Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

TAXES AND DIVIDEND POLICY

CHUNCHI WU

ABSTRACT

This paper examines the effect of differential personal taxation on corporate dividend policy. The tempo-
ral pattern of corporate dividend payout and dynamic dividend behavior are examined over a period with
significant changes in the tax code. The paper finds that there are structural shifts in the aggregate divi-
dend payout and these shifts often coincide with tax law changes. The empirical results provide a modest
support for the hypothesis that the elimination of the preferential capital gains tax treatment in 1986 has a
positive impact on the aggregate corporate dividend payout ratio.

I. INTRODUCTION

Financial theory has suggested that taxes affect corporate dividend policy (Farrar & Sel-
wyn, 1969; Brennan, 1970; Masulis & Trueman, 1988). If this theory is correct, we should
observe changes in corporate dividend payout whenever the government changes its
income tax policy. However, Miller and Scholes (1978, 1982) have shown that dividend
income could, to a large extent, be sheltered from taxation. Therefore, the effect of taxes on
corporate dividend policy depends on the extent to which individuals can elude taxes. If
individuals and institutions can successfully reduce the tax burden, the effect of taxes on
corporate dividend policy would be minimal. Previous studies (Peterson, Peterson, & Ang,
198.5; Chaplinsky & Seyhun, 1990) have shown that individuals can reduce some but not
all tax liability. Recently, Bolster and Janjigian (1991) and Papaioannou and Savarese
(1994) examine the effect of the 1986 tax reform on corporate dividend policy and report

Direct all correspondence to: Chunchi Wu, School of Management, Syracuse University, Syracuse, NY
13244.

International Review of Economics and Finance, S(3): 291-305 Copyright 0 1996 by JAI Press Inc.
ISSN: 1059-0560 All rights of reproduction in any form reserved.

291
292 CHUNCHI WU

conflicting results.’ The controversial evidence for the impact of taxes on corporate divi-
dend policy suggest the need for a further examination of this issue. In this paper, we ana-
lyze the temporal pattern of aggregate dividend behavior to shed more light on the effect of
taxes on dividend policy.
The purpose of this paper is to examine the effect of changes in capital gains and income
taxes on corporate dividend policy. We measure the extent of the tax effect on dividend
policy from two different aspects. First, we analyze the intertemporal pattern of dividends
to see whether dividend payout ratios are sensitive to changes in income taxes. In this anal-
ysis, we attempt to detect whether there is a significant shift in corporate dividend payout
surrounding the years with changes in the income tax code. Second, we examine time vari-
ations of the parameters of the dynamic dividend models to detect whether there are
structural shifts in corporate dividend behavior as a result of changes in taxes. The tax fac-
tor may affect the firm’s target dividend payout ratio and the dynamic adjustment pattern of
dividends. If so, the parameters and the structure of the dynamic dividend models may
change. An examination of the parameter and the structure of dividend models provides
some clues to the effect of taxes on dividend policy. Our analysis focuses on the aggregate
dividend behavior. The aggregate analysis allows us to assess the overall impact of taxation
on dividends while bypassing the complicated issue of information signalling.
The present study focuses on the impacts of tax reforms on the temporal pattern of cor-
porate dividend payout and dynamic dividend behavior. This approach contrasts previous
studies on dividends and taxes that emphasize individuals’ tax reduction strategies (Miller
& Scholes, 1978; Chaplinsky and Seyhun, 1990). Our analysis shed more light on the role
of taxes in the determination of corporate income distribution and dynamic dividend
adjustment. Furthermore, we study the effects of tax reforms in recent years which were
not covered by previous studies.
The remainder of this paper is organized as follows. Section 2 discusses the relationship
between taxes and dividend policy. Section 3 describes the empirical methodology while
Section 4 discusses data and empirical results. Finally, Section 5 summarizes the findings.

