See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332671292

Dividend Policy, Signaling Theory: A Literature Review

Article in SSRN Electronic Journal · April 2019

CITATION READS

1 9,090

1 author:

Taleb Lotfi
Ecole Supérieure des Sciences Economiques et Commerciales de Tunis
39 PUBLICATIONS 220 CITATIONS

SEE PROFILE

All content following this page was uploaded by Taleb Lotfi on 26 April 2019.

The user has requested enhancement of the downloaded file.

 Dividend Policy, Signaling Theory: A Literature Review

Taleb Lotfi*

(*) Taleb Lotfi, ESSEC, Tunis, Tunisia.

Abstract:
With imperfect market hypothesis, it is widely accepted that announcements of dividend
payouts affect firm value. An explanation has been proposed with the cash flow signaling theory
and the dividend information content hypothesis. This original explanation, was developed in
theoretical models by Bhattacharaya (1979), John and Williams (1985) and Miller and Rock
(1985). All these authors argue that since managers possess more information about the firm's
cash flow than do individuals outside the firm and they have incentives to convey that
information to investors in order to inform the true value of the firm. This paper aims at
providing the reader with a comprehensive understanding of dividend policy by reviewing the
main theories and empirical findings under this signaling hypothesis.

Classification JEL: G35, D82

Key words: Dividend, dividend policy, signaling theory, signaling equilibrium; literature
review

1. Introduction
Financial markets exhibit asymmetric information, to reduce this information asymmetry, there
must be one or more reliable mechanism to adequately inform investors in the market so that
they can correctly evaluate the value of the firm. It is, especially through financial decisions,
that almost all information circulates, and among these financial decisions the dividend policy
occupies an important place.
In this sense, Modigliani and Miller (1961) [MoMi] admit that investors in the markets can
interpret any change in dividends as a sign of an anticipated change in profits in the market.
More recently, the theory of signals, and through structured theoretical models, has endowed
this hypothesis known as the dividend informational content hypothesis (ICH) of a theoretical
framework to explain certain aspects of the issue on dividends.
Under this signaling hypothesis, the distribution of dividends allows the managers to signal to
the market the true type of their company. Indeed, starting from the idea that there is an
information asymmetry between the initiates or the best informed, the insiders, and the
uninitiated or the badly informed, the outsiders, is such that investors can evaluate the firm
from the distribution of returns that they perceive from signals transmitted to the market.

Electronic copy available at: https://ssrn.com/abstract=3359144

Based on signaling theory, several authors [Kalay (1980), Bhattacharaya (1979), John and
Williams (1985), Miller and Rock (1985)] support the idea that dividend policy allows to
managers to signal the market the true type of he firm based on their futures perspectives. The
announcement and an increase (decrease) in the dividend level would mean that managers have
an improvement (deterioration) of the firm's prospects. As a result, the market must, normally
react to these decisions by revising the shares price.
The objective of the present article is to analyze the issue of dividend distribution under the
assumptions of signaling theory. The first section will be devoted to the presentation of the
main theoretical models of signaling by dividends. The second section will focus on the
development of the main empirical work. In the third section we discuss the issue of forecasting
share prices via dividends. The fourth section is reserved for the conclusion.

2. Dividend policies, signaling theory: theoretical models

2.1 The model of Kalay (1980)
Since the 1970s, financial theory has been enriched by several signaling models strongly
inspired by the work of Spence (1974) and Riley (1975) in order to develop a logical
explanation of the behavior of firms in terms of capital structure and dividend policy.
Kalay in 1980 proposed a model of signaling by dividends, the objective of this model, which
is also a transposition of the model of Ross (1977) applied to the problem of dividend, is the
explanation of Lintner's (1956) findings about managers' aversion to reducing the usual level
of dividends. This model implicitly assumes that if the managers attempt to emit false signals,
a penalty will be imposed on them.
To develop its model and based on certain restrictive assumptions1, Kalay (1980) assumes that
there are two periods t 0 and t1 , at the beginning of the first period ( t 0 ), the managers, who
are supposed to have more information on the future profitability of the firm, take the decisions
(choice of the dividend) and at the end of the first period ( t1 ), they pay the dividend and
liquidate the firm.
In addition, Kalay assumes managers ' compensation scheme ( M ) is assumed to be a function
of the value of the firm at the beginning of the period ( V0 ) and the cash flows generated at
the end of the period ( E1 ). This remuneration is supposed to be of the form:
M  (1  r ) 0V0   1 E1 (1)

With,

 0 and  1 represent the nonnegative weight

r is the interest rate.

