Research Proposal AC9101-Seminar in Capital Markets in Accounting by scl14029

VIEWS: 7 PAGES: 39

									      Information Risk, Dividend Changes, and Market reaction

                                Linna Shi and Huai Zhang




[Abstract]


       In this paper, we investigate how information risk can influence market

reaction to dividend changes, and how information risk can influence the

management’s dividend strategy. Taking into account all possible situations when

firms disclose different information to the market through dividend changes, we find

confirmative results to support dividend signaling theory: (1) for high information risk

firms, market reacts more (both positively and negatively) to dividend changes

information; and (2) for high information risk firms, management choose to use larger

scale of dividend changes to signal information. In sum, information risk can explain

cross-sectional variations of dividend signaling roles across firms.




*Authors’ Names and Affiliations:
Linna Shi, corresponding author, Nanyang Technological University. Tel: +65-8158-
5446. Email: shil0005@ntu.edu.sg. Huai Zhang , Nanyang Technological University
and University of Hong Kong
I. Introduction


       Traditional literatures on dividends argues that dividends have important

information content which will signal a firm’s future information (including Linter

(1956), Bhattacharya (1979), Miller and Rock (1985), Johns and Williams (1985),

etc.). Specifically, under asymmetric information circumstance, dividend policy

indicates information either on a firm’s future earnings prospect (including including

Ofer and Siegal (1987), Healy and Palepu (1989), Michaely, Thaler, and Womack

(1995), Benartzi, Michaely, and Thaler (1997), Nissim and Ziv (2001), and etc.), or

on its discount rate prospect, in another word, risk prospect (including Venkatesh

(1989), Dyl and Weigand (1998), Grullon, Michaely, and Swaminathan (2002), Chen,

Shelvin, and Tong (2007), and etc.). While empirical evidence on dividend signaling

theory is mixed, all of these studies agree with one common underlying assumption

that the same type of dividend strategy will signal same, or at least very similar,

information among different firms. As a matter of fact, to support this assumption,

many previous studies have found that the market generally positively reacts to

announcements of dividend increase or dividend initiation while negatively responses

to announcements of dividend decrease or dividend omission.


       However, some studies have shown it is not necessarily the case. From the

findings of Dhillon and Johnson (1994), one may notice that in as many as one third

of all the dividend adjustments cases, market’s reaction are in opposite to the direction

of dividend adjustment, i.e. a negative association between market reaction and

dividend changes. Because market reaction to any news can be seen as market’s

interpretation to the information content in the news, we can thus take the market

reaction as a proxy for information content in dividend announcement. Accordingly,


                                           1
different or even opposite market reaction to same type of dividend news as shown in

Dhillon and Johnson (1994) is a good example indicating the significant variations to

same dividend policy among firms with different characteristics. Indeed, increasing

dividend may signal increasing future earnings of one firm; while the same increase

may suggest less growth opportunities or high default risk of another firm. Therefore,

if we oversimplify the dividend signal issue without considering each firm’s

characteristics, we may miss important factors and even get lost when we study

information content of dividend.


       Among all the firm characteristics (such as growth, capital structure, default

risk, profitability, size, institutional structure and etc., refer to Haw and Kim (1991),

Banker, Das, and Datar (1993), Healy, Hathorn, and Kirch (1997), etc. ) that may

influence market reaction to dividend policy, we believe information risk is one of the

most important characteristics. Information risk has been proved to be highly

correlated with a firm’s market performance (including Easely and O’Hara (2004),

O’hara (2003), Leuz and Verrecchia (2005), Zhang (2006), Aboody et al. (2004),

Francis et al. (2005), (2006), etc.). With large (small) information risk, investors will

be less (more) confident on the firm’s current information and future investment

opportunities, and thus they will price lower (higher) for such kind of firms because

of higher (lower) cost of capital.


       Following dividend signaling theory, in a perfect market where there is no

information asymmetry, dividend will signal no information at all because all

information is publicly available; while in the reality when information asymmetry

exists, dividend will signal more information for higher information risk firms than

for lower information risk firms, because such information in dividend policy is more


                                           2
precious and not available otherwise. Consequently, stronger market reaction should

be observed for high information risk firms than for lower ones.


       We find empirical evidence to support this predication. Using a sample with

all significant dividend changes events, we find confirmative results to support the

dividend signaling theory. The two-day cumulative abnormal returns (“CAR”) during

dividend announcement are larger in absolute value when information risk is high

(low). In another word, the relation between intensity of market reaction and dividend

changes is significantly influenced by the level of information risk. Different with

most previous studies, we also consider the possibility that market may negatively

interpret announcement of dividend changes because of different firm’s

characteristics. We find that no matter dividend changes actually disclose good or bad

information, higher information risk firms always obtain stronger reactions from the

market. We thus show information contents of dividend changes are significantly

different under different information risk scenarios.


       Using the same sample, we also examine how information risk can affect the

dividend policy for different firms. Dividend signaling theory predicts that, when

information risk is high, there is more needs to use dividend changes to signal

information to the market. In our empirical test, we use scale in dividend changes to

proxy the needs to use dividend policy for signaling, and find high information risk

firms are more likely to use larger scale of changes in their dividend policy. Therefore

this finding is also consistent with the dividend signaling theory.


         Our findings contradict to those in Li and Zhao (2007). We believe that,

other than signaling consideration, management will have other consideration on

firm’s dividend policy. It is possible that the information risk and a firm’s general

                                            3
dividend policy is affected by the same underlying characteristics, i.e. the volatility of

the firm’s earnings may significantly influence a firm’s decision to pay dividend, to

initiate dividend, to increase dividend, to pay large amount of dividend, as well as

information risk. However, if we focus on dividend changes samples, this concern can

be relaxed to the minor level. Because the general dividend policy (such as whether to

pay, initiate, or increase dividend and how much the dividend level) has been decided

ex ante according to a comprehensive consideration, the decision left on the scale of

changes can be used by the manager to exert the signal function.


       We believe our findings will contribute both to the literature and to the real

world from the following aspects:


       First of all, this paper offers new evidence of support to the signaling effect of

dividends. Past literature on dividend signaling theory mostly focus on the general

relation between stock price changes with and dividend changes, i.e. using the

positive relation between market reaction and dividend changes to prove the signaling

role of dividend to the market. We, however, from a new point of view, further

demonstrate that the information content in dividend can vary across firms according

to their information risk characteristics. Thus we use the intensity of the market

reaction, rather than the direction of stock price changes, to substantiate the signaling

theory of dividend.


       Secondly, this paper reminds us the cross-sectional variations in the market

response to dividend announcement and provides us one additional factor to explain

such variations. In addition to the traditional explanations (firm’s characteristics such

as growth opportunities, profitability, default risk, size, ownership structure, and etc.)

which may affect the market response to dividend signaling information, information

                                            4
risk is also an important factor. More specifically, information risk, which we can use

firm-specific accounting information to measure, can well predict the difference in the

magnitude of market response to the same dividend changes announcement.


       Thirdly, our study also documents the different management decision

regarding dividends changes when they decide to use such changes to signal

information to the market. We show empirical evidence that management tend to use

larger scale of dividend changes when the firm’s information risk is high. Though

some recent literature such as Chen et al. (2007) has shown that dividend can help to

decrease information risk, we provide more direct evidence on whether high

information risk firms do use dividends more in order to signal information.


       Lastly, because our results indicate the substitute role of dividend and

accounting earnings information, our paper may have some implications to the

management on their choice of dividend policy. If a firm has high quality of earnings

information, management may avoid the costs to using dividend to signal the future

expectations, since market reaction will be comparatively lower to such dividend

changes; alternatively, when a firm’s earnings is very volatile, the management may

want to use dividend to disclose their confidence to their firm and may expect strong

market’s response.




