses: (H1): Following earnings increases, Y0,T is on average higher for the firms that increase their subsequent dividends than those that leave their dividends unchanged. (H2): Following earnings increases, Y0,T is on average lower for the firms that reduce their subsequent dividends than those that leave their dividends unchanged. (H3): Following earnings increases, Y0,T is on average higher for the firms that increase their subsequent dividends than those that reduce their dividends. (H4): Following earnings declines, Y0,T is on average higher for the firms that increase their subsequent dividends than those that leave their dividends unchanged. (H5): Following earnings declines, Y0,T is on average lower for the firms that reduce their subsequent dividends than those that leave their dividends unchanged. (H6): Following earnings declines, Y0,T is on average higher for the firms that increase their subsequent dividends than those that reduce their dividends. Throughout our paper, we test the difference in the average Y0,T for our portfolios based on t-tests of the mean differences and Wilcoxons test of the median differences.

II. Data and Evidence A. Variable Definitions

The two main variables of interest are dividend changes and earnings changes. We measure change in dividend as Di , 0 ( Di , 0 Di , 1 ) / Di , 1 , where Di,t is firm is dividend in quarter t. Measuring changes in earnings is tricky as different models of expected earnings deliver different measures. We initially measure the change in prior earnings as the scaled change in earnings in the most recent four quarters over the preceding four quarters, adjusted for the fouryear earnings drift (prior studies document drift in earnings; e.g., Foster (1977)).5 Thus, the change in earnings for firm i in quarter 0 is computed as:
Yi ,0 100 * ((Ei ,0, 3 Ei , 4, 7 ) ( Ei , 4, 7 Ei , 20, 23 ) / 4) / BEi ,0 ,

where Ei,t,x is the sum of firm is earnings from quarter t to quarter x and BEi,t is the firms book value of common equity at the end of quarter t. Like Nissim and Ziv (2001), book values of equity that are less than 10% of the firms total assets are set to 10% of the total assets to minimize distortions from the deflation (this affects .8% of our sample). In addition, we winsorize changes in earnings at 1% to minimize outlier effects. Similar to previous studies, we focus on analyzing the changes in earnings in the four and the eight quarters following the quarters for which the dividends were announced. These are computed as :
Yi ,1 100 * (( Ei , 4,1 Ei ,0, 3 ) ( Ei ,0, 3 Ei , 16, 19 ) / 4) / BEi ,0 and Yi , 2 100 * (( Ei ,8,5 Ei ,0, 3 ) ( Ei ,0, 3 Ei , 16, 19 ) / 4) / BEi ,0 .
5

Adjusting for the three-year or five-year drift does not affect our qualitative results.

10

We measure the future earnings changes with respect to the same benchmark (Ei,0,-3) as the permanence hypothesis is about whether the change in current earnings will reverse in the future or not. Furthermore, gauging the change in future earnings with respect to the same benchmark minimizes the problems with measuring earnings changes discussed in Nissim and Niv (2001). Thus, we compute the average of Yi,1 and of Yi,2 for each of our portfolio of firms, and test the differences in these averages to test hypotheses (H1) through (H6).

B. Sample Selection We extract data for New York Stock Exchange (NYSE), American Stock Exchange (AMEX), and NASDAQ listed common stocks that have dividend information in the Center for Research in Security Prices (CRSP) database and have quarterly earnings before extraordinary items (Data No. 8) in the COMPUSTAT database. Our sample covers January 1979 through December 2009 and to remain in the sample in a quarter, a firm must meet the following criteria: (i) the firm must declare at least two consecutive quarterly regular cash dividends (CRSP distribution code 1232) and the declaration dates must be in CRSP, (ii) no other distributions were announced between the declaration of the previous dividend and the declaration of the current (e.g., Nissim and Zvi (2001)), (iii) the firm must have earnings announcement dates prior to the dividend announcement and quarterly earnings information for the twenty-four consecutive quarters prior to and the eight consecutive quarters following the dividend announcement in COMPUSTAT,

11

(iv) the dividend announcement does not represent a dividend omission or initiation since omissions and initiations are associated with pronounced market reactions (e.g., Asquith and Mullins (1983) and Michaely, Thaler and Womack (1995)), (v) the firm must maintain dividends or change dividends by at least 10% in absolute value to ensure that the dividend change is informative (e.g., Yoon and Starks (1995)), (vi) past earnings must change by at least 0.5% in absolute value to ensure that the earnings change is informative (this removes about 7% of our observations), and (vii) the dividend announcement must occur at least 2 days after the earnings announcement.6 Furthermore, a dividend changer must not reverse that policy in the following four quarters (our qualitative results remain the same if we require that the policy is not changed in the following eight quarters). To reduce contamination of the future earnings of the firms that leave their dividends unchanged, dividend changers are never included in the non-dividendchanging sample in the following four quarters even if they do not change their dividends in these quarters. After applying these filters, we are left with a sample of 2,988 instances of dividend increases, 23,229 cases with unchanged dividends, and 513 cases of dividend reductions across 2,625 firms.

