Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to The Review of
Economics and Statistics.
http://www.jstor.org
[ 68 ]
PB - PA a
TABLE 1.- DATA ON THE STATISTIC
D
Standard Probability Implied
Number of Deviation True Mean Is Tax
Observations Mean of Mean One or More Bracket
Raw Data 4148 .7767 .0990 .013 36.4
Data Adjusted by Index 4148 .7868 .0875 .0075 35.1
PB PA (1-to)
a
D (1 -t)
b to will be one-half of to or 25 per cent, whichever is lower. For any particular ordinary income tax rate one and only one capital gains
tax rate is appropriate. For example, in table 1 the assumption that the capital gains tax is one-half the income tax rate holds because the im-
plied ordinary tax rate is less than 50 per cent.
day and the prior day.6 We defined PB as the The presence of a bias favorable to our hy-
closing price on the day before a stock went pothesis was checked in two ways. First, the
ex-dividend. PA was measured by the closing movement of the New York Stock Exchange
price on the ex-dividend day. We used closing index during the period we calculated our sta-
rather than opening prices because all orders tistic was examined.10 As can be seen from
on the specialists' books are adjusted by the 8From the central limit theorem, the mean is normally
amount of the dividend. Consequently,the first distributed with a standard deviation equal to the standard
trade is likely to be a biased estimate of the deviation of the population divided by the square root of
the number of observations. This result is used to test the
equilibriummarket price. Since we had no way statistical significance of our statistic.
of handling this bias, we had to reject opening 'Weston and Brigham [171 made an earlier attempt to
prices.7 define stockholder tax brackets using a method analogous to
Jolivet. They found a rate of 46 per cent for the year 1965.
Table 1 lists the average value of the statistic Neither of these results are fully comparable with ours.
(PB - PA)/D for our entire sample. The proba- Both authors exclude institutional investors from their
bility that the statistic has a value of one or studies and both look at a weighted (by ownership) average
tax rate. Nevertheless, their estimates can be considered as
more (price fell by the amount of the dividend) defining a general area in which we would expect to find
marginal tax rates.
6Foreign corporations and closed end mutual funds were '0The New York Stock Exchange index is not the ideal
eliminated f-rom the sample. index for our purposes. We would like to have an index
7Problems, because of movements in the market between that weights equally a movement in any stock's price. Un-
opening and closing prices, will be discussed shortly. fortunately, the New York Stock Exchange index weights
PB - PA a
D
Probability
DIP Standard Z True Mean is Implied
Decile Mean Mean Deviation Value One or More Tax Bracket
1 .0124 .6690 .8054 .411 .341 .4974
2 .0216 .4873 .2080 2.465 .007 .6145
3 .0276 .5447 .1550 2.937 .002 .5915
4 .0328 .6246 .1216 3.087 .001 .5315
5 .0376 .7953 .1064 1.924 .027 .3398
6 .0416 .8679 .0712 1.855 .031 .2334
7 .0452 .9209 .0761 1.210 .113 .1465
8 .0496 .9054 .0691 1.369 .085 .1747
9 .0552 1.0123 .0538 .229 .591 b
PB - PA
a Spearman'srank correlation coefficient between DIP and is .9152 which is significant at the 1 per cent level.
b Indeterminate.
yield and the statistic measuring ex-dividend The implied tax rates for the last two deciles
behavior is positive and statistically significant cannot be computed. In both deciles the mar-
at the one per cent level.'4 In fact, with the ket exhibits a preference for dividends over
exception of the first and eighth decile the test capital gains, with the preferencebeing stronger
and in fact statistically significant at the 1 per
The possibility of bias arises in our rank correlation cent level for the 10th decile.'7
13
- P a
PB
D
Probability
DIE Standard Z True Mean is Implied
Decile Mean Mean Deviation Value One or More Tax Bracket
1 .204 .677 .801 .403 .3446 .4883
2 .316 .674 .197 1.655 .0485 .4945
3 .371 .756 .134 1.818 .0344 .3889
4 .409 .730 .124 2.179 .0146 .4254
5 .447 .741 .104 2.486 .0064 .4108
6 .486 .680 .108 2.959 .0015 .4848
7 .533 1.024 .125 .191 .5754 b
a Spearman'srank correlation coefficient between DIE and is 0.7939 which is significant at the 1 per cent level.
D
b Indeterminant.
might account for some of the dividend pref- seem to attract a clientele but they attract a
erence exhibited by the last two deciles, the rational clientele - one which should prefer
authors do not believe that this phenomena is their dividend policy.
widespread enough to account for the extent of
dividend preference. Part of the preference for
Summary
dividends might result from behavioralpatterns
which, while not wholly rational, have become In this paper, we have derived marginal
widely recognized over time. One behavioral stockholder tax brackets by studying the ex-
pattern that could account for part of the pref- dividend behavior of common stocks and have
erence for dividends is the propensity for the shown that these tax brackets are related to a
beneficiary of trust accounts to be restricted or firm's dividend policy. These findings . . ...
to restrict himself to dividend income from the 1. Provide data that is necessary in order to
account. Since capital gains cannot be used as calculate a firm'scost of retained earnings
income, the trust manager would prefer to re- and to test investment models.
tain the right to the dividend. This type of 2. Provide evidence in support of Modi-
behavior along with the tax advantage of divi- gliani's and Miller's clientele effect, sug-
dends for a few types of institutional investors gesting that a change in dividend policy
could account for the dividend preference ex- could cause a costly change in shareholder
hibited by the last two deciles. wealth.
Table 4 contains the same type of data that 3. Illustrate one form of market rationality
is contained in table 3 except that the analysis in that stockholdersin higher tax brackets
has been performed in terms of the firm's pay- show a preference for capital gains over
out ratio. The relationship between our test dividend income relative to those in lower
statistic and the firm's payout ratio is signifi- tax brackets.