II. Literature Review and Hypothesis Development


1. Dividend Signaling Theory




                                          5
       Under traditional dividend signaling theory, outside investors and inside

managers have asymmetric information about the firm’s future performance, and thus

management can use dividend policy to signal insider’s information on future cash

flows. Theoretical analyses (including Linter (1956), Bhattacharya (1979), Miller and

Rock (1985), Johns and Williams (1985), Ofer and Thakor (1987), and etc.) all imply

the managers’ motivation to use dividend policy to disclose inside information to the

market and align the asymmetric information problem. Indeed, as shown in the survey

to corporate managers by Baker and Powell (1999), most of the respondents express

high level of agreement with the statement about signaling.


       However, empirical evidence in testing the dividend signaling role to future

cash flow is, at the best, mixed so far. Some studies (such as Ofer and Siegal (1987),

Healy and Palepu (1988), Brook et al. (1998), Nissim and Ziv (2001), and etc.)

demonstrate that future earnings (or cash flow) information is reflected through

dividend policy, while many other studies fail to find significant relationship between

current dividend change and future earnings (see Macquieria and Meggison (1994),

Michaley et al. (1995), Bernartzi el al. (1997), Grullon et al. (2002), and etc. ),

arguing that dividend changes reflect not future earnings information, but

contemporaneous and previous earnings changes. Some of these studies even show

negative relation between dividend policy and future earnings, for instance, improved

earnings after dividend omission or decrease and deteriorate profitability following

dividend initiation or increase.


       Attempting to explain the puzzle between earnings performance and dividend

signaling policy, some papers start to argue that management use dividend strategy

not necessarily to convey information on future earnings, they may intent to signal


                                          6
information on firm’s risk factors. Studies including Venkatesh (1989), Dyl and

Weigand (1998), Grullon et al. (2002) all find supporting evidence on this concern.


2. Factors Affecting Market Reaction to Dividend Changes Announcement


       One underlying assumptions for the above studies on dividend signaling

theory is market valuation to the dividend policy is consistently positive, i.e. dividend

increase (initiation) is always a good signal to the market while dividend decease

(omission) is always bad. Although most studies find significant market returns on

average supporting this assumption, this is not necessarily the truth. As many as one

third observations show the opposite results in Asquith and Mullins (1983), Dhillon

and Johnson (1994), and Healy et al. (1997), i.e. market reaction to same dividend

news can be so different among firms.


       To explain this phenomenon, we have to understand that dividend policy is

never an isolated signal tool. Instead a firm’s dividend policy may be influenced by

many other factors, such as growth opportunities, agency cost, tax consideration,

default risk, and so on (refer to Rozeff (1982), Miller and Modigliani (1961), Allen

and Michaely (2003), and etc.). Consequently, same dividend policy may have

different implication to the market. For instance, increasing dividend may signal

improved future earnings for one firm; while the same increase may simply mean less

growth opportunities for the other firm. Accordingly, because different information

conveyed from dividend policy, market response to the same dividend news can also

be different.


       Several studies have provided us some example of firm’s characteristics

(public available information before dividend announcement) which may affect


                                           7
information content in dividend policy and market reaction. For example, Asquish

and Mullins (1983) find that market reaction depends on magnitude of dividend

changes while Ghosh and Woolridge (1988) find percentage changes matters. Healy

et al. (1997) provides evidence that growth earnings before dividend announcement,

leverage ratio, P/E ratio, and magnitude of dividend changes all affect market reaction.

Moreover, some studies, such as Haw and Kim (1991), Mitra and Owers (1995), and

Eddy and Seifert (1998), demonstrate that firm’s size may influence market reaction

to dividend announcement. These studies may implicit the effect of information risk

on dividend signaling role because firm size is often argued as a vague proxy to

information risk.


       In summary, it is not easy to thoroughly distinguish between signal role and

other role of dividend, and investors may interpret same dividend policy differently

according to firm’s characteristics.


3. Information Risk and Dividend Changes


       In recent literature, it has been widely accept that information risk is priced by

the market. Easely and O’Hara (2004) proposed a model where information risk

premium increase with private information and decrease with improved information

quality. Aboody et al. (2005) shows that insiders tend to use their private information

to obtain abnormal returns. Both while Francis et al. (2004) and Biddle and Hilary

(2006) shows low earnings quality (high information risk) is associated with high cost

of capital. Moreover, Francis et al (2005), Ecker et al. (2006), Zhang (2006) all

provide evidence that information risk affect market’s price to earnings information.




                                           8
       Besides earnings information, dividend policy is another important source of

information to reduce information asymmetry. In fact, Koch and Sun (2004) have

indicated us the evidence that changes in dividend cause investors to re-evaluate

public available information such as previous earnings. Therefore, it is natural to

bring forward such a question: whether risk on previous available public information

will affect market’s price to the new information source - dividend policy?


       To our best knowledge so far, two studies have been done to investigate the

relation between information risk and dividend signal theory, from different point of

view. Chen et al. (2007), for example, try to compare firm’s information risk

condition before and after dividend changes. They find that dividend changes are

associated with changes in information risk, measured by accounting information

such as e-loading of Francis et al. (2006), analysts forecast performance such as

forecast dispersion in Zhang (2006), or firm’s market performance such as stock

volatility. Their findings confirm signaling theory by telling us an alternative piece

of information content dividend: market reaction to dividend changes can be

explained by changes of information risk.


       Another example is Li and Zhao (2007). They investigate whether

management dividend policy are affected by information asymmetries. Because they

find that firms subject to more information asymmetry are less likely to pay, initiate,

or increase dividend and disburse smaller amount of dividend, they do not support

dividend signaling theory. However, as we discussed above, management decision on

dividend policy is not simply draw by signaling consideration, but made by many

other considerations, the results of this paper just tell us that signaling theory is not




                                            9
the most important consideration for dividend policy, while they are not sufficient

evidence to deny dividend signaling role.


       From a different angle, we can test dividend signaling theory by examining the

link between market reaction and information risk. We can predict one simple fact

under dividend signaling theory: information risk will affect not only information

content in dividend but also market reaction to it as well.


4. Hypothesis Development


       Upon the above discussed literature, we can test the existence of dividend

signaling theory by the link between information risk and information content in

dividend announcement. Several testable hypotheses for this paper are developed as

followings:


       For firms with more information risk, since little public information is

available before dividend changes, information content disclosed in dividend changes

is more. Accordingly, investors will response to dividend information more heavily.

Thus we have our first hypothesis as follows:


       H1: Absolute value of abnormal return around announcement date of

dividend changes is larger (smaller) for high (low) information risk firms.


       Other than market reaction to dividend changes, we believe how management

use dividend changes to signal information to the market is another relevant issue.

When earnings information is more volatile, managers may more heavily depend on

dividend changes to disclose information to the market. Therefore a testable

prediction is the management would like to use larger scale of dividend to convey


                                            10
information when firm’s information risk is high. Thus our second hypothesis is

developed as follows:


         H2: The management of high (low) information risk firms is more likely

to use larger (smaller) scale of dividend changes to convey information to the

market.