II. DIVIDENDS AND TAXES

Miller and Modigliani (1961) argue that dividend policy is irrelevant for the cost of capital
and the value of the firm in a world without taxes or transaction costs. They show that when
investors can create any income pattern by selling and buying shares, the expected return
required to induce them to hold firm shares will be invariant to the way the firm packages
dividend payments and new issues of stocks. Since the firm’s assets, investment opportu-
nities, expected future net cash flows and cost of capital are not affected by the choice of
dividend policy, its market value is unaffected by any change in the firm’s payout pattern.
Thus, dividend policy is irrelevant and firms can choose any payout pattern without affect-
ing their value. Miller and Modigliani’s (MM) theory implies that dividend payout will
fluctuate as a by-product of the firm’s investment and financing decisions, and therefore,
will not exhibit a systematic pattern over time.
MM’s irrelevance argument also applies to the case with corporate taxes but without per-
sonal taxes. It can be shown that the value of firms with or without debts is not affected by
Taxes and Dividend Policy 293

dividend policy if there are no transaction costs and investment decisions are independent
of dividend decisions.
The situation becomes more complicated when personal taxes are considered. In general,
dividend policy becomes relevant if investors’ dividend income tax rates are higher than
their capital gains tax rates. However, Federal tax laws are complex and individuals can
shelter their income through tax-exempt and tax-deferred investments. Also, reducing the
tax liability of dividend income often involves transaction costs. The existence of transac-
tion costs may preclude individuals from fully using tax-reduction strategies. Despite these
complications, dividends generally subject individual investors to a higher income tax rate
than capital gains do. Thus, dividend policy may affect investors’ after-tax income and
therefore, the cost of capital and the value of the firm. As Brennan (1970) indicates, for a
given risk level, investors require a higher total return on a security, the higher its prospec-
tive dividend yield is, if a higher tax rate is levied on dividends than on capital gains.
Therefore, from the standpoint of maximizing individuals’ net after-tax income, firms
should pay the lowest cash dividends to reduce shareholders’ tax burden.
If taxes are relevant to the firm’s value and cost of capital, we should observe a change
in corporate dividend policy whenever there is a change in personal income and/or capital
gains taxes. Firms and investors are expected to respond to the change in the tax code in
making their dividend and investment decisions. For instance, when there is a decrease in
the capital gains tax rate relative to the regular income tax rate, firms should reduce their
dividend payout, and vice versa.
The tax laws in the U.S. have gone through a couple of changes since 1980. In 1981,
we had the largest tax cuts in the history of the United States. Top individual tax rates
decreased from 70 to 50 percent and allowable contributions to Keoghs and other retire-
ment systems were increased. The Tax Reform Act of 1984 lowered the holding period
on capital gains from more than one year to more than six months, while the minimum
holding period for the 85% exclusion for corporate investors was extended from 16 to
46 days.
In 1986, the most sweeping revision of the tax code in the U.S. history took place. One
of the most significant changes in 1986 is the elimination of the preferential tax treatment
of long-term capital gains. An important question is whether these changes in tax laws have
affected corporate dividend policy.
To examine the effects of changes in tax laws on corporate dividend policy, we first ana-
lyze shareholders’ after-tax income under differential taxes. We assume that there are
corporate and personal debts and the tax rates for capital gains and dividends are different.
Also, for simplicity, assume that the firm’s net after-tax income is positive and the borrow-
ing rate is equal for the firm and the shareholder.2 If the firm pays out all its cash flows as
dividends, the shareholder will receive the following after-tax income $:

Yd = [(Y- iD,)( 1 - tJ - iD,](l - t&J (1)

where

Yd = the shareholder’s after-tax income if corporate income is distributed totally as


dividends
Y = the firm’s net operating income
294 CHUNCHI WU

i= the borrowing rate


tc = the corporate income tax rate
tp = the personal income tax rate
D, = corporate debt, and
Dp = personal debt.

Alternatively, if the firm pays no dividends to let the shareholder realize capital gains, then
the shareholder’s after-tax income is

F = (Y- iD,)(l - t,)(l - tg) - iD,,(l - tp) (2)

where

k’s = the shareholder’s after-tax income if the firm does not pay dividends
tg = the shareholder’s capital gains tax rate.

The shareholder’s after-tax income is greater when the firm pays no dividends, that is,
PTplyd> 1 if the capital gains tax rate is lower than the regular income tax rate.
Two types of changes in personal taxes can affect the shareholder’s after-tax income.
First, an increase in the capital gains tax rate, relative to the regular income tax rate,
reduces the ratio YRlyd. In the special case where the capital gains tax rate is equal to the
regular income tax rate, the ratio becomes one. Second, an across-the-board increase in
personal income tax rates, if the capital gains rate is a percentage (p) of the regular income
tax rate, raises the ratio Pl@. It can be shown that

a( YVf> (1 -P)W-iD,)(l-$)I >.