In addition, Kalay assumes that there are two types of firms, good firms (type A ) and bad firms
(type B ) with respectively income Ea and Eb .

1
In particular, an economy where there are only two types of firms: the good ones and the others. In addition, it is
assumed that there are no taxes, transaction fees or agency costs. However according to Kalay, the only
imperfection is the existence of a certain information asymmetry

Electronic copy available at: https://ssrn.com/abstract=3359144

If the announced dividend ( D ) exceeds a certain critical value ( D * ), the firm is perceived by
the market as being very profitable (type A). And then its value is estimated as follows:
Ea
V0A  (2)
1 r
With E A the cash flows paid in the form of dividends by good firms (type A)

If not, the firm will be perceived by the market as unprofitable (type B) and will only be worth:

Eb
V0B   V0A (3)
1 r

With, Eb are the cash flows paid in the form of a dividend by bad firms (type B)

Suppose that a firm on the market, arbitrarily chosen, has a certain probability ( q ) of being of type A
and a probability ( 1  q ) of being of type B, and if the type of the firm is only known by the manager.
the market the two types of firms will have the same value V0 as:

qE A  (1  q ) E B
V0  (4)
1 r

V0A  V0  V0B

To arrive at a signaling equilibrium situation, Kalay assumes that it is necessary to add another
assumption regarding the executive's compensation function. This assumption takes into
account that the manager is downright at any decision to decrease the director's compensation
function dividend and alternatively if he tries to reduce this level, a penalty ( c ) will be inflicted
on him. Its compensation function2 taking into account this cost (the penalty) becomes of the
form:

 E1 si E 1  D0


M   0 (1  r )V0   1  (5)
 c
 E1  si E 1  D0
 1

With

D0 : the dividend announced on period t 0

Taking into account this remuneration function, Kalay demonstrates that there exists a
"Spencian" signalling equilibrium situation in which the firms type A will choose a dividend

2
This manager compensation function is supposed to be known by all investors.

Electronic copy available at: https://ssrn.com/abstract=3359144

such as [ E b  D *  E a ] and the market will identify all the firms that pay a lower than optimal
dividend level ( D0  D * ) as a type and all firms that pay a higher than optimal dividend (
D0  D * ) as type B.

The compensation scheme of the type A managers is as follows:

 0 E a   1 E a si D *  D0A  E a


MA 

 E   E si D 0A  D *
 0 b 1 a

As long as, E a  Eb , we have M A ( D0A  D * )  M A ( D0A  D * )

In this case the manager of the type A firms will choose a dividend level such as D0A  D *

In the same way, the compensation scheme of the type B managers is of the form:

 E   E - c si D 0B  D *  Eb
 0 a 1 b


MB  (6)

 E   E si D 0B  E b
 0 B 1 B

The manager of the type B firms who correctly report will choose whether the amount of the
penalty is greater than the profits that would result from false signals. Formally we can write:

 0 E b   1 E b   0 E a   1 Eb  c (7)

So we obtain:
c   0 ( E a  Eb ) (8)

According to Kalay, it is precisely the existence of this penalty and the managers' aversion to
lowering the level of the dividend that makes a signaling equilibrium through dividends may
exist and without an assumption about managerial reluctance to cut dividends, a signaling
(using dividends) equilibrium cannot exist.
Following the formulation of this model, Kalay proposed a testable version largely inspired by
the basic model of Lintner (1956), the objective being to verify the attitude of the manager when
the reduction of the dividend. For this, Kalay calculated 1248 variations of dividends on a
sample of 100 companies taken at random. He identified 197 reductions and distinguished the
forced cuts due to cash shortfalls and voluntary cuts. He finds that about 5% of the decreases