III. Sample Selection and Variable Measurement


         Following the similar methodology as Benartzi, Michaely, and Thaler (1997),

we select data from CRSP and COMPUSTAT, for the period from 1963-2006. We

require the following criteria for a firm-year observation to be included in the sample:


         1. Closed-end funds, REITs, stock certificates and ADRs are excluded;

         2. The distribution is a regular quarterly cash dividend paid in U.S. dollars

             (distribution code 1232). Excluded are dividends defined as special, year-

             end interim or non-recurring and dividends paid at other frequencies or in

             foreign currency. Also excluded are dividend initiations and resumptions 1 ;

         3. The dividend announcement must have a valid announcement date. The

             firm must have returns information from day -5 to day +5 around the

             dividend announcement date;

         4. There is no announcement of other distributions in a 30-day window (days

             -15 to 15 day surrounding the announcement); Excluded are those

             dividend changes events that concurred earnings announcement, i.e.


1
 The same test can be applied to investigate initiation, omission, resumption of dividend, rather than
dividend changes, as robustness test.

                                                  11
           observations when dividend declare date = (earnings announcement date -

           1 to announcement date +1);

       5. In order to exclude those insignificant dividend changes events, we require

           the quarter dividend to be increased / decreased by more than 12.5% and in

           order to prevent other extreme simultaneous event, we require the dividend

           changes less than 500%;

       6. We exclude all the firms in financial industry and utilities industry, i.e.

           SIC Code = 4049-4999 or 6000-6999.

       7. In order to calculate the accruals quality, we need to require at least 3

           years accrual information before dividend announcement date and we

           require accruals quality to be non-missing before the announcement date

           of dividend changes.

       8. We require total asset, market value of equities, market to book ratio, sales

           growth information, and leverage ratio to be non-missing before the

           announcement date of dividend changes.

       9. To minimize any influence of extreme value on our tests, we winzorize our

           sample if any variable in 6 and 7 fall outside the range of 0.5% to 99.5%.


       Since Kathy and Wu (2001) remind us the effect of survivorship on inference

of dividend signaling, we decide, unlike in Benartzi, Michaely, and Thaler (1997), not

to require a firm to keep operation and disclose earnings information in the following

2 years after dividend announcement. The resulting sample contains 18217

observations.


       Similar to Michaley et al. (1997), we choose either market-capitalization

decile returns as benchmark to calculate cumulative abnormal returns (“CAR”)


                                          12
around announcement of dividend changes. We then calculate 3-day CAR (from 1

day before announcement date to 1 day after announcement date) from a buy-and-

hold strategy:


                           1              1
         CAR j ( −1, +1) = ∏ (1 + Rt ) − ∏ (1 + MRt )
                         t = −1         t = −1




         where CAR j ( −1, +1) = 3-day cumulative abnormal return around announcement

of dividend changes from day -1 to +1; Rt = raw return for observation firm j on day t;

MRt = corresponding benchmark return on day t. We define market reaction to

dividend news as the absolute value of CAR j ( −1, +1) , i.e. ABS j ( −1, +1) = CAR j ( −1, +1)   2




         Table 1 reports the frequency of dividend increasing and decreasing events

and the 3-day cumulative abnormal return conditions around dividend changes

announcement. Among the totally 18217 dividend changes event in our sample,

14492 events are dividend increase, accounting for 80% of the whole sample, while

3725 are dividend decease, around 20% of the whole sample. Consistent with

previous literature, when firms announce to increase dividend, market reacts to such

news positively to 57.5% firms and reacts negatively to as many as 42.5% firms.

When firms decrease dividend, more than 53.7% firms get opposite market response,

i.e. positive market response.


         [Insert Table 1]




2
 We also test CAR when benchmark return = market return or industry matching firm return. Results
are robust in all the tests in this study.


                                                  13
           Accruals quality (“AQ”) is proxy for earnings information risk. We follow

Francis et al. (2005) to measure AQ as standard deviation of residuals from the

Dechow-Dechiv model (2002) for the previous 3 years from the following regression:


                                                                                                                              3
           TCAi ,t = φ 0, j + φ1, j CFO j ,t −1 + φ 2, j CFO j ,t + φ 3 CFO j ,t +1 + φ 4 ΔREV j ,t + φ 5 PPE j ,t + σ j ,t

           We run the above regression for each year and industry according to the first

two digits of SIC code. We require at least 12 observations for each industry-year for

the regression. AQ is then calculated as standard deviation of σ j ,t for the past three

years. We then rank AQ for 10 deciles (i.e. 0-9). High ranking of AQ stands for low

accruals quality or high information risk. Any missing value of AQ is then assigned to

even higher ranking (we code it 10) because missing value indicates no earnings

quality information at all.


           In addition to information risk measured by AQ in our study, other

information may also affect investors’ interpretation to news of dividend changes. We

measure these variables as follows. 4


           1. Magnitude of dividend changes (“DDIVY”) are scaled by stock market

                price before the announcement of dividend changes;

           2. Firm size is measured by total asset (“TA”);

           3. Growth opportunities is measured by average growth rate of sales in

                previous 2 years (“GR_S”) and market-to-book ratio (“M/B”);

           4. Profitability is measured by return on asset (“ROA”);

           5. Capital structure is measured by leverage ratio (“LEV”).



3
    Please refer to the Appendix for the meaning and calculation of each variable in this regression.
4
    Please refer to Appendix for the detailed meaning and calculation methods for these control variables.

                                                             14
       Table 2 presents the descriptive statistics for all the key variables for dividend

increase groups (in Panel A) and dividend decrease groups (in Panel B) respectively.

During dividend increasing events, average 3-day CAR around announcement is

0.84%, which is significantly larger than zero; comparatively, average 3-day CAR

around dividend decrease announcement is 0.34%, also significantly larger than zero,

although p value decreased from 25.02 to 3.98. Because there are around 40% to 50%

firms get opposite market reaction to dividend changes, ABS under both dividend

changes groups are much larger than CAR in either group, average to 2.90% and

3.71% respectively.


       [Insert Table 2]


       In dividend increase group, average dividend increase is around 0.5% of the

stock price, while average dividend decrease is much larger in magnitude, at around

-1.71%, indicating that management are reluctant to decrease dividend and will wait

to decrease dividend until they have to do so.


       Mean and median values of AQ are 2.29% and 1.81% respectively for

dividend increase firms, and 2.77% and 2.15 respectively for dividend decrease firms.

Standard deviation of AQ is 1.9% and 2.31 for the two groups. Comparing the two

groups from each other, AQ are lower for dividend increase firms, showing

confirming cross-sectional evidence to Chen et al. (2007).


       Average size for dividend increasing firm is M$1700 in total assets, while for

dividend decrease firms M$1338, which suggests that larger firms are more likely to

increase dividend than to decrease dividend. Average growth rate in sales is 13.69%

and 13.13% and market to book ratio for the two groups are 2.05 and 2.17


                                          15
respectively, indicating that there growth opportunities are similar within the two

groups. Moreover, dividend increase firms normally have lower leverage ratio than

dividend decrease firms, in mean (e.g. debt ratio equal to 21.49% and 23.00% for the

two groups respectively), median, and other percentiles. If comparing past

profitability, dividend increase firms have higher return on asset (mean=8.59%) than

dividend decrease firms (mean=7.62%) in fiscal year before dividend changes.


       Table 3 tabulates correlations of the key variables in this study. The top-right

triangle shows the results from Pearson test, while the bottom-left triangle show those

from Spearman test. AQ does not have significant correlation with CAR, but on the

contrary, its correlation with ABS is significantly positive at 1%. Moreover, AQ also

demonstrates strong negative effect on magnitude of dividend changes. Size, growth

rate of sales, leverage ratio, previous profitability all suggest significant correlation

with both CAR and magnitude of dividend changes (“DDIVY”). Specially, size is

negative correlated to CAR and DDIVY; at the same time, growth rate, leverage ratio,

and profitability have positive correlation with CAR and DDIVY. Market to book

ratio is negatively correlated with CAR and DDIY, indicating that high M/B firms

tend to decrease dividend or increase smaller dividend and that market reaction to

dividend changes are more likely to be negative.