(3)
at, = [(Y-iD,)(1-tc,-iDpl(l-$)2

The effect of tp on Plyd is positive because any increase in the regular income tax rate
widens the difference between the capital gains and the regular income taxes. Conversely,
a decrease in the income tax rate (as in 1981) will narrow the gap between the capital gains
and regular income taxes.
A change in the capital gains tax rate relative to the regular income tax rate should affect
corporate dividend payout if the tax theory of dividends is correct. To that extent, we
should observe changes in aggregate payout ratios around the tax reform years. The past
decade provides an excellent opportunity for tracking the temporal effect of taxes on divi-
dend policy because several unprecedented tax reforms were introduced in this period. In
the following, we discuss the empirical methodology.

III. EMPIRICAL METHODOLOGY

To assess the effect of changes in taxes, we analyze the dividend payout pattern over a
period with major tax reforms. In addition, we examine firms dynamic dividend behavior
and see how they respond to tax law changes.
Taxes and Dividend Policy 295

A. /ntertempora/ Dividend Payout

A direct analysis of the tax effect is to examine the intertemporal pattern of dividend pay-
out to see whether there are significant changes in payout ratios around the years with
income tax changes. Two measures of dividend payout are used in this analysis. The first
payout measure is dividends divided by earnings (D/E). This is the type of payout ratio
usually defined in the traditional financial analysis. The second payout measure is divi-
dends divided by stock price (D/P). In this alternative measure, we assume that stock price
is equal to the intrinsic value of the firm or the present value of expected future cash flows
available to existing stockholders. Finance literature has suggested that stock price con-
tains information about the firm’s permanent earnings (Marsh & Merton, 1987; Beaver,
Lambert, & Ryan, 1987). Assuming that the required rate of return (a) for the firm is con-
stant, the firm’s permanent earnings can be set equal to aP, the perceived returns on the
intrinsic value of the firm. Thus, the D/P ratio is theoretically equal to l/a times the long-
term payout ratio (dividends divided by permanent earnings). Changes in D/P therefore
provide valuable information for changes in corporate payout policy.
To analyze the intertemporal behavior of corporate dividends, we propose the following
linear regression model:

z, = m + I_+ (4)

where

Z, = the dividend payout measure (D/E or D/P)


m = the mean of Z,

and -ot - N(0, o*) is the disturbance term. We then apply the tests of structural changes pro-
posed by Brown, Durbin and Evans (1975) to the above linear model. Brown, Durbin and
Evans (1975) show that the residuals from recursive regressions contain useful information
for detecting structural changes over time. They suggest CUSUM and CUSUM-square
tests for detecting such changes. Let X = (xt, x2, . ... x,)‘, where xP j = 1, 2, 3 ,..., IZare the
observation of the independent variables (including the intercept term) at time j and n
equals the number of observations. The orders of X and Xj are nxk and krl, respectively,
where k is the number of explanatory variables in the regression, which is equal to one in
the present case. Then, the recursive residual of the regression model based on the first r
observations is

Zr-xr’mr- 1
w, = (5)
(1 +xr’(XJ_J1x,)

where r = 2, 3, . ... n and wt - N(0, o*). The CUSUM and CUSUM-square test statistics are
defined as

w, = f i wj
j=k+l
296 CHUNCHI WU

(7)

where s is the sample estimate of the standard deviation (T.


A series of these parameter estimates can be generated by a recursive estimation procedure.
We initially estimate the model using the first k + 1 observations, and then expand the sample
to include the next observation. This procedure is repeated until all the observations in the
sample are exhausted. If taxes are indeed an important determinant of corporate dividend pol-
icy, we should observe a significant structural shift around the time of a major tax reform.
To detect such a structural shift, we construct the CUSUM and CUSUM-square test statistics.
Under the null hypothesis, the means of these two statistics are 0 and (I - k)l(n - k), where
r= k+ 1, .... n, is the number of data points included in the estimation and II is the total obser-
vations in the sample. Brown, Durbin and Evans (1975) provide cutoff values for different
statistical confidence levels. Details about the test procedure are described in the Appendix.

