Académique Documents
Professionnel Documents
Culture Documents
Blackwell Publishing and American Finance Association are collaborating with JSTOR to digitize, preserve
and extend access to The Journal of Finance.
http://www.jstor.org
The
VOL. XXXV
3ourna1
of
FINANCE
MARCH1980
No. 1
where:Dj,q= expected dividendper share for the j-th firm in the q-th quarterand
Dj,q= actual dividend per share announcedby the j-th firm in the q-th quarter.
Accordingly, a dividend announcement is considered favorable7if Dj,q> 1)j,q,
neutral if Dj,q= 1)j,qand unfavorable8if Dj,q<1)j,q
Justificationfor the naive expectationmodel is derivedfrom the reluctance-tochange dividendsassertation,which states that managersdo not change dividend
payments unless they have reasons to expect a significant change in the future
prospects of the firm. Hence, an increase in dividends should signal a favorable
change in managers'expectations,whereasa decreasein dividendsshould indicate
a pessimistic view of the firm's future prospects. The empiricalobservationthat
most firms follow a policy of dividend stabilizationis consistent with the reluctance-to-changedividendsassertion.In our sample,for example,about 87 percent
of all cases falls in the category of no change in quarterlydividendpayments (see
Table I). Furthermore, Laub [6] has shown that the adjustment process of
quarterlydividends is more likely to be discrete.
B. The Timing of QuarterlyDividend and Earnings Announcements
The major issue to consider is whether quarterly dividend announcements
provide informationbeyond that already provided by quarterly earnings numbers.9A major difficulty lies in the fact that quarterly earnings and dividend
figures often are released to the public at approximatelythe same time. In these
5An alternativedividendexpectationmodel was also used in this study. This is a modifiedversion
of the Lintner[8] model proposedby Fama and Babiak [3] and adjustedhere for quarterlydata:
Dj,q - Dj,q-1 = bljDj,q-1+ b2jEj,q + b3jEj,q-4 +
6j,q,
where:
Ej,q= earningsper share reportedby the j-th firm in the q-th quarter,and
Sj,q= a disturbanceterm assumedto satisfy the usual requirementsof the OLS regressionmodel.
It representsthe unexpectedchangein quarterlydividends.
This model providedresults (not reported)similarto those presentedbelow.
6 No seasonality is observedin the data with respect to changes in dividendsfrom one quarterto
another.Furthermore,Laub[6, p. 558] providesevidencethat ". . . there seems to be some additional
informationin the quarterlydividendsand earningsnot presentin the annualdividendsand earnings."
7 The cases examinedincludedividendincreasesdue to the distributionof extra cash dividendonly
if the extra dividenddiffersfromthat distributedin the same quarterof the previousyear.
8 Only decreases in regular dividends are considered.
cases, any observableadjustmentof stock prices may be the result of a confounding of the informationconveyed by earningsand dividends.
In order to isolate possible dividend effects from those of earnings,this study
examines only those quarterlydividend and earnings announcementsconveyed
to the public on differentdates within any given quarter.In this way, a distinction
is made between earnings announcementsthat precede or follow and those that
accompanydividend announcementswithin any given quarter.The sample data
frequency is presented in Table I.
The sample data are grouped accordingto the direction of dividend changes
from one quarter to another,10and by the number of trading days between
earnings and dividend announcement dates in any given quarter. The sample
includes 2612dividendannouncementsthat follow (Panel A) and 787that precede
(Panel B) quarterly earnings announcements by at least eleven trading days.
Among these are 384 increases, 47 decreases, and 2968 cases of no change in
dividends.
C. Measurement of Abnormal Performance
Assumingthat security returnsare distributedmultivariatenormal,the following market model is used:'1
Rjt = aj + fjRmt + ijt,
(2)
where:
Rj = the geometricmean'2 of the daily rates of return of securityj, over a five
consecutive trading-dayperiod, t,
=
the geometric mean of the daily rates of return of Standard and Poor's
1imt
Industrial - tock Price Index, over a five consecutive trading-dayperiod,
t,
= E(Ajt) - 8jE(Rmtt),and
disturbanceterm of securityj at period t, and E(,jt) = 0.
Using an OLS regression,equation (2) is run for each stock over the entire period
studied.'3Retum observations are excluded for fifteen days before and fifteen
days after each quarterlydividend and eamings announcementbecause if these
announcements affect stock prices, the residual terms ,jt wil not have a zero
expected value in the period surroundingeach announcementdate.
In order to derive a dividend effect that is not confounded with an earnings
aj
t=
0 All three categories exclude cases that involve stock splits or stock dividends (largerthan 10
percent) in either the currentor previousquarterin question.
See Fama [4, pp. 63-132] for a discussionof this model.
12As pointed out by Scholes and Williams[13], many securitieslisted on organizedexchangesare
tradedonly infrequently,and this introducesinto the marketmodel a potentiallyseriouseconometric
problemof errorsin variables.This nonsynchronoustradingproblemappearsparticularlysevere with
daily prices.To overcomesuch a potentialproblemwe estimatethe marketmodelusingthe geometric
means of the daily rates of returnover five consecutivetrading-dayperiods.The coefficientsobtained
(&,,/,) are then used to estimate daily residuals.(Notice that Scholes and Williams[13] suggest a
differentestimationprocedureto attack this problem.)