       [Insert Table 3]




V. Empirical Tests


V.1. Information risk and market reaction



                                            16
         We start by testing how information risk will affect market reaction to

dividend change announcement.


         According to the ranking of AQ, we divide the dividend increase sample and

decrease sample into 10 deciles respectively. High AQ means higher information risk.

We check market reaction (CAR) to dividend change announcement date over the 10

AQ deciles. Following previous signaling literature that dividend increase firms

disclose good news and dividend decrease firms convey bad information to the market

through dividend news, we should expect increasing trend of CAR along AQ decile

rankings in dividend increase group, and decreasing trend in dividend decrease group.


         Table 4 reports our test results on the above prediction. In dividend increase

group, generally speaking, both mean and median CAR increase along decile ranking

of AQ. For example, mean value of CAR is 0.73% in to the lowest AQ firms, 0.80%

to middle decile of AQ firms, and 1.03% to the highest AQ firms. This finding

confirms our hypothesis that high information risk firms will get stronger market

response to dividend increase information. On the contrary, however, in dividend

decrease group, neither mean nor median of CAR show clear trend along AQ decile

firms.


         [Insert Table 4]


         Because we realize that as many as more than 40% observations have opposite

sign on CAR to dividend change announcement, and because the findings in table 1

remind us that CAR of dividend decrease firms are, on average, positive, we decide

that CAR around announcement date is not a good measurement to market reaction.




                                           17
       Let us take dividend increase group as an example. On the one hand, if

dividend increase signals good news about its future cash flows, we should expect

positive CAR; on the other hand, if dividend increase conveys information on firm’s

absence of growth opportunities, market will response to such news by lowering

valuation of the firm, in another word, negative CAR. However, in either situation

dividend change news still provides investors additional source of information and

decrease information asymmetry. Therefore, stronger market reaction does not

necessarily equal to high CAR. When dividend increase means bad news to the

market, stronger market reaction means more negative CAR (i.e. lower CAR). The

same logic works for dividend decrease firms. Henth information risk works for the

absolute value of CAR, not the actual value of CAR.


       We test absolute value of CAR (“ABS”) over different AQ deciles for both

groups. This time we find more obvious increasing trend of ABS in either group. ABS

steadily increase with information risk, meaning that market reaction to dividend

change announcement is more intensively for higher information risk firms.


       To further clarify our above explanation on how information risk affect market

reaction to dividend changes, we further separate the whole sample into four

subgroups (i.e. dividend increase & CAR>0 subgroup, dividend increase & CAR<0

subgroup, dividend decrease & CAR>0 subgroup, and dividend decrease & CAR<0

subgroup) and introduce control variables, other than information risk, which are also

supposed to affect market reaction to dividend change announcement. Then we run

the following regression:


       CAR = β 0 + β 1 AQ + β 2 DDIY + β 3TA + β 4 M / B + β 5 GR _ S + β 6 LEV + β 7 ROA



                                          18
       To measure market reaction to dividend change announcement, we use CAR,

3-day cumulative abnormal returns adjusted by decile returns according to the firm’s

capitalization, as dependent variable. We want to test how the coefficient on AQ, β1 ,

changes across the four subgroups. Under our hypothesis, if information risk affect

intensity of market reaction, we predict β1 significantly larger than zero for subgroups

when CAR>0 while significantly negative for subgroups when CAR<0.


       The control variables in the regression are explained below:


   1. DDIVY: Both Aquish and Mullins (1983) and Ghosh and Woolridge (1988)

       find value of dividend changes will affect market reaction. Because magnitude

       of dividend changes may partly proxy the information content.

   2. TA: We use total assets as a proxy for firm size. Previous studies including

       Haw and Kim (1991), Mitra and Owers (1995), and Eddy and Seifert (1998),

       demonstrate that firm’s size may negatively influence magnitude market

       reaction to dividend announcement.

   3. M/B: M/B is a good indicator for growth opportunities. From Healy et al.

       (1997), firms with high growth opportunities are more likely to have negative

       (or lower) market reaction to dividend increase announcement.

   4. GR_S: For the same reason as 3.

   5. LEV: From Healy et al. (1997), high leverage firms are less likely to increase

       dividend, and therefore increase dividend do signal good information for the

       firm.

   6. ROA: As indicated in Koch and Sun (2004), if dividend changes confirm

       pervious performance, market will reflect this confirming effect.




                                          19
   Table 5a presents the regression results for the four subgroups of observations.

When dividend increase is taken as good news by the market (CAR>0), coefficient on

AQ is 5.43, significantly positive; if such increase is interpreted as bad news

(CAR<0), coefficient on AQ is -4.44, significantly negative. On the contrary, when

dividend decease is associated with positive CAR, coefficient on AQ is significantly

positive at 3.50; when dividend decrease signal bad news to the market, coefficient on

AQ is -1.57 but not significant. These findings provide persuasive evidence that

information risk strongly affected market reaction to dividend news, whether the

market understanding of the news is in the same sign with dividend changes.


   [Insert Table 5a]


   In order to minimize the effect of extreme values of the independent variables on

CAR, we use decile rankings to run the regression. Results are reported in Table 5b.

All our test conclusions are robust after we use decile rankings, proving our results

are not driven by outliers.


   [Insert Table 5b]


V.2. Information risk and Dividend Changes


       Knowing information risk of accounting earnings will affect the market needs

and reactions to dividend news, we next test whether such information risk will

influence the management choice to use dividend signaling information to the market.

In the regression, we use DDIVY as the dependent variable to measure how

management chooses the amount of dividend changes. From previous literature, we

know dividend policy should be related to the firm’s size, growth opportunities,

leverage ratio, and profitability. Therefore we put all these variables into the
                                           20
regression as control variables. Then we check whether firm’s information risk in

accounting earnings has any impact on management’s decision on the amount of

dividend changes.


       Table 6a shows the regression results for the four subgroups of observations.

When the management decides to increase dividend, they increase more when

information risk is high, no matter the market will take the signal as a good news

(CAR>0) or bad news (CAR<0). However, we do observed that the t value of

coefficient to AQ is much higher when CAR>0 (t=4.98) than when CAR<0 (t=2.49).

This finding indicates that when the management attempt to use dividend increase to

signal good news, information risk will influence on their decision on the amount of

increase more. For the other two subgroups when the management has to decrease

dividend, the amount of dividend decrease also positively related to AQ, meaning that

high AQ firms tend to use larger amount of decrease. However, we should notice that

the coefficient of AQ is not significant anymore.


       [Insert Table 6a]


       Table 6b presents the regression results when we use decile ranking values to

all the independent variable and control variables. For the two subgroups when

dividend increases, the association between decile ranking of AQ and DDIVY is still

significantly positive. For the dividend decrease subgroups, the decile ranking of AQ

signifantly affect the management choice on amount of dividend decrease, with t-

value equal to 2.63 (when CAR>0) and 2.74 (when CAR<0) respectively. The

significance of AQ is still much lower for dividend decrease subgroups than for

dividend increase subgroups however.