B. Dynamic Dividend Adjustment

Previous studies (Lintner, 1956; Fama & Babiak, 1968; Brittain, 1966; Marsh & Merton,
1987; Lee, Wu & Djarraya, 1987; Kao & Wu, 1994a, 1994b) have suggested that corporate
dividend behavior follows a dynamic adjustment process. If taxes are relevant, changes in
the tax law will affect the firm’s (desired) target payout ratio and possibly the dividend adjust-
ment pattern also. Either of these two effects can cause a shift in the structure of the dividend
adjustment model. Therefore, an alternative way for examining the tax effects is to detect
whether there is a significant change in the dynamic dividend behavior surrounding the time
with changes in the tax code.
We consider two dividend models. The first is Lintner’s partial adjustment model:

D,-D,- 1 = h(D;-D,_,)+u, (8)

where 0; is the desired dividend level, his the speed-of-adjustment coefficient and uI is the
disturbance term. The desired dividend is traditionally related to reported earnings Et. For
instance, Lintner set the desired dividend to yEl where y is the long-term payout ratio. How-
ever, reported earnings may not represent the true long-term permanent earnings of the firm.
To remedy this problem, we also use stock price as a proxy for permanent earnings and set
the desired dividend to y*P, where ‘y*equals ay, and a is the required rate of returns.
The second dividend model is suggested by Marsh and Merton ( 1987):3

log(D,+ ,/D,) = a0 + ~logWt + D,)/P,_ 1l -Glog(D,/P,_ ,I = q+ 1 (9)

where

a0 = the constant term


h = the speed-of-adjustment parameter
6 = the coefficient of error-correction that drags short-run dividends toward the long-
run payout ratio
Taxes and Dividend Policy 297

D, = the firm’s dividend at time t


P, = the firm’s stock price at time t, and
Et + 1 = the disturbance term at time t + 1.

The Marsh-Merton model imposes a long-run steady-state condition for the target dividend
payout on the short-run dividend dynamics.4
A series of parameter estimates for the dividend models can be generated by the
CUSUM recursive procedure. Similar to the analysis of intertemporal dividend payout, the
CUSUM and CUSUM-square tests of the structural changes can be applied. The test statis-
tics will provide an indication whether there is a significant structural change in the
corporate dividend behavior when the tax law changes.

IV. DATA AND EMPIRICAL RESULTS

Aggregate quarterly dividends, prices and earnings data were collected from the COM-
PUSTAT PDE tape. The COMPUSTAT tape contains quarterly dividends, prices and
earnings for major market indexes. We select S&P 500 and 400 indexes for our study. The
S&P 500 composite index includes utilities which are regulated firms adopting somewhat
different dividend policy. The composition of the S&P 500 index is usually less consistent
than that of the 400 index. We choose both indexes for comparison because the temporal
behavior of these two indexes may be different. Major changes in the tax code are reported
by Internal Revenue Service Statistics of Income. The IRS reports allow us to identify the
years of significant changes in income and capital gains taxes. The study period covers
1965-91. The selection of a longer sample period increases the degree of freedom and
reduces the sensitivity of parameter estimates to the sample size. Among the major tax law
changes since 1965, we are particularly interested in the impact of the 1986 tax reform.
Table 1 reports the results of CUSUM and CUSUM-square tests for dividend yields
(D/P). The tests are performed recursively forward and recursively backward based on
equation (4). The tests indicate that there are four shifts in the D/P ratio for the period
1965-91. These shifts occur in the years of 1970, 1978, 1982, and 1986. There were
major changes in the tax code for capital gains and income taxes in 1978 and 1986. The
Revenue Act of 1978 reduced corporate income taxes and changed tax shelter rules. The
1986 Tax Reform Act eliminated the preferential tax rate for capital gains. The rise in the
capital gains tax rate after 1986 increases the relative value of dividends to investors. As
a result, firms may have an incentive to distribute more income as dividends. The results
of recursive estimation show that there is a significant increase in the D/P ratio in the late
1986 and early 1987. Thus, there appears to be a correlation between corporate dividend
payout and income tax changes.
Table 2 reports the results for the D/E ratio. The CUSUM and CUSUM-square tests
identify five major shifts. These shifts occur in 1971, 1976, 1977, 1988, and 1989. The
shift in 1971 might be due to external shocks associated with oil crisis and price freeze. In
1977, the Tax Reduction and Simplification Act was introduced. The shifts in 1988 and
1989 could be associated with the 1986 tax law change. Although the Tax Reform Act was
introduced in 1986, the year of 1987 is a transitional period. The full implementation of
the tax reform took place in 1988. Therefore, the significant shifts (increases) in the D/E
298 CHUNCHI WU