13Joy, et al. [5, Appendix]estimate the slope coefficientof equation (2) in altemative ways and
concludethat empiricalresults obtainedusing these estimates are similar.
Table I
Category
Increase in Dividendsa
No Change in Dividends
Decrease in Dividends
0-10
11-20
21-40
41-60
61+
11-61+
(Total)
379
2726
68
132
974
13
136
988
10
33
274
1
9
42
0
310
2278
24"
1
2
0
74
690
23c
1
13
1
1Rji -oi -
j R,,,,d,
(3)
where i denotes the i-th day relative to a given announcement date for firm j.
Then, for any day i within the interval of ten days before to ten days after an
announcement date, the average residual (AR) across sample members is
ARi
z>N
V,q _1'1Ejqi
Qq-1 Nq
(4
=
ARiStQ, .t(Q - 1),
S(eqi)
(5)
where:'4
eqi=
Ejqi,
) XQ(eqi-ARX)2.
CARK= S2K
. . . , 0, . . ., + 10):
ARi.
(6)
sions.
1 Day AD is the dividendannouncementdate when the informationis assumedto becomepublicly
availablethroughthe media. Day AD-1 is the declarationdate and in many cases the announcement
is revealedby the wire serviceson that day.
C.
B.
I.
II.
A.
I.
II.
I. No
11.
t
t
t
t
2278 t
AR
AR
AR
AR
AR
AR
Dividend
CAR
CAR
CAR
CAR
Dividend
CAR
CAR
Days
Change
tValue
(%)
Value
Value
(%)
Value
Value
(%)
Value
(%)
(%)
(%)
(%
(%
(%
cases;
*Significant
(%)
Earnings
(%)
Earnings
Earnings
(%)
Earnings
Earnings
Eamnings
in
**Significant
*Significant
at at h
at
Increase
Decrease follow
Relative
5 2.50.5690
follow
follow
precedeto
precede
precede
Dividend
cases;
AD
percent
percent
percent
Dividend
Dividend'
Dividend"
d
Dividend'
Dividend'
Dividend'
level
310
level
level
- - cases;
.20.46
d
.20
(one-tailed
(one-tailed
74 - (one-tailed
test).
.31
.26
test).
test). cases; 11
.40.90.40
.05.27.05 .151.37
.05.80.05
-1.22
.04
.55
.151.18
.15-.13
.351.74
.40 .02-1.24
.12
.171.72
.03
.04-.04
.00
- -
.08
.251.12
-04.0400
- - -
-8
.31.94.06
04 .1400
-7
.62
.47-1.78
-1.53
.2208
cases.
.72
.95-1.39
.551.29
.20
.23.86.24
1.00
1.76
.45
- .97.22
03
- -
.83.21.12
84 .71
-04-79-.451.25
.38
-2.23
-2.36
-
24 -1.01
-1.39
-70
- cases;
.9844
-1.45
'23 - -
-1.57
.18
- - -
- -
.16.87
.14
- .14.21
02-
.09-1.71
23
- - -
.29.29.02
1.03
.00 .04
-
.13.4304
.30.16.01
-1.42
07
71 -.5213-13.061.94
.19
- - - -
.31.12.01
1.09
-.07-.02
.05
- -.04.2502
63 .3608
.08.9214
.13
.441.74
67
-2.57
-2.68
.66.14.03
.14
.061.13
.06
.38-1.01
- -
.37
.67 .433.23*
1.33
3.00*
.33.73.05
-.04.8003
.36
1.69
2.29*
*
.782.35*
.35
*
.28.58.05
.01.92.03
-4.27
.67.35
.17
.36.06 .951.44
1.75
.31.52.03
.051.42
.06
.59 -4.02
.30.25
-3.85
1.00
.46
.95.00.00
1.82
.07
-
.32.14.01
.121.77
.07
- -
1.18
.35
.33.07 -3.67
-3.78
-2.23
1.41
41
-
.421.67
.10
.01.7711
.47.78.05
.02.16.01
- -
-1.08
-1.46
-5.39 -1.13 -4.62
1.97**
.95
-4.44
1.79
-1.17
14
1.27
1.03
.72
.08
- .94.6209
.12.03
-3.75
- .6424
-3.99
-4.15
-1.20
.43
.27
1.33
.06
-
.88.6606
.50.40.03
.071.18
.05
- -
.07.02
-3.70
.76.29 -4.13
-1.89
1.07
26
- -
.51.22.01
.04.7503
- -
.27.13
.24.10 -4.00
-3.60
- -
.2804
1.03
.97.77
.09
- .90.44
-1.25 -4.33
.80.33
-1.92
-4.85
.29.22
-4.46
.74.39 -4.55
-6
-5
Measures
for
-4
Days
-3
- -
.2701
.03
- - -
.80.13 .81.8309
1.16
- -
.54.59.03
00 .81.03
.81.0200
.48.71.06
.63.09
1.25
-1
AD
07-
Table
II
Surrounding
+2
+3
Announcement
+4
Dates
+5
+6
+7
+8
.01.34.01
Dividend
+1
.51.05.00
-2
-3.72
.14.05
- -
Performance
-.01.73.03
- -
-9
- -
.92.64.47
1.13
-1.90
.46
.86.13.06
-10
+9
CAR(%)
Cumulative
a.