Our qualitative results remain the same if we include dividends that were announced a day after the earnings announcement or if all dividend changes are included in the sample.

12

C. Descriptive Statistics Each quarter, we sort the firms by the signs of their prior earnings changes and their dividend changes. Panel A of Table 1 reports the distribution of prior earnings increases of the firms by their dividend announcements. For these firms, the mean (median) past earnings increases are 5.38 (3.57), 5.83 (3.49), and 4.99 (3.33) for the firms that increase, maintain, and reduce their dividends, respectively. Thus, there is no monotonic pattern in earnings increases across firms that increase, maintain, or reduce dividends. Turning to firms that experience declines in their prior earnings (Panel B), the mean (median) prior earnings growths are -4.48 (-2.39), -6.57 (-3.87), and -7.54 (-5.35) for the firms that increase, maintain, and reduce their dividends. These results suggest that the increases in past earnings increase monotonically across firms that increase, maintain, and reduce their subsequent dividends. Panels C (D) of Table 1 presents the distributions of dividend increases (reductions) by the sign of prior earnings changes. In Panel C, the mean (median) dividend increases are 0.25 (0.17) and 0.38 (0.17) for the firms that experience past earnings increases and declines, respectively. Thus, the average dividend increase is not higher among the firms that experience past earnings increases compared to those that experience past earnings declines. In Panel D, the mean (median) dividend reductions are 0.39 (0.40) and 0.44 (0.47) for the firms with increases and those with declines in past earnings, respectively. Thus, it appears dividend reductions are larger among firms that experience prior earnings declines than those that experience prior earnings increases.

13

D. Prior earnings changes, dividend changes, and future earnings performance Table 2 presents the mean and median earnings changes for our sample firms sorted by their past earning changes and dividend announcements. Panel A reports the results for the firms with prior earnings increases and, therefore, the tests for hypotheses (H1) through (H3). The results show that the average change in future earnings is higher for the firms with corroboratory dividend increases than those that leave their dividends unchanged. For instance, the median earnings changes for the firms with corroborating dividend increases are 0.37 and 1.60 in the four and eight quarters after the dividend increases. Each of these median increases is significantly higher than the corresponding changes of -0.01 and 0.56 for the firms that do not change their dividends (the p-values of the differences are each less than 0.01). The results from the means are qualitatively the same. This replicates the Benartzi et al. (1997) result that higher earnings are more permanent for firms that increase their subsequent dividends than those that leave their dividends unchanged, supporting (H1). Turning to the firms that reduce their dividends after earnings increases, we see that their median earnings changes are 0.44 and 1.59 in the four and eight quarters after the dividend reductions. Like the firms with confirmatory dividend increases, these median changes are significantly higher than corresponding changes for the firms that maintain their dividends. The results from the means deliver the same conclusion. This contradicts (H2) that higher earnings should reverse more for firms that subsequently reduce dividends than for the firms that maintain their dividends. Further, contrary to (H3), the change in earnings for firms with corroboratory dividend increases is not significantly higher than that for firms with contradictory dividend reductions. These results suggest that the future earnings performance

14

of the firms with contradictory dividends reductions is stronger than the permanence hypothesis suggests. We discuss this result in the next section. Panel B of Table 2 presents the results for the firms that experience declines in prior earnings, testing hypotheses (H4) through (H6). The median changes in earnings for the firms with contradictory dividend increases are 1.95 and 2.49 in the four and eight quarters after the dividend increases, respectively. These increases are significantly higher than the corresponding values of 1.18 and 1.91 for firms that do not change their dividends (p-values for the difference test are less than .01). The results from analyzing the means are similar. Consistent with the permanence hypothesis, this result suggests that declines in prior earnings reverse more for the firms that increase their dividends than those that maintain theirs, supporting (H4). For the firms with confirmatory dividend reductions, the median earnings changes are 0.24 and 1.66 in the first four and eight quarters after the dividend reductions. The change over the first four quarters is significantly lower than the change for the firms that maintain their dividends (p-value of the difference is less than .01). Analysis of the means provides same conclusion. This is consistent with hypothesis (H5) that lower earnings are more persistent among firms that subsequently reduce their dividends than their counterparts that do not change their dividends. Finally, our results show that the median increases in the future earnings of the firms with contradictory dividend increases are higher than those with confirmatory dividend decreases. Thus, declining earnings reverse more for the firms that increase subsequent dividends than those that reduce theirs, supporting hypothesis (H6). In summary, our tests do not reject hypothesis (H1) that higher earnings are more permanent for firms that increase their subsequent dividends than the firms that maintain their

15

dividends. As well, we do not reject hypotheses (H4) through (H6), suggesting that declines in earnings display greater persistence for firms that subsequently reduce dividends, and they reverse more for firms that subsequently increase dividends compared to those that leave their dividends unchanged. This is new evidence in support of the permanence hypothesis. In contrast, our tests reject hypotheses (H2) and (H3), suggesting higher earnings do not reverse more for the firms that subsequently reduce dividends than those that maintain or increase their dividends. In the next section, we discuss the implications of these results.