Electronic copy available at: https://ssrn.com/abstract=3359144

are forced. This small percentage did not allow him to reject the hypothesis that dividend
reductions convey information.3
It should be noted that, despite the theoretical interest of the Kalay model (1980), this model
has certain limits to the signaling equilibrium. The signal used by the leaders meets the
conditions set out by Spence (1974). Indeed, the optimal dividend is high enough to dissuade
bad firms from imitating good ones. For that, it is necessary that the cost of the signal, which is
in this case the penalty, is relatively important so as to force the less efficient firms not to emit
false signals.
In his proposed model, Kalay (1980) also assumes that the managers' aversion to dividend
reduction, which in fact was highlighted in 1956 by Lintner, is a necessary condition for the
signaling equilibrium. To test this hypothesis, although developed under a highly restrictive
hypothesis, Kalay's signaling model is particularly important insofar as it allows theoretical
wording to be given to dividend ICH. Nevertheless, in practice this model does not provide a
satisfactory explanation of the dividend distribution behavior of firms. Indeed, it stipulates that
the dividend conveys information only if the managers are strongly penalized when they reduce
the coupon promised. Moreover, this model remains too partial insofar as it refers to an
economy with two firms and two periods, in addition it is based on certain restrictive hypotheses
that do not make it possible to translate a more complex reality.
Other signaling models, notably the Bhattacharaya (1979) model, have also always used the
dividend as a signal and tried to explain the behavior of the firm in the context of an information
asymmetry situation. These models have changed the manager's objective and incorporated into
their initial assumptions a more realistic economic structure.

2.2 The Bhattacharaya model (1979)

Bhattacharaya's (1979) model establishes, like most dividend signaling models, that in a context
of asymmetric information the dividend can be an excellent signal for inferring the true value
of the firm to investors on the market. The difference with the model of Kalay (1980), which
implicitly assumes that the signaling by the dividends affect the remuneration of the manager
through a given penalty, is that the model of Bhattacharaya, even the penalty exists4, the penalty
is indirect. This penalty is reflected in the loss of the value that the company may suffer as a
result of a signaling by dividends.
Bhattacharaya, to develop its model, assumes that the managers are chosen by the shareholders
in order to represent their interests within the company. In these circumstances, it seems that
decisions made by managers are motivated by maximizing the wealth of current shareholders
and that, the announcement of the dividend allows investors to set the value of the firm [ V (D )

3
“We found, however, that only 5 percent of them (firms) were forced reductions. Hence, we cannot refute the
informational content of dividend”. Kalay (1980), the Journal of Financial and Quantitative Analysis, vol 15, p.
863.
4
In the Bhattacharaya (1979) model, there are two types of costs: a tax cost and an illiquidity cost. The first cost
is represented by the tax differential between dividend and capital gain, while the second is when the reported
dividends exceed the profit actually generated.

Electronic copy available at: https://ssrn.com/abstract=3359144

]. This value is determined under the assumption that there is a relationship between the
dividend and the real value of the firm.5
This new concept is based on the stability condition known as Spence-type (1974) signaling
equilibrium6 [ V ( D )  V ]. In addition, Bhattacharaya takes into account the personal taxation
of dividends at a given rate [ (1   ) ], supposed to be the tax cost of the signaling activity.
Regarding the capital gain it is supposed to be exempt from tax. A high refinancing penalty [(
 )] complements the tax cost if future cash flows are insufficient to pay the announced
dividend.
In this model, Bhattacharaya also assumes that undistributed cash flows to shareholders are
reinvested in investment projects. It assumes that the assets held by the firm have a life longer
than the investment horizon of the shareholders. Thus, according to Bhattacharaya, the present
value of the firm [ V0 ( D ) ], is supposed to be equal to:

1  
X D
V0 ( D )  V ( D )  D   (1
X  D ) f ( X )dX   (1   )( X  D ) f ( X )dX 
442443 1444424444 3
1  r  D 1 X 2 
(9)

1  
D
 V ( D )  M  (1   ) D    F ( X )dX 
1  r  X 

With,
- r is is the per period rate of interest after personal income taxes.
- M , is the mean cash flow linked to the investment project whose value is the subject of the
signal.
- V (D ) represents the value of the firm when the managers pledge to pay a dividend and respect
their commitments.
Since the investment horizon of the shareholders is one period, the current value of the firm is
equal to the flows received by the investors to which must be added the average cash flows
generated by the investment when [ X  D ], or the necessary cost refinancing the company if
[ X  D ].
~
Bhattacharaya also assumes that the random cash flow [( X )] is evenly distributed over the
interval 0, t  with an average [ t / 2 ], and that the goal of the managers is to maximize
shareholder wealth. For this they maximize the discounted value of the firm:

1 t D2 
Max V0 ( D)    V ( D )  (1   ) D    (10)
D (1  r )  2 2t 

5
This hypothesis has been the subject of numerous empirical studies [Petit (1972), Watts (1972), Aharony and
Swary (1980), Asquith and Mullins (1983) and Healy and Palepu (1988), Michaely, Thaler and Womack (1995).
and Benartzi, Michaely and Thaler (1997)].
6
Ex-ante expectations must be verified ex-post

Electronic copy available at: https://ssrn.com/abstract=3359144

The first-order condition of this maximization problem is as follows:

V0 ( D ) D*
 V ( D * )  (1   )   0 (11)
D t

D * , represents the optimal amount of dividend paid to shareholders, which is supposed to be a

function of the horizon t
Since this model is multi-periodic (n periods) but investors' investment horizon is one period,
the optimal dividend amount [( D * )] chosen at the beginning of the period (period 0) will be
identical for n future periods.7Equation (10) implies that:

D*
V ( D * )  (1   )   0 (12)
t

A stable signaling equilibrium also implies that the value signaled by the dividend at the
beginning of the first period [ V ( D * (t )) ], corresponds to the value of the firm at the end of the
period. If the cash flows generated by the company are constant and perpetual and the dividend
policy is stable, then the value of the firm can be considered as a series of cash flows discounted
to infinity as given as follow:

 
 t D * 2 (t ) 
V ( D (t ))  K  (1   ) D (t )  
* *
(13)
2 2t 
 

Where K  1 / r is the discount rate of a perpetual cash flow

By replacing V ( D * (t )) with its value found in the previous equation, we obtain the following
differential equation:

 D  D  D2 
(1  K ) (1   )     K 1 / 2   2  (14)
 t  t  2t 

Bhattacharaya also assumes that bad firms do not pay dividends to their shareholders, therefore
[ D * (0)  0 ]. Moreover, it shows that the equation [ D * (t )  At ], where A , represents the
distribution ratio and it is obtained as the solution of the following differential equation:

(1   )( K  1) (1   )  ( K  1)  K ( K  2) 
A   1  (15)
 ( K  2)   ( K  2) (1   ) 2 ( K  1) 2 

These results found proved that investors, knowing their tax rate as well as the interest rate,
deduce the value of the distribution rate and the announcement of the dividend, and they are

7
This assumption is unrealistic for several reasons. First of all, it assumes that the dividend thus defined carries
all the future information available on the periods. In addition, managers must reinvest undistributed cash-flow.
For this, they must have enough investment opportunities. Finally, for this model to be stable over time, it is
necessary for the different rates (tax, discount and refinancing) to be constant over time.

Electronic copy available at: https://ssrn.com/abstract=3359144

able to estimate the cash- anticipated flows by the firm's managers. In this way, the firm
prospects’ can be correctly evaluated by the market.