                                          21
       [Insert Table 6b]




V. Conclusions and Future Research

       In this paper, we propose that investors’ need for dividend payments is closely

related to the information risk on the firm’s earnings information. When investors are

highly uncertain (certain) with the firm’s earnings, information content in dividend

policy becomes to be more (less) valuable, and thus investors will response to such

dividend changes more intensively. We find empirical evidence to support this view:

the absolute value of two-day dividend announcement period cumulative abnormal

returns are more (less) related to dividend changes when information risk is high

(low). This finding helps to explain the cross-sectional variations of market reaction

to same dividend changes announcement. Moreover, we also find management’s

choice on the amount of dividend changes is significantly depend on information risk,

especially when the management attempts to use dividend increase to signal good

news to the market.


       In this paper, we have examined the cross-sectional difference for signaling

role of dividend policy for different firms. From time-series point of view, we can

predict one possible cause for the changing trend of dividend premium upon this link:

Dividend premium should be lower during the period when investors have more

confidence and provide smaller premium to information risk. During the bull (bear)

market, investors have more (less) confidence, provide larger (smaller) premium to

information risk, and accordingly pay higher (lower) dividend premium. Then the

extension of this paper can help explain why firm’s dividend is changing during

different period, as shown in Fama and French (2001) and Baker and Wurgler (2004) .

                                         22
23
Reference

Aboody, D., J. Hughes, and J. Liu 2005. Earnings quality, insider trading, and cost of
capital. Journal of Accounting Research 43 (5): 651-674

Amihud. Y., and K. Li, 2006. The declining information content of dividend
announcements and the effects of institutional holdings. Journal of Financial &
Quantitative Analysis 41 (3): 637-660.

Baker, H. K., and G. E. Powell, 1999. How corporate managers view dividend policy.
Quarterly Journal of Business and Economics 38 (2): 17-35

Baker, M., and J. Wurgler, 2004, Appearing and disappearing dividends: The link to
catering incentives. Journal of Financial Economics 73(2): 271-288

Banker, R., S. Das, and S. Datar, 1993, Complementarity of prior accounting
information: The case of stock dividend announcements. The Accounting Review 68
(1): 28-47.

Benartzi, S., R. Michaely, and R. Thaler. 1997. Do changes in dividends signal the
future or the past? The Journal of Finance 52 (3): 1007-1034.

Biddle, G., and G. Hilary, 2006. Accounting quality and firm-level capital investment.
The Accounting Review 81 (5): 963-982.

Bhattacharya, S., 1979. Imperfect information, dividend policy, and ‘the bird in the
hand’ fallacy, Bell Journal of Economics 10(1): 259-270.

Brav, A., J. Graham, and R. Michaely, 2005. Payout policy in the 21st century.
Journal of Financial Economics 77 (3): 483-527.

Brook, Y., W. Charlton, and R. Hendershott, 1998. Do firms use dividends to signal
large future cash flow increases? Financial Management 27 (3): 46-57.

Chen, S., T. Shevlin, and Y.H.Tong, 2007. Does the Pricing of Financial Reporting
Quality Change Around Dividend Changes? Journal of Accounting Research 45 (1):
1-40.

Dechow, P., and I. Dichev, 2002. The quality of accruals and earnings: the role of
accrual estimation errors. The Accounting Review 77 (supplement): 35-59.

Divecha, A., and D. Morse, 1983. Market Responses to Dividend Increases and
Changes in Payout Ratios. Journal of Financial and Quantitative Analysis 18 (2):
163-173.

Dhillon, U., and H. Johnson, 1994, The effect of dividend changes on stock and bond
prices. The Journal of Finance 39(1): 281-289.




                                         24
Docking, D. S., and P.D. Koch, 2005. Sensitivity of investor reaction to market
direction and volatility: dividend change announcements. Journal of Financial
Research 22 (1): 21-40.

Dyl, E., and R. Weigand, 1998. The information content of dividend initiations:
Additional evidence. Financial Management 27 (3): 27-35.

Easely, D., and M. O’Hara, 2004. Information and the cost of capital. The Journal of
Finance 59 (4): 1553-1583.

Eker, F., J. Francis, I., Kim, P., Olsson, and K. Schipper. 2006. The return based
representation of earnings quality. The Accounting Review 81(4): 749-780.

Fama, E., and K. French, 1993. Common risk factors in the returns on stocks and
bonds. Journal of Financial Economics 33 (1): 3-56.

Fama, E., K. French, 2001, Disappearing dividends: Changing firm characteristics or
lower propensity to pay? Journal of Financial Economics 60 (1): 3-44.

Francis, J., R. LaFond, P. Olsson, and K. Schipper. 2004. Cost of equity and earnings
attributes. The Accounting Review 79 (4): 976-1010.

Francis, J., R. LaFond, P. Olsson, and K. Schipper. 2005. The market pricing of
accruals quality. Journal of Accounting & Economics 2005 (39): 295-327.

Gorman, L., Weigand R., Zwirlein T., 2005. The information content of dividend
resumptions. University of Colorado at Colorado, Working Paper.

Grullon, G., R. Michaely and B. Swaminathan, 2002. Are dividend changes a sign of
firm maturity? Journal of Business 75 (3): 387-424.

Haw, I. and W. Kim, 1991. Firm Size and Dividend Announcement Effect. Journal
of Accounting, Auditing, and Finance 6 (3):325-347.

Healy, P. M., and K. G. Palepu, 1988. Earnings information conveyed by dividend
initiations and omissions. Journal of Financial Economics 21 (September): 149-75.

Healy, P.M., J. Hathorn, and D. Kirch, 1997, Earnings growth and the differential
information content of initial dividend announcements. Accounting Enquiries 6 (2):
187-220.

Ho., K, and C. Wu, 2001. The earnings information content of dividend initiation and
omissions. Journal of Business, Finance & Accounting 28 (7): 963-977.

Howe, J., and Y. Shen, 1998. Information associated with dividend initiations: Firm-
specific or industry-wide? Financial Management 27 (3): 17-24.

John, K., and J. Williams, 1985. Dividends, dilution, and taxes: a signaling
equilibrium. The Journal of Finance 40 (4): 1053-1070.


                                         25
Koch, P. D., and C., Shenoy, 1999. The information content of dividend and capital
structure policies. Financial Management 28 (4): 16-35.

Koch, A.S., and A.X, Sun, 2004. Dividend Changes and the Persistence of Past
Earnings. Changes. The Journal of Finance 59 (5): 2093-2116

Leuz, C, and R. Verrecchia, 2005. Firms’ capital allocation choices, information
quality, and the cost of capital. University of Pennsylvania working paper.

Li, K., and X. Zhao, 2007. Asymmetric information and dividend policy. Financial
Management, forthcoming.

Lintner, J., 1956. Distribution of income of corporations among dividends. Retained
earnings and taxes. American Economic Review (may): 97-113.

Lipson, M. L., C. P. Maquieira, and W. Megginson, 1998. Dividend initiations and
earnings surprises. Financial Management 27 (3): 36-45.

Michaely, R., R. H. Thaler, and K. L.Womack, 1995. Price reactions to dividend
initations and omissions: overreaction or drift? The Journal of Finance 50 (2): 573-
608.

Miller, M., F. Modigliani, 1961. Dividend policy, growth and the valuation of shares.
Journal of Business 34: 411-433.

Miller, M., K. Rock, 1985. Dividend policy under asymmetric information. The
Journal of Finance 40 (4): 1030-1051.

Nissim, D., and A. Ziv, 2001. Dividend changes and future profitability. The Journal
of Finance 56 (6): 2111-2133.

Ofer, A., D. Siegel, 1987. Corporate financial policy, information, and market
expectations: An empirical investigation of dividends. The Journal of Finance 42:
889-911.

O’Hara, M, 2003. Presidential address: liquidity and price recovery. The Journal of
Finance 58: 1335-1354.