III. Discussions of Results Our evidence from corroboratory dividend changes supports the permanence hypothesis. While the evidence from confirmatory dividend increases is not new (Benartzi et al. (1997)), that from confirmatory dividend reductions is new. That is, reduced earnings persist more for the firms that subsequently reduce their dividends than for those that leave their dividends unchanged. This result is inconsistent with the documented rebounds in earnings after dividend reductions (e.g., Benartzi et al. (1997) and Jensen et al. (2010)). However, it is in line with DeAngelo et al. (1992), who report that losses tend to persist more for firms that reduce their dividends than those that maintain theirs and is also consistent with market reactions to confirmatory dividend reductions reported in Koch and Sun (2004). This result suggests that managers reduce dividends when they expect persistence in reduced earnings. With regards to firms with contradictory dividend changes, we find reduced earnings tend to reverse more among firms that subsequently increase their dividends than those that leave their dividends unchanged. Again, this evidence supports the permanence hypothesis and it suggests that managers do increase dividends in anticipation of future earnings increases. This

16

is consistent with a direct relation between dividend increases and future earnings predicted by the models of Bhattacharya (1979) and Miller and Rock (1985). These models suggest that, in the presence of information asymmetry, management can use dividend increases to convey information about future earnings. Our result shows that managers use dividend increases to signal future earnings growth if current earnings declines convey poor earnings information, which contradicts their belief that earnings will grow. Also, the result sheds light on the mixed evidence on the relation between dividend increases and future earnings growth. For instance, while Aharony and Dotan (1994) and Nissim and Ziv (2001) document a positive relation between increases in dividends and future earnings changes, Bernatzi et al. (1997) and Grullon et al. (2005) suggest that there is no relation. Our result suggests that the positive relation between dividend increases and future earnings is peculiar to firms with declines in their prior earnings. The result from contradictory dividend reductions, however, does not support the permanence hypothesis. Specifically, following earnings increases, the future earnings changes of the firms that reduce their dividends are not lower than those that maintain or increase their dividends. Thus, it does not appear that contradictory dividend reductions signal reversal of prior earnings increases or declines in future earnings.7 We provide an explanation that hinges on managers reluctance to provide bad news (reduce dividends) due to the associated negative market reactions. In particular, since management is not enthusiastic about reducing dividends it will try not to reduce the dividend even if it anticipates that the current earnings increases will reverse. It will rather wait until it is compel by actually deteriorations in earnings to reduce dividends. Thus, dividend reductions that follow earnings increases may be motivated by

Several studies document growth in earnings after dividend reductions (e.g., Benartzi et al. (1997) and Jensen et al. (2010)).

17

concerns that the increases in current earnings will reverse. Thus, unlike contradictory dividend increases where eagerness to provide goods news can result in dividend increases in anticipation of reversals in current earnings declines, it is hard to see why bad news adverse managers will reduce dividends in anticipation of reversals in current earnings increases. We, therefore, investigate the policy of reduced dividends after an increase in earnings. The strong earnings performance of the firms with contradictory dividend reductions points to growth-motivated dividend reductions. The future earnings growth can be due to increased investments or to cost savings from letting future growth options expire (Jensen et al. (2010)), in which case investment will fall. To investigate this, we calculate the median capital expenditure and median change in capital expenditure of our sample firms and report the results in Table 6. From the table, the median investment per book value of assets in the four and eight quarters after the dividend announcement is highest for the firms with contradictory dividend reductions. Also, these firms display the highest increases in capital expenditure in the same periods.8 This suggests that contradictory dividend reductions are motivated, at least in part, by cash needs to fund investment. Thus, with the exception of contradictory dividend reductions, our evidence shows that dividend changes are managerial signals about the permanence of prior earnings changes.

These results are contrary to those from several prior studies that report declines in capital spending after dividend reductions (e.g., Yoon and Starks (1995) and Jensen et al. (2010)). Our finding, however, is consistent with Grullon et al. (2002), who report increases in capital expenditure after dividend reductions. Our results show that firms with confirmatory dividend reductions reduce their capital expenditures while those with contradictory dividend reductions increase them.

18

IV.