The Bhattacharaya model has all the merit of being the first theoretical work dealing with the
dividend policy as part of the signaling theory. In addition, this model is mainly explained the
distribution behavior of firms, as it allowed to present the fundamental concepts necessary for the
equilibrium of signaling. Thus, according to Allen and Michaely (2003), this signaling model has all the
merit of specifying both the strengths and weaknesses of dividend signaling models. The main strength
is that these models, in particular that of Bhattacharaya, are able to explain the positive reaction
of the market when announcing the increase of the dividend or the repurchase. This explanation
is based essentially on an intuitive notion that dividends can provide information on the current
situation or even the future of the firm (the ICH). However, the somewhat restrictive basic
assumptions have given rise to several criticisms.
The main criticism of this model concerns the economic environment described in this model
which remains too restrictive. Indeed, the author does not give a precise definition of the job-
resource balance. Nevertheless, it assumes that the firm holds cash flows, which are released
from the assets held by the company at the beginning of each period, which it allocates between
a constant dividend amount over time and between a residual used to finance investment
projects. On the other hand, if the cash flows generated by the investment are insufficient to
finance the dividend, the company will resort to debt.
The Bhattacharaya model also states that dividend policy can be used by managers to signal the
true firm quality. However, the restrictive assumptions on which the model is based appear
difficult to conceive in reality. Indeed, it is not clear that companies go into debt to pay the
promised dividend when it is higher than the cash-flow actually achieved. In addition, the cost
of illiquidity defined by Bhattacharaya is hardly feasible in reality, since it assumes that a non-
performing company borrows at a rate higher than market conditions to finance the portion of
the dividend that exceeds its ability to self-financing.
To fix all these weaknesses, recognized even by Bhattacharaya8, the author proposes in 1980 a
signaling multi-periodic dividend signaling model (n periods), where he finds that the longer
the investment horizon of the shareholder is remote the lower the required distribution rate. The
efficiency of the dividend as a signal is thus substantially increased.
Michaely and Allen (2001) have criticized the Bhattacharaya model, the main criticism is that
this model is not able to explain certain behaviors on the part of the firm and in particular to
explain why firms choose to opt for this particular type of signal (the dividend) to signal their
future prospects, knowing that there may be other less expensive tools, including the share
buyback.

2.3 The model of Miller and Rock (1985)

The two previous signaling model [Kalay (1980) and Bhattacharaya (1979)], assume that
signaling through dividends is done via a direct cost. The cost being the penalty paid by the
managers [Kalay (1980)] and the additional tax paid [Bhattacharaya (1979)]. Other models are
based on the principle of signaling by dividends via an indirect cost. The best known are those

8
“We have developed our model in terms of one-period planning horizon for shareholders. This somewhat
unsatisfactory, for the following raisons. First, in reality shareholder horizons are far longer than the time periods
over which corporations can change their dividend. Second, as consequence, the low response of V (D ) to D
appear to be unrealistic”. Bhattacharaya (1979), the Bell Journal of Economics, p.267.

Electronic copy available at: https://ssrn.com/abstract=3359144

of Miller and Rock (1985); John and Williams (1985), Ambarish, John and Williams (1987)
and Allen, Bernardo and Welch (2000).
As part of the Miller and Rock (1985) model, the signaling activity, like the other models
previously developed, is based on the dividend to signal to investors the value of the current
profits.

In this study, Miller and Rock show that the equilibrium obtained assumes that the firm rejects
projects with a net positive present value and it is the opportunity cost that arises that constitutes
the cost of signaling (indirect cost).
Miller and Rock observe that under certain hypotheses, a signaling equilibrium is created since
the managers do not have interest to report falsely.
To develop their model Miller and Rock have described an economic environment that tends to
be closer to reality. Indeed, they assume that the firm holds at the end of the first period ( t1 )
resources that it allocates at the beginning of the second period ( t 2 ) between the investment (
I1 ) necessary for the second period and the dividend that it pays at the beginning of the second
period ( t 2 ), so that the employment-resource equality can be written in the following way:

X 1  B1  I 1  D1 9 (16)

With, B1 corresponds to the company's external sources of financing available in date t1

At the beginning of the first period, the firm invests an initial amount ( I 0 ) in a production
process whose distribution function [ F ( I 0 ) 10] is supposed to be known by all investors. At the
end of the period, the firm obtains an income ( X 1 ) corresponding to the invested funds
increased by a random term (  1 ), supposed to be object of the signal. The income obtained,
assumed to be a random income, at the end of the first period is defined by Miller and Rock as
follows:
~
X 1  F ( I 0 )  ~1 (17)

In the same way:

~
X 2  F(I1 )  ~2 (18)

With,

E ( ~1 )  E ( ~ 2 )  0 ; E ( 2 /  1 )   1 (19)

9
According to Miller and Rock the profit declared at the end of the period to the value of invested funds increased
~
by an error term, so that X 1  F (I 0 )  1

F ( I )  0 ; F ( 0)  0 ; F   0 ; F   0
10
The investment / production function must meet certain characteristics:

Electronic copy available at: https://ssrn.com/abstract=3359144

The term  represents an assumed coefficient of persistence11. At the end of the second period
( t 2 ), the realized income ( X 2 ) is distributed among the shareholders, then the firm is liquidated.
Managers choose a dividend amount that maximizes maximize current shareholder income.
Knowing that this dividend must signal the real value of income ( X 1 ), Miller and Rock assume
that investors know the value of the adjustment factor.
At the end of the first period ( t1 ) the value of the firm attached coupon, is equal to the expected
profits in the second period ( t 2 ) discounted at the rate of return required by the shareholders (
i ), increased by the dividend and decreased by external financing:
1
V1  D1  F ( I1 )   1   B1
1 i
(20)
1
 D1  B1  F ( I1 )   1 
1 i

Miller and Rock also assume that the firm's managers may be incentivized to set an investment
level below the optimal level12 in order to favor a temporary overvaluation of the share price.
They include in their model the two categories of investors and they assume that a fixed number
( k ) of shareholders will try to sell their securities after the announcement and payment of
dividends just before the publication of the results. They trade their shares at a market price (
V1d ), depending on the information held by investors.13
According to who owns the information, Miller and Rock express the quality of information
held by executives (  d ) and that held by investors in the market (  m ) as follows:

 d  X 1 , I 1 , D1   I 0 ,  1 , I 1 , D1  (21)

 m  I 0 , D1  (22)

Based on the information they hold, the value of the attached coupon firm, as estimated by the managers,
would be:
1
V1d  D1  F ( I 1 )   1  (23)
1 i

While investors, who do not have all the information, for them the value of the firm will be
expressed as follows:

11
The coefficient  is a coefficient of inertia which is equal to 0 if the random element of the first period is only
transient, it is equal to 1 if this element is permanent. This coefficient can take any value greater than 1, less than
zero or between 1 and 0
12
This is possible because Miller and Rock assume that the dividend announcement is positively correlated with
the stock price. As a result, an increase in the dividend results in an increase in the market value of the firm.
13
Miller and Rock assume that there is information asymmetry because executives know the amount of
investments made at the beginning of the period, profits and investments made at the end of the period while
investors only know the amount of dividends and external financing at the end of the period.

10

Electronic copy available at: https://ssrn.com/abstract=3359144

 V1m  D1 
1
1 i

E1m ( F ( I 1 )  ~1 /  m  (24)

The problem for managers is to maximize the wealth of current shareholders by equitably
considering both groups of shareholders. The choice of the level of the dividend ( D ) and that
of the investment ( I ) that meets this objective must be proportional to the share of the interests
of each shareholder group:

max W1  kV1m  (1  k )V1d

D1 , I1

(25)
  
 k D1 
1
E1m ( F ( I )  ~1 ) m   (1  k )  D1 
1
F ( I 1 )   1 
 1 i   1 i 

Under budget constraint: D1  I 1  X 1

For investors, the market value of the share is a function of the announced dividend (
V1  V1 ( D) ). On the other hand for managers, it is also function of the profit ( X ) that they are
the only ones to know. Their goal then becomes:

max W ( X ; D,V m ( D))  kV m ( D)  (1  k )V d ( X , D) (26)

D

Thus, for each level of profit, there is a dividend amount that maximizes the objective function
above. If is the relationship that binds the profit of the firm to the dividend is unique, and if the
shareholders are rational, a stable equilibrium can be defined only when:

V m (D)  V d X (D), D  V d ( X , D) (27)

By replacing V (D ) with, V d C ( D), D , at the signaling equilibrium, the firm must choose the
amount of the dividend ( D ) such as:

V d
X ( D), D X ( D)  k V X ( D), D   (1  k ) V ( X , D)  0
d d
k (28)
X D D

This differential equation describes the optimal path of the values of. However, to reach a
maximum, it is necessary that the condition of second order is respected and which is written
in the following way:
V d
 ( X , D ) X ( D )  0 (29)
DX

This found result is one of the first fundamental results of Miller and Rock's (1985) model.
It should be noted that the main objective of the Miller and Rock study was to highlight a new
concept in signaling equilibrium, the concept of indirect cost of signaling and this, through the
policy of underinvestment. This result can be mitigated if we assume that managers have the