Table 1. Structural Shifts in Temporal Dividend-Price Ratios


(D/P), = 171 + 2),

Data Tests Years and Qtrs

Forward 1978: 1
S&P 500 Index CUSUM _
Quarterly Backward 1982:2
D/P Forward 1970:3
CUSUM
(1965:1-1991:4)
SQUARE Backward 1986:3
Forward 1978:4
S&P 400 Index CUSUM
Quarterly Backward 1982:2
D/P Forward 1970:3
CUSUM
(1965:lL1991:4)
SQUARE Backward 1986:3

Table 2. Structural Shifts in Temporal Dividend-Earnings Ratios


(D/E), = m + 2),

D&2 Tests Yam and Qtrs

Forward 1977: 1
S&P 500 Index CUSUM
Quarterly Backward 1989:2
DiE Forward 1971:l
CUSUM
(1965:lL1991:4)
SQUARE Backward 1988:3

Forward 1976:2
S&P 400 Index CUSUM
Quarterly Backward 19x9:3
D/E Forward 1971:2
CUSUM
(1965:1-1991:4)
/ SQUARE Backward 1988:3
I I I

ratio for the years 1988 and 1989 may be due to the 1986 tax law change. These increases
in the D/E ratio are consistent with the tax theory of dividends which predicts an increase
in dividend payout as the capital gains tax rate rises. The increases in the D/E ratio is much
stronger than the increases in the D/P ratio after 1986. It is possible that increases in the D/
E ratio after 1986 are affected by short-term fluctuations in the reported earnings figures.
To smooth out the effect of short-term cyclical earnings fluctuations, we also performed
the tests using three-years moving averages of earnings figures. It is interesting to note
that the results based on the smoothed data exhibit a similar pattern as in Table 2.
Figures la and 1b present the plots of temporal D/E ratios estimated recursively forward
and backward for S&P 400 and S&P 500 indexes. The D/E ratios for both 500 and 400
indexes are quite stable for most of the study period. The results show a significant upward
shift in the D/E ratio after 1987, which is confirmed by both the CUSUM and CUSUM-
square tests. The results of the CUSUM estimation refect the pattern of the underlying tem-
poral D/E series. The historical dividend payout ratios for the S&P 500 (S&P 400) index
increase from about 38% (39%) in 1986 to 67% (69%) in 1989. During this period, the
CUSUM backward estimates of dividend payout ratios for the S&P 500 (S&P 400) index
increase from 55% (50%) to 60% (56%).
Taxes and Dividend Policy 299

Figure la. Plot of CUSUM Estimates of S&P 500 D/E Ratios

0.75 I P
0.7

0.65

0.6

0.55

0.5

0.45

Figure I b. Plot of CUSUM Estimates of S&P 400 D/E Ratios

0.75

0.7
I t
0.65

0.6

0.55

0.5

0.45

Figure 1. Plot of CUSUM Estimates of Dividends Payout Ratios (1965:1-1991:4)

Figures 2a and 2b show the plots of temporal D/P ratios estimated recursively forward
and backward for the 500 and 400 indexes. The results for both indexes exhibit a similar
pattern. The backward estimations indicate a short-term upward shift of the dividend pay-
out in late 1986. This upward shift is also confirmed by both the CUSUM and CUSUM-
square tests. The CUSUM estimates of dividend yields for the S&P 500 (400) index
increase from 0.0081 (0.0071) in 1986 to 0.0086 (0.0077) in 1988. It is also worth noting
that historical dividend yields exhibit variations. The mean dividend yield of the S&P 500
index is around 0.008 before 1973. The mean dividend yield increases to 0.013 between
300 CHUNCHI WU