Robustness

Previous studies show that alternate models of earnings dynamics can deliver fundamentally different results (e.g., alternative assumptions about mean revision in earnings resulted in Nissim and Ziv (2001) and Grullon et al. (2005) reaching different conclusions regarding the relation between dividend changes and future earnings changes). To minimize the possibility that our results are driven by earnings dynamics, we check our results by using a matched samples as well as different measures for changes in earnings.

A. Matching firms Our previous results compare the dividend changers to all their non-dividend-changing counterparts. In this section, we use past earnings performance and industrial matches to check our results. To match on past performance, we follow Benartzi et al. (1997) and use as controls the non-dividend-changing firms whose prior earnings changes are within two percent of the respective dividend changers mean (median) in the mean (median) analysis.9 Relative to the results reported in Table 3, the mean (median) prior earnings changes for the dividend changers and their matched groups are closer and the significant differences disappear. Despite this, our conclusions are unchanged, suggesting they are not driven by differences in past earnings performance. To match by industry, we adjust each firms earnings change by the average change for the firms in the industry that do not change their dividends. Thus, firm is earnings change is computed as:
1 Yi ,0 100 * ((Ei ,0, 3 Ei , 4,7 ) / BEi ,0 J
9

(E
j 1

j ,0, 3

E j ,4, 7 ) / BE j ,0 )

We do not match dividend reducers against dividend increasers due to the limited number of observations.

19

where j = 1,...,J are all the firms that do not change their dividend in the quarter and are in the same industry as firm i, defined by the two-digit Standard Industrial Classification (SIC) code. We adjust earnings changes for the industry means before classifying them by the sign of their past earnings changes and, hence, the means for the non-dividend-changing groups are not equal to zero. The results, presented in Table 4, are qualitatively the same as those reported earlier, indicating that our results are not influenced by differences in industrial composition of our test groups.

B. Other measures of earnings change The other metrics that we use to capture changes in earnings are: (i) the seasonal random walk earnings change scaled by the book value of common equity, calculated as Yi ,0 100 * ( Ei ,0 Ei , 4 ) / BEi ,0 ,where Ei,t is the earnings of firm i in quarter t (e.g., Koch and Sun (2004)). Yi,4 and Yi,8 are the sum of the quarterly earnings changes in the four and eight quarters after the dividend announcements. (ii) the change in earnings in the first four quarters over the preceding four quarters scaled by the book value of common equity, computed as Yi ,0 100 * ( Ei ,0, 3 Ei , 4, 7 ) / BEi ,0 . (iii) the earnings change adjusted for linear mean revisions in earnings, computed as:
0 ), where are the estimates of 0 and from the following pooled 0 and Yi , 4 100 * (

regression: (Ei,4,1 Ei,0,-3)/BEi,0 = 0 + 2ROEi,0 + 3(Ei,0,-3 Ei,-4,-7)/BEi,0 + i,4,0, (1)

and ROEi,0 = Ei,0,-3/BEi,0. We compute Yi,8 in the same fashion by replacing Ei,4,1 in equation (1) with Ei,8,5. ROE and prior earnings performance capture mean reversion in future earnings and we obtain negative coefficient estimate on each (e.g., Nissim and Ziv (2001)). Adding

20

dummies for positive and negative dividend changes as additional regressors does not affect our conclusions. (iv) change in earnings adjusted for nonlinear mean revisions, computed as
0 ), and 0 and are the estimates of 0 and from the following pooled Yi , 4 100 * (

regression: (Ei,4,1 Ei,0,-3)/BEi,0 = + 1REi,0 + 2ND0*REi,0 + 3ND0*(REi,0)2 + 4PD0*(REi,0)2 + 1CEi,0 + 2AD0*CEi,0 + 3AD0*(CEi,0)2 + 4BD0*(CEi,0)2 + i,4,0 (2)

RE0 = ROE0 E[ROE0], where E[ROE0] is the fitted value from the cross-sectional regression of ROE0 on its lagged value, and on the logarithm of total assets and of the market-to-book ratio of equity in the preceding quarter; CE0 = (E0,-3 E-4,-7)/B-1; ND (PD) is a dummy variable set to one if RE0 is negative (positive) and zero otherwise, and AD (BD) is a dummy variable set to one if CE0 is negative (positive) and zero otherwise. Again, Yi,8 is obtained by replacing Ei,4,1 in equation (2) with Ei,8,5. The dummy variables and the squared terms account for the findings that large changes in earnings revert faster than small changes and negative changes revert faster than positive changes (e.g., Grullon et al. (2005)). To conserve space, we only report the results from the medians in these alternate tests in Panels A through D in Table 5 (the mean results are similar). It is evident that our conclusions remain the same in each of these alternative specifications of earnings changes. Thus, it does not appear that our results are influenced by different evolution processes in earnings across our test groups.