Rozeff, M. 1982. Growth, beta and agency costs as determinants of dividend payout
ratios. Journal of Financial Research 5 (3): 249-259.

Short, H., H. Zhang, and K. Keasey 2002. The link between dividend policy and
institutional ownership. Journal of Corporate Finance 8: 105-112.

Skinner, D., 2004. What do dividends tell us about earnings quality? University of
Michigan, working paper.

Venkatesh, P. C., 1989. The impact of dividend initiation on the information content
of earnings announcements and returns volatility. Journal of Business 62 (2): 175-197.


                                         26
Zhang, F., 2006. Information uncertainty and stock returns. The Journal of Finance
61 (1): 105-137.




                                       27
 Table1: Frequency on dividend changes and sign of CAR, 1970-2006, 18234 obs.

                                          Div Increase                                  Div Decrease         Total
 CAR<0           # of obs.                                        6159                                1725            7884
                 sample %                                        33.81                                9.47           43.28
                 row %                                           78.12                               21.88
                 col %                                           42.50                               46.31
 CAR>0           # of obs.                                        8333                                2000           10333
                 sample %                                        45.74                               10.98           56.72
                 row %                                           80.64                               19.36
                 col %                                           57.50                               53.69
 Sum                                                          14492                                  3725            18217
 Sum%                                                            79.55                               20.45       100.00

Note:

Variable definitions:

Div Increase: Dividend increase event, if 12.5% < (quarterly cash dividend in quarter t - quarterly cash
dividend in quarter t-1) / quarterly cash dividend in quarter t-1 < 500%;

Div Decrease: Dividend decrease event, if -500% < (quarterly cash dividend in quarter t - quarterly
cash dividend in quarter t-1) / quarterly cash dividend in quarter t-1 < -12.5%;

CAR: 3-day cumulative abnormal return around announcement of dividend changes from day -1 to +1

                                                          1                         1
by a buy-and-hold strategy: CAR    j ( − 1, + 1 )   =   ∏ (1 + R
                                                        t = −1
                                                                         t   )−   ∏ (1 + MR
                                                                                  t = −1
                                                                                                 t   )




                                                              - 28 -
Table2: Descriptive statistics on key variables, 1970-2006, 18234 obs.

                                                      Panel A: Div Increase
  Variable           n          mean          std_dev     10%       25%                                    50%        75%           90%              t
 CAR               14492        0.84%           4.04% -3.45% -1.40%                                        0.54%      2.75%         5.41%           25.02
 ABS               14492        2.90%           2.94%     0.37%     0.95%                                  2.08%      3.89%         6.30%          118.78
 AQ                14492        2.29%           1.90%     0.59%     1.04%                                  1.81%      2.98%         4.47%          145.67
 DDIVY             14492        0.51%           0.48%     0.12%     0.21%                                  0.37%      0.65%         1.03%          126.80
 TA                14492       1700.30        4784.99      43.36    103.47                                 331.42    1268.11       3820.73          42.78
 M/B               14492          2.05            1.93      0.66       0.97                                  1.57       2.52          3.81         127.39
 GR_S              14492       13.69%          12.67%     1.17%     6.25%                                 11.88%     18.79%        27.94%          130.12
 LEV               14492       21.49%          16.77%     2.23%      7.69%                                18.15%     32.09%        45.64%          154.29
 ROA               14492        8.59%           4.28%     4.02%     5.76%                                  8.03%     10.77%        14.02%          241.82
                                                      Panel B: Div Decrease
  Variable            n         mean          std_dev     10%       25%                                    50%         75%           90%             t
 CAR                 3725        0.34%          5.20% -5.37% -2.24%                                        0.34%       3.06%         6.29%           3.98
 ABS                 3725        3.71%          3.66%     0.47%     1.19%                                  2.66%       5.04%         8.12%          61.87
 AQ                  3725        2.77%          2.31%     0.71%     1.24%                                  2.15%       3.57%         5.46%          73.11
 DDIVY               3725       -1.71%          2.04% -3.48% -2.17%                                       -1.25%      -0.62%        -0.32%         -51.15
 TA                  3725      1388.45        4005.95      35.36      76.28                                232.52      862.72      3108.23          21.15
 M/B                 3725          2.17           2.09      0.64       0.96                                  1.63        2.69          4.12         63.20
 GR_S                3725      13.13%          15.07% -3.01%        4.35%                                 11.48%      19.84%       30.86%           53.17
 LEV                 3725      23.00%          18.21%     2.34%     7.86%                                 19.03%      35.00%       50.50%           77.08
 ROA                 3725        7.62%          5.54%     1.18%     4.36%                                  7.48%      10.70%       14.14%           83.94


Note:

Variable definitions:

Div Increase: Dividend increase event, if 12.5% < (quarterly cash dividend in quarter t - quarterly cash
dividend in quarter t-1) / quarterly cash dividend in quarter t-1 < 500%;

Div Decrease: Dividend decrease event, if -500% < (quarterly cash dividend in quarter t - quarterly
cash dividend in quarter t-1) / quarterly cash dividend in quarter t-1 < -12.5%;

CAR: 3-day cumulative abnormal return around announcement of dividend changes from day -1 to +1
                                                                  1                           1
by a buy-and-hold strategy: CAR
                                           j ( − 1, + 1 )   =   ∏ (1 + R
                                                                t = −1
                                                                                  t   )−    ∏ (1 + MR
                                                                                            t = −1
                                                                                                             t   )


ABS: Absolute value of CAR.

AQ: Accrual’s Quality, calculated as standard deviation of                                 σ j ,t    for the past three years.   σ j ,t   is the
error term for the cross-sectional regression within each industry and each year:
TCAi ,t = φ0, j + φ1, j CFOj ,t −1 + φ2, j CFOj ,t + φ3CFOj ,t +1 + φ4 ΔREVj ,t + φ5 PPE j ,t + σ j ,t

DDIVY: Change in Dividend, calculate as ( DIVt * 4 − DIVt −1 * 4) / Pτ −1 , where DIVt is quarterly cash
dividend from CRSP (DISTCD=1232) for firm j in quarter t, and P −1 is the closing stock price in the
                                                               τ
fiscal year end before dividend changes announcement.

TA : Total Assets (Compustat#6) in the year before dividend changes announcement.




                                                                         - 29 -
M/B: Market to Book Ratio, calculate as MVτ −1 / BVτ −1 , where BVτ −1 is book value of common
shares (Compustat#60) , for firm j in the year before dividend changes announcement.

GR_S: Average Growth in Sales, calculated as   1 [( Sales − Sales ) / Sales + ( Sales − Sales ) / Sales ] ,
                                                2        τ −1    τ −2      τ −2      τ −2    τ −3      τ −3

where sales is total sales (Compustat#12) for firm j in the year before dividend changes announcement.

LEV: Leverage Ratio, calculated as Detbτ −1 / Capitalτ −1 , where Debt= short-term debt
(Compustat#9) + long-term debt (Compustat#44) respectively for firm j in the year before dividend
changes announcement, and Capital = Debt + MV.

ROA: Return on Assets, calculated as Earning / TA, where Earnings = Income before extraordinary
items (Compustat#18) for firm j in the year before dividend changes announcement.




                                                - 30 -
Table 3: Correlations of the key variables, 1970-2006, 18234 obs.