Figure 2a. Plot of CUSUM Estimates of S&P 500 D/P Ratios

0.011
0.0105
0.01
0.0095
0.009
0.0085
0.008
0.0075

Figure 2a. Plot of CUSUM Estimates of S&P 400 D/P Ratios

Figure 2. Plot of CUSUM Estimates of Dividends Yields (1965: l-1 99 1:4)

1973 and 1982. It then drops to 0.009 after 1982. Dividend yields appear to be affected by
other economic factors.5 Nevertheless, a short-term shift in dividend yields is identified by
the CUSUM tests in 1986, which is associated with the 1986 tax refotm6
In summary, the results above suggest that there is an upward shift in the aggregate cor-
porate dividend payout after the 1986 tax reform. This shift may be attributed to the
elimination of the favorable capital gains tax treatment after 1986. Also, there is a slight
difference between the timing of shifts for the D/P and D/E ratios. The shift in the D/P ratio
occurs in 1986, while the shifts in the D/E ratio occur after 1987. This discrepancy may be
Taxes and Dividend Policy 301

Tubk 3. Structural Shifts in Dynamic Dividend Behavior


log@, + IQ) = oo + hlog[(P, + D,)lP, - 11- ~log(q& - 1) + Et, 1

D&t Tests Yeur.s and Qrrs

Forward 1978:4
S&P 500 Index CUSUM _~ ~____~
Backward -
Quarterly
Data Forward 1974:2
CUSUM
(1965:1-1991:4)
SQUARE Backward 1981:4

Forward -
S&P 400 Index CUSUM
Quarterly Backward
Data Forward 1977:2
CUSUM
(1965:1-1991:4)
SQUARE Backward 1981:3

due to the fact that stock price reacts more quickly to changes in the tax law, whereas it
takes longer time for companies to alter the dividend payout ratio. It may also possibly be
related to the fact that the full scale effect of the 1986 tax reform took place in 1988 or later.
Therefore, the time series of D/P and D/E might exhibit somewhat different patterns.
To further assess the effect of income tax changes on dividend policy, we examine the
dynamic adjustment pattern of aggregate dividends. We first estimate the Marsh and Mer-
ton model in equation (9). The coefficients in equation (9) are estimated recursively
forward and backward separately. Following this, the CUSUM and CUSUM-square tests
are applied.
Table 3 reports the results of Marsh and Merton’s dynamic dividend model. The tests for
the S&P 500 index identify shifts in 1974, 1978 and 1981, respectively. On the other hand,
the tests for the S&P 400 Index identify only shifts in 1977 and 1981. Consistent with the
findings for the D/P and D/E ratios, the results of the dynamic dividend model indicate that
there are structural shifts around 1977, 1978 and 1981. However, no major shifts are iden-
tified around 1986.
We also estimate Lintner’s partial adjustment dividend model. The Lintner model
involves fewer parameters and therefore, is less sensitive to the sample size of the recursive
estimation. Table 4 reports the results of the partial adjustment model where the desired
dividend is related to the stock price (Pt). The CUSUM and CUSUM-square tests based on
the 500 index indicate that there are structural shifts in 197 1, 1976, 1988 and 1989. These
results are generally consistent with the findings in Tables 1 and 2. Thus, there appear to be
structural shifts after the 1986 tax reform. For comparison, we further estimate the Lintner
model using reported earnings as a proxy of permanent earnings. Table 5 reports the results
of the partial adjustment model of dividends using reported earnings (Et) as an explanatory
variable. Again, the results indicate structural shifts in 1988 and 1989.
In summary, the results for the structural shifts of dynamic dividend models are mixed.
The result of the Lintner model indicates that the aggregate dividend payout ratio shifts
(increases) after 1986, whereas the result of the Marsh and Merton model does not show
such a shift. Notwithstanding this discrepancy, the structural shifts identified from the
Lintner model appear to be consistent with those identified from the time series of D/E
302 CHUNCHI WU