21

C. Alternative Deflators for Earnings Change and Dividend Change We repeat our tests using alternative deflators for earnings changes. These include market value of common equity, market value of the firm, and book value of the firm. In addition, we repeat the analysis deflating the dividend change by the book value of common equity and by the closing price on the dividend-announcement day and excluding firms whose dividend change is in the decile closest to zero. In all cases, the results are qualitatively the same as those tabulated.

V. Concluding Comments It is well-documented that dividend changes induce stock price changes, but there is no consensus on the exact information that the markets react to. Cash flow models suggest that dividend changes are a signal about future earnings growth. Collectively, the evidence from future earnings patterns does not support this view. More recent empirical work by Koch and Sun (2004) finds that market reactions to dividend changes suggest that investors interpret dividend changes as revealing private managerial information about the permanence of prior earnings changes. This view, too, has implications for future earnings behavior. In particular, it suggests that increased earnings should be more permanent (transitory) for firms that subsequently increase (reduce) their dividends than those that maintain their dividends. Likewise, reduced earnings should be more permanent (transitory) for the firms that subsequently reduce (increase) their dividends than those that maintain their dividends. Consequently, we study the hypothesis that dividend changes are signals about the permanence of past earnings changes based on the future earnings patterns of firms sorted by their past earnings performance and their dividend announcements.

22

Our results from confirmatory dividend changes are consistent with the permanence hypothesis. In particular, they show that increases in earnings are more persistent for the firms that subsequently increase dividends than those that maintain their dividends. Also, declines in earnings are more persistent for the firms that subsequently reduce dividends than those that leave their dividends unchanged. Thus, in addition to evidence that managers increase dividends when they expect permanence in higher current earnings, we find that managers also reduce dividends when they expect persistence in lower current earnings. The evidence from contradictory dividend changes shows that declines in current earnings tend to reverse more for the firms that subsequently increases dividends than those that maintain their dividends. However, contrary to the permanence hypothesis, we do not find that increases in current earnings tend to reverse more for the firms that subsequently reduce dividends than those that leave their dividends unchanged. Thus, the evidence from contradictory dividend increases is mixed. We explain the mixed results by the asymmetry in managements willingness to reduce versus increase dividends. Specifically, while managers are reluctant to reduce dividends (reveal bad news), they are eager to increase dividends (reveal good news). Thus, we argue that managers increase dividends in anticipation of reversals in lower current earnings, but they do not reduce dividends in anticipation of reversals in the higher current earnings. This argument is consistent with the information asymmetry models that suggest managers use dividend changes to signal good news about future earnings if current information is to the contrary (declines in current earnings). In contrast, bad news adverse managers may not reduce dividends even if they expect reversals in higher current earnings. Consequently, it is unlikely that contradictory dividend reductions are motivated by managers expectation that higher current earnings will reverse.

23

Since the contrary evidence from contradictory dividend reductions stems from stronger future earnings growth, we investigate the capital expenditures of our sample firms and find that contradictory dividend reducers display the largest increases in capital expenditure. This suggests that the policy to reduce dividends following earnings increases are likely driven by cash needs to boost investment activities. Overall, therefore, the evidence from future earnings patterns supports the view that dividend changes reflect private managerial information about the persistence of prior earnings changes.

24

References Aharony, Joseph, and Amihud Dotan, 1994, Regular dividend announcements and future unexpected earnings: An empirical analysis, Financial Review 29, 125-151. Asquith, Paul, and David W. Mullins, 1983, The impact of initiating dividend payments on shareholders wealth, Journal of Business 56, 77-96. Benartzi, Shlomo, Roni Michaely, Richard Thaler, 1997, Do changes in dividends signal the future or the past? Journal of Finance 52, 1007-1034. Bhattacharya, S., 1979, Imperfect information, dividend policy, and the bird in the hand fallacy, Bell Journal of Economics 10, 259-270. Brickley, J. A., 1983, Shareholder wealth, information signaling and the specially designated dividends: An empirical study. Journal of Financial Economics 12, 187-209. DeAngelo, Harry, Linda DeAngelo, and Douglas J. Skinner, 1992, Dividends and Losses, Journal of Finance 47, 1837-1863. DeAngelo, Harry, Linda DeAngelo, and Douglas J. Skinner, 1996, Reversal of fortune: Dividend signaling and the disappearance of sustained earnings growth, Journal of Financial Economics 40, 341-371. Foster, George, 1977, Quarterly accounting data: Time series properties and predictive ability results, Accounting Review 52, 1-21. Gonedes, Nicholas J., 1978, Corporate signaling, external accounting, and capital market equilibrium: Evidence on dividends, income and extraordinary items, Journal of Accounting Research 16, 26-79.