 Variable     CAR        ABS         AQ       DDIV2         TA       M/B        GR_S        LEV          ROA
 CAR                      0.29*       0.00      0.11*     -0.03*     -0.03*       0.03*        0.01       0.04*
 ABS            0.24                 0.06*     -0.07*     -0.07*     -0.07*       0.03*      0.10*       -0.06*
 AQ             0.00      0.05*                -0.07*    -0.02**      0.09*       0.02*     -0.09*        0.02*
 DDIVY         0.10*    -0.01**     -0.10*                -0.02*     -0.03*       0.08*    -0.02**        0.11*
                                                                                                              -
 TA           -0.08*     -0.14*     -0.09*     -0.17*                0.24*       -0.04*      -0.07*      0.02**
 M/B          -0.05*     -0.10*      0.10*     -0.38*        0.33*                0.06*      -0.43*       0.39*
 GR_S          0.03*      0.03*     -0.03*      0.07*       -0.11*    0.10*                    0.00       0.30*
 LEV           0.02*      0.07*     -0.10*      0.20*       -0.04*   -0.63*   -0.01***                   -0.57*
 ROA           0.03*     -0.05*       0.01       0.01       -0.11*    0.47*       0.31*      -0.64*

Note:

The upper triangle reports results on Pearson correlation tests; the lower triangle reports results on
Spearman correlation tests.

* denotes significant at 1% level, ** denotes significant at 5% level, and *** denotes significant at
10% level.

Please refer to Table 2 for the definition of the variables.




                                                   - 31 -
Table 4: CAR around dividend announcement on RAQ deciles

Dividend increasing group


 RAQ         Lowest          1         2           3         4        5        6       7       8   Highest
 N             1432       1455      1453        1451      1448     1455     1456    1452    1454     1436
 CAR
 Mean         0.73%     0.61%     0.87%       0.73%     0.81%     0.80%    0.81%   0.91%   1.10%    1.03%
 Median       0.54%     0.48%     0.57%       0.43%     0.59%     0.65%    0.56%   0.51%   0.69%    0.51%
 ABS
 Mean         2.65%     2.61%     2.78%       2.72%     2.93%     2.84%    2.87%   3.00%   3.12%    3.47%
 Median       1.97%     1.93%     2.02%       1.95%     2.15%     2.08%    2.08%   2.04%   2.23%    2.39%

Dividend decreasing group


 RAQ         Lowest          1         2           3          4       5        6       7       8   Highest
 N              356        378       373         378        369     384      375     375     377      360
 CAR
 Mean        -0.11%     0.30%     0.89%      -0.12%     0.79%      0.07%   0.22%   0.07%   0.85%    0.43%
 Median       0.16%     0.58%     0.67%       0.08%     0.54%     -0.04%   0.27%   0.25%   0.31%    0.39%
 ABS
 Mean         3.45%     3.31%     3.58%       3.58%     3.60%     3.49%    4.18%   3.74%   4.07%    4.14%
 Median       2.55%     2.32%     2.53%       2.53%     2.44%     2.70%    2.83%   2.83%   2.91%    2.89%

Note:

RAQ: Decile ranking of AQ.

Please refer to Table 2 for the definition of the variables.




                                                   - 32 -
Table 5a: Regression results, 1970-2006

CAR = β 0 + β 1 AQ + β 2 DDIY + β 3TA + β 4 M / B + β 5 GR _ S + β 6 LEV + β 7 ROA

Dividend increasing & CAR>0 (# of obs.=8333)

 Adj_RSR:       0.0513
 Variable       Intercept    AQ          DDIVY       TA       M/B      GR_S     LEV      ROA
 Estimate       0.022        0.102       1.152       0.000    0.000    0.011    0.007    -0.007
 t-value        14.890       5.429       15.405      -5.619   0.520    3.818    2.648    -0.602
 p-value        0.000        0.000       0.000       0.000    0.603    0.000    0.008    0.547

Dividend increasing & CAR<0 (# of obs.=6159)

 Adj_RSR:       0.0135
 Variable       Intercept    AQ          DDIVY       TA       M/B      GR_S     LEV      ROA
 Estimate       -0.021       -0.069      -0.258      0.000    0.000    -0.012   0.000    0.014
 t-value        -17.273      -4.438      -3.356      5.011    -1.266   -4.907   -0.091   1.519
 p-value        0.000        0.000       0.001       0.000    0.205    0.000    0.927    0.129

Dividend decreasing & CAR>0 (# of obs.=2000)

 Adj_RSR:       0.0150
 Variable       Intercept    AQ          DDIVY       TA       M/B      GR_S     LEV      ROA
 Estimate       0.034        0.128       -0.061      0.000    -0.001   0.018    0.004    -0.022
 t-value        11.265       3.570       -1.277      -1.732   -1.551   2.977    0.684    -1.074
 p-value        0.000        0.000       0.202       0.083    0.121    0.003    0.494    0.283

Dividend decreasing & CAR<0 (# of obs.=1725)

 Adj_RSR:       0.0778
 Variable       Intercept    AQ          DDIVY       TA       M/B      GR_S     LEV      ROA
 Estimate       -0.030       -0.057      0.171       0.000    0.001    0.003    -0.029   0.045
 t-value        -9.510       -1.527      4.235       0.334    2.070    0.491    -4.631   2.064
 p-value        0.000        0.127       0.000       0.738    0.039    0.623    0.000    0.039

Note:

Please refer to Table 2 for the definition of the variables




                                                   - 33 -
Table 5b: Regression results, 1970-2006

CAR = β 0 + β 1 RAQ + β 2 RDDIY + β 3 RTA + β 4 RM / B + β 5 RGR _ S + β 6 RLEV + β 7 RROA


Dividend increasing & CAR>0 (# of obs.=8333)

 Adj_RSR:       0.0566
 Variable       Intercept    RAQ         RDDIVY       RTA        RM/B      RGR_S     RLEV      RROA
 Estimate           0.033      0.001       0.001        -0.002     0.000     0.000     0.000     0.000
 t-value           17.464      5.766       8.105       -15.007    -0.782     2.656     1.926    -1.904
 p-value            0.000      0.000       0.000         0.000     0.434     0.008     0.054     0.057

Dividend increasing & CAR<0 (# of obs.=6159)

 Adj_RSR:       0.0205
 Variable       Intercept    RAQ         RDDIVY       RTA        RM/B      RGR_S     RLEV      RROA
 Estimate          -0.027      0.000       0.000       0.001       0.000     0.000     0.000    0.000
 t-value          -17.055     -2.922      -0.967       9.626      -0.985    -2.937    -0.131    2.274
 p-value            0.000      0.003       0.334       0.000       0.325     0.003     0.896    0.023

Dividend decreasing & CAR>0 (# of obs.=2000)

 Adj_RSR:       0.0376
 Variable       Intercept    RAQ         RDDIVY       RTA        RM/B      RGR_S     RLEV      RROA
 Estimate           0.049      0.001       0.000       -0.002     -0.001    0.001      0.000    -0.001
 t-value           11.805      3.470      -1.040       -6.106     -1.851    2.246     -0.832    -1.748
 p-value            0.000      0.001       0.299        0.000      0.064    0.025      0.405     0.081

Dividend decreasing & CAR<0 (# of obs.=1725)

 Adj_RSR:       0.0778
 Variable       Intercept    RAQ         RDDIVY       RTA        RM/B      RGR_S     RLEV      RROA
 Estimate          -0.048      0.000       0.002       0.000      0.001      0.000    -0.001    0.001
 t-value          -10.530     -1.016       4.835       0.938      2.323     -0.798    -1.730    1.774
 p-value            0.000      0.310       0.000       0.348      0.020      0.425     0.084    0.076

Note:

Please refer to Table 2 for the definition of the variables




                                                   - 34 -
Table 6a: Regression results, 1970-2006

 DDIVY = β 0 + β 1 AQ + β 2 TA + β 3 M / B + β 4 GR _ S + β 5 LEV + β 6 ROA

Dividend increasing & CAR>0 (# of obs.=8333)

 Adj_RSR:       0.1916
 Variable       Intercept    AQ          TA          M/B       GR_S      LEV       ROA
 Estimate           0.002      0.014       0.000      -0.001    -0.001     0.011     0.025
 t-value           11.515      4.979      -1.919     -22.730    -2.519    26.854    15.494
 p-value            0.000      0.000       0.055       0.000     0.012     0.000     0.000

Dividend increasing & CAR<0 (# of obs.=6159)

 Adj_RSR:       0.1840
 Variable       Intercept    AQ          TA          M/B       GR_S      LEV       ROA
 Estimate           0.002      0.006       0.000      -0.000    -0.001     0.010     0.020
 t-value            9.916      2.486      -3.899     -17.590    -3.240    24.753    13.445
 p-value            0.000      0.013       0.000       0.000     0.001     0.000     0.000

Dividend decreasing & CAR>0 (# of obs.=2000)

 Adj_RSR:       0.0847
 Variable       Intercept    AQ          TA          M/B       GR_S      LEV       ROA
 Estimate           0.015      0.019       0.000      -0.001    -0.011    -0.019     0.011
 t-value           10.655      1.151       1.528      -6.551    -4.047    -6.866     1.092
 p-value            0.000      0.250       0.127       0.000     0.000     0.000     0.275

Dividend decreasing & CAR<0 (# of obs.=1725)

 Adj_RSR:       0.1287
 Variable       Intercept    AQ          TA          M/B       GR_S      LEV       ROA
 Estimate           0.017      0.019       0.000      -0.002    -0.022     0.027     0.003
 t-value            8.990      0.880       2.073      -5.276    -5.721     7.526     0.212
 p-value            0.000      0.379       0.038       0.000     0.000     0.000     0.832

Note:

Please refer to Table 2 for the definition of the variables




                                                   - 35 -
Table 6b: Regression results, 1970-2006

DDIVY = β 0 + β1 RAQ + β 2 RTA + β 3 RM / B + β 4 RGR _ S + β 5 RLEV + β 6 RROA

Dividend increasing & CAR>0 (# of obs.=8333)

 Adj_RSR:       0.1230
 Variable       Intercept    RAQ         RTA         RM/B        RGR_S     RLEV      RROA
 Estimate           0.006      0.000       0.000       -0.001      0.000     0.000     0.000
 t-value           21.716     16.154      -4.935      -19.779     -3.057     5.452     6.185
 p-value            0.000      0.000       0.000        0.000      0.002     0.000     0.000

Dividend increasing & CAR<0 (# of obs.=6159)

 Adj_RSR:       0.1254
 Variable       Intercept    RAQ         RTA          RM/B       RGR_S     RLEV      RROA
 Estimate           0.006      0.000       0.000         0.000     0.000     0.000     0.000
 t-value           21.665     13.135      -4.959       -16.901    -4.895     3.919     4.513
 p-value            0.000      0.000       0.000         0.000     0.000     0.000     0.000

Dividend decreasing & CAR>0 (# of obs.=2000)

 Adj_RSR:       0.0973
 Variable       Intercept    RAQ         RTA          RM/B       RGR_S     RLEV      RROA
 Estimate           0.022      0.000       0.001       -0.000     -0.002    -0.000     0.000
 t-value           11.856      2.628       4.852       -0.621     -9.932    -2.813     0.394
 p-value            0.000      0.009       0.000        0.535      0.000     0.005     0.693

Dividend decreasing & CAR<0 (# of obs.=1725)

 Adj_RSR:       0.1162
 Variable       Intercept    RAQ         RTA         RM/B        RGR_S     RLEV      RROA
 Estimate           0.023      0.001       0.001      -0.002      -0.001     0.000    -0.000
 t-value            8.890      2.738       3.681      -6.935      -4.208     1.507    -0.371
 p-value            0.000      0.006       0.000       0.000       0.000     0.132     0.711

Note:

Please refer to Table 2 for the definition of the variables.




                                                   - 36 -
Appendix:


Definition of the Key Variables:


   1. AQ – Accrual’s Quality

       We run the following regression for each year and industry according to the first two

       digits of SIC code. We require at least 12 observations for each industry-year for the

       regression. AQ is then calculated as standard deviation of σ j ,t for the past three years.


       TCAi ,t = φ0, j + φ1, j CFO j ,t −1 + φ2, j CFO j ,t + φ3CFO j ,t +1 + φ4 ΔREV j ,t + φ5 PPE j ,t + σ j ,t

       Where:

        TCAi ,t        = Total current accruals of firm j in year t scaled by average

                            total assets. Using COMPUTSTAT Annual Industry Data

                            base to select related Balance Sheet accounting items, we

                            calculate TAC in the same way as in Sloan (1996):

                            TCA = ACCRU _ BS = {(ΔCA − ΔCash) − (ΔCL − ΔSTD) − Dep}/ Avass
                            where: ΔCA = change in total current assets (Compustat#4);

                            ΔCash =        change in cash and short-term investments

                            (Compustat#1); ΔCL = change in total current liabilities

                            (Compustat#5); ΔSTD = change in debt in current liabilities

                            (Compustat#34); Dep = depreciation and amortization

                            (Compustat#14);           and      Avass       =     average       total     assets

                            (Compustat#6).

        CFO j ,t       =    CFO j ,t = Earnings j ,t − TAC j ,t
                                                                        where Earnings = Income

                            before Extraordinary Items (Compustat#18) scaled by avass.

        ΔREV j ,t      = Change of sales (data12) of firm j in year t, scaled by average



                                                    - 37 -
                             total assets.

         PPE j ,t      = Gross Property, Plant, and Equipment of firm j in year t,

                             scaled by average total assets.

.

    2. DDIVY – Change in Dividend:

        DDIVY = ( DIVt * 4 − DIVt −1 * 4) / Pτ −1 , where DIVt is quarterly cash dividend from

       CRSP (DISTCD=1232) for firm j in quarter t, and Pτ −1 is the closing stock price by

       the fiscal year end before dividend changes announcement.

    3. TA – Total Asset:

       TA= Total Assets (Compustat#6) in the year before dividend changes announcement.

    4. M/B – Market to Book Ratio:

       M/B=         MVτ −1    /    BVτ −1    ,    where      MVτ −1 = priceτ −1       (Compustat#199)*

        # sharesτ −1 (Compustat#25), and BVτ −1 is book value of common shares

       (Compustat#60) , for firm j in the year before dividend changes announcement.

    5. GR_S – Growth in Sales

       GR_S=         1 [( Salesτ −1 − Salesτ − 2 ) / Salesτ − 2 + ( Salesτ − 2 − Salesτ −3 ) / Salesτ −3 ] ,
                      2

       where Salesτ −1 is total sales (Compustat#12) for firm j in the year before dividend

       changes announcement.

    6. LEV – Leverage Ratio

       LEV= Debtτ −1 / Capitalτ −1 , where Debtτ −1 = short-term debt (Compustat#9) + long-

       term debt (Compustat#44) respectively for firm j in the year before dividend changes

       announcement, and Capitalτ −1 = Debtτ −1 + MVτ −1 .

    7. ROA- Return on Assets

       ROA= Earnings / TA , where Earnings = Income before extraordinary items

       (Compustat#18) for firm j in the year before dividend changes announcement.



                                                  - 38 -

								
To top