Table 4. Structural Shifts in Dynamic Dividend Behavior


Q-D,_, =3L(y*P,-D,_t)+u,

Data Tests Years and Qtrs


I
Forward 1976: 1
S&P 500 Index CUSUM
Quarterly Backward 1988:4
Data Forward 1971:4
CUSUM
(1965:1L1991:4)
SQUARE Backward 1989:l

Forward 1977:3
S&P 400 Index CUSUM
Quarterly Backward 1988:l
Data Forward 1971:2
CUSUM
(1965:1L1991:4)
SQUARE Backward 1988: 1

Table 5. Structural Shifts in Dynamic Dividend Behavior


Dt-Dt_l =h(yEt-Dt_l)+ut

Data Tests Years and Qtrs

Forward 1983:l
S&P 500 Index CUSUM
Quarterly Backward 1988:4
Data CUSUM Forward 1972: 1
(1965:1-1991:4)
SQUARE Backward 1989: 1

Forward 1979:l
S&P 400 Index CUSUM
Quarterly Backward 1988:3
Data Forward 1972:4
CUSUM
(1965:1-1991:4)
SQUARE Backward 1989: 1

ratios. Both results provide some evidence of changes in the aggregate dividend behavior
after 1986.

V. SUMMARY

This paper examines the effect of differential capital gains and income taxes on corporate
dividend policy. The paper analyzes the temporal pattern of dividend payout and the
dynamic behavior of dividends for the S&P firms. The analysis of temporal dividend pay-
out shows that structural shifts in the corporate dividend payout often occur around those
years when there are significant changes in the tax code. In particular, the results from the
analysis of the temporal payout pattern show an increase in the dividend payout ratio after
the 1986 tax reform. In addition, the Brown-Durbin-Evans test results for the Lintner
model show significant structural shifts after 1986. The overall results thus provide a mod-
est support for the hypothesis that the 1986 tax reform has a positive effect on the aggregate
dividend payout.
Taxesand DividendPolicy 303

APPENDIX

The procedure of the CUSUM tests is as follows. For the CUSUM test, we first compute
the recursive residuals and the CUSUM statistics:

Zr--+,_,
WI = (A.1)
(1 + ,x,_J’q
gcx;_
wj (A.2)

where r = k + 1, 3, . ... n, k is the number of independent variables in the regression, s is


the sample standard deviation. Following this, we calculate cv = a(n - k)“* + 2a(r - k)
(n - k)1’2 where a = .948 at the five percent significance level. If IW,.l> cv, we identify r
as the switching point.
For the CUSUM-square test, we first compute the test statistics:

(A.3)

Following this, we compute

CS, = MaxlS,. - (r - k)/(n - k)l (A-4)

and compare CS, with the critical value (c) at the five percent significance level. For
instance, if rr = 108, and k = 2 then the degree of freedom is l/2 (n - k) - 1 = 52 and the crit-
ical value c = .173 16. If ICS,I > c, then r is identified as the switching point.

ACKNOWLEDGMENTS

The author acknowledges the research support by the George E. Bennett Center for Tax
Research, Syracuse University. An earlier version of this paper was presented at the 1993
FMA meetings. The author is grateful to an anonymous referee and especially to the editor,
Carl Chen for very helpful comments.

NOTES

1. Bolster and Janjigian (1991) find no evidence that dividend payouts increased in
response to the 1986 tax reform. On the other hand, Papaioannou and Savarese (1994) find
that dividend payout ratio increases for the low and medium payout-ratio firm groups after
the tax reform. They argue that the lowering of the ordinary income tax rates and the elim-
CHUNCHI WU