25

Grullon, Gustavo, Roni Michaely, Shlomo Benartzi, and Richard H. Thaler, 2005, Dividend changes do not signal changes in future profitability, Journal of Business 78, 16591682. Healy, P. M. and K. G. Palepu, 1988, Earnings information conveyed by dividend initiations and omissions, Journal of Financial Economics 21, 149-175. Jensen, Gerald R., Leonard Lundstrum, and Robert Miller, 2010, What do dividend reductions signal? Journal of Corporate Finance 16, 736-747. John, K. and J. Williams, 1985, Dividends, dilution, and taxes: A signaling equilibrium. Journal of Finance 40, 1053-1070. Kane A, Y. K. Lee, and A. Marcus, 1984, Earnings and dividend announcements: Is there a corroboration effect? Journal of Finance 39, 1091-1099. Koch, A. S. and A. X. Sun, 2004, Dividend changes and the persistence of past earnings changes, Journal of Finance 59, 2093-2116. Lintner, John, 1956, The distribution of incomes of corporations among dividends, retained earnings and taxes, American Economic Review 46, 97-113. Michaely, Roni, Richard Thaler, and Kent Womack, 1995, Price reactions to dividend initiations and omissions: Overreaction or drift? Journal of Finance 50, 573-608. Miller, M. H. and F. Modigliani, 1961, Dividend policy, growth, and the valuation of shares, Journal of Business 34, 411-433. Miller, M. H. and K. Rock, 1985, Dividend policy under asymmetric information, Journal of Finance 40, 1031-1051. Nissim, Doron and Amir Ziv, 2001, Dividend changes and future profitability, Journal of Finance 56, 2111-2133.

26

Penman, S. H., 1983, The predictive content of earnings forecasts and dividends, Journal of Finance 38, 1181-1199. Watts, R., 1973, Information Content of Dividends. Journal of Business 46, 191-211. Yoon, Pyung S., and Laura T. Starks, 1995, Signaling, investment opportunities, and dividend announcements, Review of Financial Studies 8, 995-1018.

27

Table 1 Distribution of Changes in Earnings and Dividends


The table presents the distribution of prior earnings changes by dividend announcements for firms with quarterly regular cash dividend information in CRSP and quarterly earnings, along with their announcement dates in COMPUSTAT. The sample covers January 1979 through December 2009. To remain in the sample the firm must have twenty-four consedecreaseive quarters of earnings before and eight consedecreaseive quarters of earnings after the dividend announcements. Each quarter, the firms are sorted by the changes in earnings, computed as Ei , 0 100 * ( Ei , 0 , 3 Ei , 4 , 7 ) / BE i , 0 , where Ei,t,x is the sum of firm is earnings from quarter t to quarter x and BEi,t is the book value of the firms common equity at the end of quarter t. The firms are then classified by their subsequent dividend announcements.
Percentiles P1 P10 P25 P50 P75 P90 P99 Mean Firmquarters

Panel A: Distribution of Prior Earnings Increases by Dividend Announcements Dividend Increases Unchanged Dividends Dividend Decreases 0.37 0.16 0.17 1.47 0.88 1.20 2.26 1.75 2.13 3.57 3.49 3.33 6.02 6.86 5.86 10.62 13.24 9.79 29.77 34.55 23.56 5.38 5.83 4.99 2,374 11,882 232

Panel B: Distribution of Prior Earnings Declines by Dividend Announcements Dividend Increases Unchanged Dividends Dividend Decreases -0.07 -0.13 -0.10 -0.34 -0.66 -0.62 -0.87 -1.62 -1.66 -2.39 -3.87 -5.35 -5.49 -8.41 -10.94 -10.59 -14.98 -15.76 -28.30 -39.38 -42.98 -4.48 -6.57 -7.54 614 11,347 281

Panel C: Distribution of Dividend Increases by Prior Earnings Changes Earnings Increases Earnings Declines 0.10 0.10 0.11 0.11 0.12 0.12 0.17 0.17 0.25 0.25 0.40 0.50 1.40 2.00 0.25 0.38

Panel D: Distribution of Dividend Decreases by Prior Earnings Changes Earnings Increases Earnings Declines -0.12 -0.13 -0.20 -0.20 -0.30 -0.32 -0.40 -0.47 -0.50 -0.5 -0.50 -0.71 -0.71 -0.90 -0.39 -0.44