Dividend
0.15-
Average
Abnormal Returns:
Cases Where Earnings
Precede Dividend Announcements
Decrease
CAR(%)
b.
Dividend
Announcemelnts
Increase
1.04A
-0. 37.
0.91-
-0.90
0.78.
-1.4:
0.65-
-1.96
0.52l
-2.49
0.39-
-3.02
0.26l
-3.55,
0. 13X
-4.00
0.
-4.61
'
-10 -8
-3
AD 2
-01
710
Days Relative
to the Announcement
of DIVIDENDS
-10-8
Date
-3 AD 2
7 10
Days Relative
to the Announcement
of DIVIDENDS
Date
earningsnumbers,we examine stock performancein the days surroundingearnings announcement dates (hereafter AE) in quarters where both dividend and
earnings changes provide favorable signals. For this purpose, a naive earnings
expectation model is applied to the cases included in each dividend increase
group.'6This model forecastsno change in earningsannouncedin a given quarter
from those announcedin the same quarterof the previous year. Accordingly,an
increase in earningsover those of the correspondingquarterfor the previousyear
is considereda favorablesignal and a decreasein earnings,an unfavorablesignal."7
Of the 310 dividend increase cases where earnings precede dividend announcements, 89 percent (i.e., 276 cases) were also in the earnings increase category.
Similarly, of the 74 dividend increase cases where earnings follow dividend
announcements, 86 percent (i.e., 64 cases) were also in the earnings increase
category.Results are presentedin Table III and Figure2 and are discussedbelow.
Results in Table III indicate that stockholders of companies that announced
both earnings and dividend increases in the same quarter realized, on average,
significantpositive abnormalreturns at the earningsannouncementdates (or at
AE-1) whether these earnings announcements preceded or followed the corresponding dividend increase announcements.These results, combined with evidence presented in Table II, provide further support of the hypothesis that
announcements of quarterly dividend changes provide informationbeyond that
already provided by correspondingquarterlyearningsnumbers.When dividend
increases are announced before or after earnings increases, stockholdersrealize
abnormalreturns in the days surroundingboth dividendand earningsannouncement dates.'8This indicates that the significantabnormalreturnsrealizedat the
time of the announcements of dividend changes do not reflect a diffusion or a
leakage of the informationconveyed by earningsnumbersbut rather, additional
informationgeneratedby the dividendannouncements.Thus, these findingshave
important implications for the effectiveness of using quarterly dividend and
earningsnumbersas devices for signalingmanagementexpectations,namely that
changes in quarterly dividends provide a signaling device that is at least as
effective as quarterlyearningsnumbers.
V. SUMMARY
**
II.
I.
t
276CAR AR CAR AR
Value
Value
(%) Days
(%)
Earnings
(%)
Earnings
(%)
cases;
Significant
Significant
b
follow
at at 64
Relative
precede
I 0.5
to
caes.
percent
percent
AE
Dividendh
Dividend'
level
level
.11.57.11
-
.00.02.00
-10
(one-tailed
(one-tailed
.19
.08-1.11
test).
test).
.07.56.07
-.21-.61-.13 .12.33.05
-9
-8
.09.82.12
.14
.02-1.20
-7
Performance
- - .09.02.00
.17
.151.82
-6
Cases
- - .12.19.03
.24
.391.41
-5
-4
Measures
where
for
.05.40.07
-.25
.14-1.27
.14.79.19
.23.88.09
-3
.25.85.11
.52
.751.71
-2
Table
Earnings
.872.62*
.62
.33
1.08
1.59
-1
andSurrounding
1.46
1.16
.29
.58
1.66
3.14*
AE
1.33
1.52
.36
.41.06
1.72
+1
1.55
1.78
.26
.25
1.64
1.97
+2
both
Days
III
Earnings
Dividends
Increase
- +3
1.84
.46.06
1.92
.28.05
1.96
.79.12
.32
2.24
1.72
+4
2.09
.87.13
2.33
.38.09
+5
.45.10
2.19
.35.05
2.28
2.22
.19.03
.85.10
2.18
Announcement
Dates:
- +6
- -
- -
+7
CAR(%)
a.
Cumulative
Average
Precede
Earnings
Abnormal
Returns:
Dividends
Increase
Dividend
CAR(%)
2.33
2.22'
2.08
1.95
1.82
1.68
1.56
1.41
1.30
1.14
1.03
0.86-
0.77
0.59'
0.51
0.32
0.25
0.06
Cases
b.
Where
11
Both
Earnings
Follow
Earnings
and
Dividend
0
-0.02
-0 .21
-106
Days
Relative
Date
-3
AE 2
7 10
to the Announcement
of EARNINGS
-10-4
Days
Relative
Date
-3
AE 2
710
to the Announcement
of EARNINGS
12