ination of the preferential tax treatment of capital gains were associated with increases in
corporate dividend payout ratios.
2. We implicitly assume that the firm’s net operating income is identical to its eco-
nomic income. In reality, this may not be true due to biases in accounting procedures. For
instance, fixed asset values may decline due to new technological development. The
decline in asset values would be capitalized and reflected in stock price, yet the firm’s net
operating income would be unaffected. This problem could introduce noise into the stock
price and earnings processes. We thank an anonymous referee for pointing out this poten-
tial problem.
3. Equation (9) was derived under the assumptions that dividends are driven by
changes in firms’ permanent earnings and that the permanent earnings are proportional to
the stock price (intrinsic value of the firm). For the derivation of equation (9), see Marsh
and Merton (1987).
4. The value of h in equation (9) should be between zero and one. The h coefficient
captures the stylized fact of the short-run dividend dynamics that managers change divi-
dends away from the anticipated path in response to an unanticipated change in permanent
earnings. The value of y should be positive. It measures the long-run average speed of con-
vergence of the payout ratio to its steady-state target. The y coefficient captures Lintner’s
stylized fact that firms typically set a long-run target for the dividend payout and move cur-
rent dividends toward that target.
5. An examination of the historical data shows that the aggregate dividend payout ratio
fluctuates over time and appears to be affected by business cycle. However, dividend pay-
out increases substantially from 38% to 67% for the S&P portfolio between 1986 and 1989
during which there is no recession. Furthermore, although stock market crashes may affect
the value of dividend yield, the shift in the dividend yield occurs before the 1987 crash.
Thus, it is unlikely that the 1987 stock market crash is the cause of the shift in dividend
yield in 1986.
6. Dividend payout or dividend yield movements may be affected by the movements
of GNP, inflation and even lagged variables (Dimson, 1979). These complicated issues can
be better treated in a separate study.

REFERENCES

Beaver, W. H., Lambert, R. A., & Ryan, S.G. (1987). The information content of security
prices: A second look. Journal of Accounting and Economics, 9, 139-157.
Bolster, P., & Janjigian, V. (1991). Dividend policy and valuation effects of the Tax
Reform Act of 1986. National Tax Journal, 44,5 1 l-5 18.
Brennan, M. (1970). Taxes, market valuation and corporate financial policy. National Tax
Journal, 23,417427.
Brown, R. L., Durbin, J., & Evans, J. M. (1975). Techniques for testing the constancy of
regression relationships over time. Journal of the Royal Statistical Society, B37,
149-192.
Brittain, J. A. (1966). Corporate dividend policy. The Brookings Institution, Washington,
DC.
Taxes and Dividend Policy 305

Chaplinsky, S., & Seyhun, H. N. (1990). Dividends and taxes: Evidence on tax-reduction
strategies. Journal of Business, 63, 239-260.
Dimson, E. (1979). Risk measurement when shares are subject to infrequent trading.
Journal of Financial Economics, 7, 197-226.
Fama, E., & Babiak, H. (1968). Dividend policy: An empirical analysis. Journul of the
American Statistical Association, 63, 1132-l 161.
Farrar, D., & Selwyn, L. (1967). Taxes, corporate financial policy and return to investors.
National Tax Journal, 17,444-454.
Kao, C., & Wu, C. (1994a). Tests of dividend signaling using the Marsh-Merton model: a
generalized friction approach. Journal of Business, 67,45-68.
Kao, C., & Wu, C. (1994b). Rational expectations, information signaling and dividend
adjustment to permanent earnings. Review of Economics and Statistics, 76,49&502.
Lee, C. F., Wu, C., & Djarraya, M. (1987). A further empirical investigation of the
dividend adjustment process. Journal of Econometrics, .58,267-285.
Lintner, J. (1956). Distribution of income of corporations. American Economic Review, 46,
97-l 13.
Marsh, T. A., & Merton, R. C. (1987). Dividend behavior for the aggregate stock market.
Journal of Business, 60, l-40.
Masulis, R. W., & Trueman, B. (1988). Corporate investment and dividend decisions under
differential personal taxation. Journal of Financial and Quantitative Analysis, 23,
369-386.
Miller, M. H., & Modigliani, F. (1961). Dividend policy, growth, and the valuation of
shares. Journal of Business, 34,411433.
Miller. M. H., & Scholes, M.S. (1978). Dividends and taxes. Journal of Financial
Economics, 6, 333-364.
Miller, M. H., & Scholes, M.S. (1982). Dividends and taxes: Some empirical evidence.
Journal of Political Economy, 6, 1118-l 14 1.
Papaioannou, G. J., & Savarese, C. M. (1994). Corporate dividend policy response to the
Tax Reform Act of 1986. Financial Management, 23,56-63.
Peterson, P. P., Peterson, D. R., & Ang, J. S. (1985). Direct evidence on the marginal rate
of taxation on dividend income. Journal of Financial Economics, 14, 267-282.

You might also like