28

Table 2 Future Earnings Changes by Current Earnings Changes and Dividend Announcements
The table presents the future earnings changes for firms classified by their prior earnings changes and their dividend announcements for firms with quarterly regular cash dividend information in CRSP and quarterly earnings information in COMPUSTAT. The sample covers January 1979 through December 2009. Each quarter, the firms are sorted by the changes in their earnings, computed as, Ei ,1 100 * (( Ei , 0 , 3 Ei , 4, 7 ) ( Ei , 4, 7 Ei , 20 , 23 ) / 4) / BE i , 0 where Ei,t,x is the sum of firm is earnings from quarter t to quarter x and BEi,t is the book value of the firms common equity at the end of quarter t. The firms are then classified by their subsequent dividend announcements. Panel A presents the results for prior earnings increases and Panel B for prior earnings declines.
Dividend Increases Panel A: Earnings Increases Previous Four Quarters First Four Quarters First Eight Quarters 4.57** [2.47]** -0.32** [0.37]** -0.18** [1.60]** 6.51 [3.48] -0.94 [-0.01] -0.52 [0.56] 4.66** [2.83]** -0.43** [0.44]** 0.04** [1.59]** Unchanged Dividends Dividend Reductions

Panel B: Earnings Decreases Previous Four Quarters First Four Quarters First Eight Quarters
*,**

-6.58**(##) [-4.25]**(##) 2.93**(##) [1.95]**(##) 2.34 [2.49]**(##)

-7.71 [-4.76] 2.14 [1.18] 2.39 [1.91]

-8.30** [-5.98]** 0.63** [0.24]** 2.45 [1.66]

Significantly different from the no-change group at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians. #,## Significantly different from the dividend reducers at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians.

29

Table 3 Matching Dividend Maintainers to Changers by Past Performance


The table presents the future earnings changes measured differently for firms by their prior earnings changes and their dividend announcements. The sample covers January 1979 through December 2009. To be included in the sample, the firm must have quarterly regular cash dividend information in CRSP and quarterly earnings information in COMPUSTAT. Each quarter, the firms are sorted by the changes in their earnings, computed as Ei ,1 100 * (( Ei , 0 , 3 Ei , 4 , 7 ) ( Ei , 4 , 7 Ei , 20 , 23 ) / 4) / BEi , 0 , where Ei,t,x is the sum of firm is earnings from quarter t to quarter x and BEi,t is the book value of the firms common equity at the end of quarter t. The firms are then classified by their subsequent dividend announcements. The dividend changers are matched against their counterparts that left their dividends unchanged and their year 0 earnings changes within two percentage points of the mean (median) in mean (median) analysis for the respectively change group. Panel A presents the results for prior earnings increases and Panel B for prior earnings declines.
Dividend Increases Increases Panel A: Prior Earnings Increases Previous Four Quarters First Four Quarters First Eight Quarters 4.57 [2.47] -0.32** [0.37]** -0.18** [1.60]** 4.31 [2.52] -1.08 [-0.16] -0.64 [0.41] 4.66 [2.83] -0.43** [0.44]** 0.04 [1.59]** 4.45 [2.71] -0.96 [-0.22] -0.20 [0.39] Matched Non-Change Dividend Reductions Reductions Matched Non-Change

Panel B: Prior Earnings Decreases Previous Four Quarters First Four Quarters First Eight Quarters
*,**

-6.58 [-4.25] 2.93** [1.95]** 2.34** [2.49]**

-6.36 [-3.98] 0.47 [0.49] 0.72 [1.12]

-8.30 [-5.98] 0.63** [0.24]** 2.45 [1.66]

-8.11 [-5.77] 1.53 [0.96] 2.74 [1.41]

Significantly different from the matched no-change group at the 10% and 1% levels using a twotailed Students t-test for the means and a two-tailed Wilcoxon test for the medians. #,## Significantly different from the dividend reducers at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians.

30

Table 4 Earnings Changes Adjusted by Industry


The table presents the future earnings changes for firms classified by their prior earnings changes and their dividend announcements for firms with quarterly regular cash dividend information in CRSP and quarterly earnings information in COMPUSTAT. The sample covers January 1979 through December 2009. Each quarter, we calculate earnings change for each firm i as

Ei,0 100 * ((Ei,0, 3 Ei, 4, 7 ) / BEi,0

1 J

(E
j 1

j , 0, 3

E j , 4, 7 ) / BE j ,0 ) ,

where j = 1, ... , J are all the firms that did not change their dividend in the quarter and have the same two-digit Standard Industrial Classification (SIC) code as firm i. The firms are then
classified by their earnings change and their subsequent dividend announcement. Panel A presents the results for prior earnings increases and Panel B for prior earnings declines. Medians are reported in square brackets.
Dividend Increases Panel A: Earnings Increases Previous Four Quarters First Four Quarters First Eight Quarters 5.15** [3.40]** 0.91**(##) [0.76]**(#) 0.83** [1.06]** 4.69 [3.04] -0.89 [-0.38] -0.99 [-0.45] 4.82 [3.48]** -0.10** [0.37]** 0.49** [1.02]** Unchanged Dividends Dividend Reductions

Panel B: Earnings Decreases Previous Four Quarters First Four Quarters First Eight Quarters
*,**

-4.12**(##) [-2.51]**(##) 2.14**(##) [1.25]**(##) 2.24**(##) [1.58]**(##)

-4.88 [-3.06] 0.93 [0.66] 1.03 [0.96]

-6.26** [-3.29] 0.16** [-0.25]** -0.03** [0.02]**

Significantly different from the no-change group at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians. #,## Significantly different from the dividend reducers at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians.

31

Table 5 Different Measures of Earnings Changes


The table presents the median percentage future earnings changes, computed with different metrics, for firms by their prior earnings changes and their dividend announcements. The sample covers January 1979 through December 2009. To be included in the sample, the firm must have quarterly regular cash dividend information in CRSP and quarterly earnings information in COMPUSTAT. Each quarter, the firms are sorted by the changes in their past earnings changes and by their subsequent dividend announcements. Ei,t, is the earnings of firm i in quarter t; Ei,t,x is the sum of firm is earnings from quarter t to quarter x, and BEi,t is the book value of common equity at the end of quarter t.
Prior Earnings Increases Dividend Increases Unchanged Dividends Dividend Reductions Prior Earnings Declines Dividend Increases Unchanged Dividends Dividend Reductions

Panel A: Seasonal Random Walk Changes computed as Ei ,0 100 * ( Ei ,0 Ei , 4 ) / BEi ,0 Subsequent Four Quarters Subsequent Eight Quarters 2.48** 2.93** 1.36 1.65 2.29** 3.16** 0.78**(##) 0.04**(##) -0.94 -0.49 -2.68** -2.09**

Panel B: Four-Quarter Changes computed as: Ei ,0 100 * ( Ei ,0, 3 Ei , 4, 7 ) / BEi ,0 Subsequent Four Quarters Subsequent Eight Quarters 1.84**(##) 1.80** -0.23 0.00 1.13** 1.71** 3.17**(##) 3.04**(##) 0.89 1.30 -1.99** 0.51**

Panel C: Earnings Changes Adjusted for Linear Mean Revisions computed using Equation (1) Subsequent Four Quarters Subsequent Eight Quarters 2.90** 2.75** 1.71 1.35 2.78** 2.69** 0.43**(##) -1.58**(##) -1.76 -3.20 -4.16** -4.45**

Panel D: Earnings Changes Adjusted for Nonlinear Mean Revisions computed using Equation (2) Subsequent Four Quarters Subsequent Eight Quarters
*,**

1.42** 1.33**

0.43 -0.11

1.34** 1.25**

1.78**(##) 0.05**(##)

-0.12 -1.81

-1.56** -3.45**

Significantly different from the no-change group at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians. #,## Significantly different from the dividend reducers at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians.

32

Table 6 Capital Expenditures by Past Performance and Dividend Changes


The table presents the median capital expenditures for our sample of the firms by their prior earnings performance and their dividend announcements. The sample period covers January 1979 through December 2009. Each quarter, the firms are sorted by the changes in their earnings, computed as Ei ,1 100 * (( Ei , 0, 3 Ei , 4, 7 ) ( Ei , 4, 7 Ei , 20 , 23 ) / 4) / BEi , 0 , where Ei,t,x is the sum of firm is earnings from quarter t to quarter x and BEi,t is the book value of the firms common equity at the end of quarter t. The firms are then classified by their subsequent dividend announcements. Capital expenditure is the expenditure in subsequent four and eight quarters after the dividend announcements, scaled by the book value of the firms assets.
Capital Expenditure Levels Dividend Increases Unchanged Dividends Dividend Reductions Percentage Changes in Capital Expenditure Dividend Increases Unchanged Dividends Dividend Reductions

Panel A: Prior Earnings Increases Most Recent Four Quarters Subsequent Four Quarters Subsequent Eight Quarters 0.045(#) 0.043 0.050** 2.1** -0.8 3.6*

0.046(##)

0.044

0.056**

2.9(#)

1.5

5.9*

0.098

0.092

0.107*

107.1

105.6

107.8

Panel B: Prior Earnings Declines Most Recent Four Quarters Subsequent Four Quarters Subsequent Eight Quarters
*,**

0.047(##)

0.049

0.035**

-4.9*

-7.2

-6.5

0.045*(##)

0.040

0.026**

-1.9**(##)

-10.5

-12.3

0.094*(##)

0.086

0.061**

105.9**(##)

80.6

69.6*

Significantly different from the no-change group at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians. #,## Significantly different from the dividend reducers at the 10% and 1% levels using a two-tailed Students t-test for the means and a two-tailed Wilcoxon test for the medians.

33