11

Electronic copy available at: https://ssrn.com/abstract=3359144

opportunity to issue new securities, and thereby increase cash flows. In this case, the under-
investment obtained by Miller and Rock at equilibrium is no longer guaranteed. However, this
issue has a capital dilution effect that must be taken into account.
The model of Miller and Rock (1985) brings new concepts in the resolution of signal equilibria.
In particular, by highlighting the fact that information asymmetries, which characterize the
financial markets, compel firms to modify their behavior in terms of investment policy. But it
should be noted that the Miller and Rock model remain unable to analyze the impact of future
cash flow variability on dividend levels, mainly because the Miller and Rock model, as well as
the Bhattacharaya model, is limited to one period and two dates and that the firm is liquidated
at the end of the period. It is therefore not necessary for the firm to be concerned about being
able to maintain its dividend for future periods.
To overcome these restrictions, Lapointe (1995) proposed a dividend signaling model in which
reputation plays a crucial role, whereas Williams (1988) considered that the random term
introduced into the Bhattacharaya model (1979) is inappropriate for that he introduced a
multiplicative random term.

3. Dividend policy, signaling theory: Empirical work

The idea that a change in dividends is associated with a change in stock returns has been largely
studied theoretically. This idea, which supports the existence of a certain informative content
of the dividends, was supported in 1961 by Modigliani and Miller neutrality thesis, which and,
without rejecting their thesis of neutrality, suppose that the variation observed at the Stock
prices when announcing a change in dividends is mainly due to some informative content, but
not to dividends as a cash flow.
The basic idea behind this assumption of the informative content is that the variation of the
dividend makes it possible to communicate to the market some information on the real quality
of the firm, so that any decision as to the distribution or the retention of the dividends will have
an impact on investor attitudes and therefore on stock prices.
This notion of the existence of a certain informative content of the dividends has been
formalized by several models. These models, which developed in the previous sections,
recommend that the dividend can have two aspects: a first aspect where the dividend can be
used as an ex-ante signal to inform on the future value of cash flows released by the firm
[Bhattacharaya (1979)]; and a second aspect assumes that dividends can convey information
on current profits. [Miller and Rock (1985) and John and Williams (1985)].
Empirically, several studies have been developed to test this hypothesis of the informational
content of dividends advocated by theoretical models. Most of this empirical study, and as Allen
and Michaely (2001) point out, has attempted to test the following implications:
(i) A change in dividends must be followed by a change in the same direction of profits;
(ii) An unanticipated change in dividends must be accompanied by a similar change in stock
prices;
(iii) An unanticipated change in dividends must be followed by a review of investor
expectations and in the same sense, as the change in dividends on the future value of returns.
It is important to note that all of these implications are necessary but not sufficient conditions
to validate signaling models by dividends [Allen and Michaely (2001)]. The condition that a
change in profits must result in a change in dividends is the most important condition. If ever
this condition is not verified, it may lead us to conclude that the dividend variable does not

12

Electronic copy available at: https://ssrn.com/abstract=3359144

actually have any informative content on current income and therefore, cannot be used as a
signal.
Most of the empirical literature dealing with the problem of dividends under the signaling
theory is focused on the second implication, that is, an unanticipated change in dividends is
associated with a change in the same sense. But that does not prevent that, several other study
have also tried to verify the hypothesis that the dividends have a certain informative content on
the future stock prices (implication 3).
Fama et al., (1969) studied the impact of the announcement of stock splits on stock prices. The
average residual yield was calculated using the market model using monthly data on a 60-month
interval around the distribution announcement date for a total of 940 distributions between
January 1927 and December 1959. Based on the results found, abnormal yields are observed
only before and not after dividend distribution. This result suggests that distribution may be the
main cause of observed abnormal return.
To test this dividend ICH, the sample selected by Fama et al., (1969) was subdivided into two
main categories of firms: (1) those that increase their dividend levels in the period following
the stock splits and, (2) those that set a low distribution rate. The result found proved that the
share price rise considerably just after the announcement of an increase in dividends. This result
asserts the assumption that the decision of a distribution is interpreted by the market as a certain
message informing about the increase of the level of the dividends. The abnormal returns and
positive signs observed following the announcement reflect a certain price adjustment that will
take place when the market is really sure of the increase in the level of distributions. On the
other hand, the average abnormal yield observed following a low distribution rate decreases
during the year following the dividend announcement period. During this period of time
investors realize that their expectations of an increase in the dividend rate have not been
realized. When combining the result of an increase and a decrease in dividends, the result found
is consistent with the assumption that the market reports unbiased expectations about the level
of dividends and that these signals are reflected in the share price one month after the
distribution decision.
Pettit (1972) is one of the first authors to study the reaction of stock prices when announcing a
dividend distribution. On the basis of a study carried out over the period 1967-1969 concerning
several variants of profit distributions (increase, cat and decrease) and using the market model,
Pettit manages to demonstrate that an abnormal performance of the order of -6.22% is observed
for the case of a catering and 2.02% in the case of an increase while, -2.24% in the case of a
reduction. These results found by Pettit confirm the idea that dividend change is a good signal
(good news) and that the reduction and omission are rather seen as an unfavorable signal (bad
news).
Based on the results found, Pettit concludes that the announcement of dividends provides useful
information and shows that the market is responding positively to the announcement of an
increase in dividends and negatively in the announcement of a decline. However, if Pettit
confirms that the announcement has an impact on the current value of the stock price, this
announcement does not have significant informational power for forecasting stock prices.
Pettit (1976), in other studies using a sample of 634 firms studied during the period 1964-1968
and based on the previous of Watts ' study , attempted to verify the results already found in his
previous study but, more to check the informational content of profits and anticipated dividends.
The empirical model developed by Pettit is of the form:

Dt  Dt  Dt 1   1 Dt 1   2 E t   3 E t 1   t (30)

13

Electronic copy available at: https://ssrn.com/abstract=3359144

With, Dt and E t designate respectively the dividend per share and the earnings per share of the
year t .
From the results found, once again, Pettit asserted the results already found in his previous study
and beyond confirming the idea that the informational content of the dividends is not negligible,
in addition, this study has confirmed another previous result that stipulates that the
informational content of dividends is significantly larger than that of profits.14
Aharony and Swary (1980), adopting an approach very similar to that of Pettit (1972) and by
using the naive model applied to a sample of firms that changed their dividends by more than
10% between year 1963 and 1976, suggest that the announcement of dividends and profits are
not perfect substitutes, and appropriate tests of the signaling hypothesis require consideration
of the possible concomitant effects on profits and dividends. After controlling the earnings
announcement effect, Aharony and Swary (1980) find a result similar to that of Pettit (1972).
In fact, the stock prices of firms, that did not change their dividends, did not change
significantly. On the other hand, for the firms having increased their dividends, an abnormal
positive profitability (+ 0.72%) was observed the day of the announcement and the day
preceding. While the firms having lowered their distribution rate experienced an abnormal
negative profitability of the order of -3.76%, as well on the day of the announcement as the day
before.
The highlighting of these reactions allows Aharony and Swary (1980) to confirm that the
reaction of the prices can only be due to the announcement of the dividends, and confirm the
hypothesis of dividend ICH advanced on the theoretical level.15
Aharony and Swary (1980) have also tried to test the effects of concomitant announcements of
profits and dividends. From the results found, the authors conclude that the announcement of a
dividend increase positively changes the stock market prices, whether it is the announcement
of the dividends is carried out before that of the profits or after.
Woolridge (1983), based on a sample of firms observed during the period 1974-1976, also tried
to observe unexpected increases and decreases in dividends in order to test whether stock
market prices react to this kind of financial information. Using the market model for calculating
unexpected returns, Woolridge finds that prices react significantly upward (down) following
the announcement of an increase (decrease) in dividends. An abnormal positive return of about
1.44% is observed in the case of an increase and a negative return of the order of -5.59% in the
case of a decrease. These results lead the author to confirm the idea of the informative content
of the dividends and then to validate the ICH.
Focusing on extreme cases observed in dividend changes, Asquith and Mullins (1983)
attempted to study the impact of dividend initiation on a sample of 168 firms observed over the
period 1964-1980. To do this, the authors examined the abnormal returns observed over a ten-
day window around the dividend announcement date. Over the two days following the
announcement, an abnormal return of 4.7% is observed. Asquith and Mullins, in addition to this
event study, used a cross-sectional regression that highlighted a positive and statistically

14
“Dividend announcement, when forthcoming, may convey significantly more information implicit in an earnings
announcement” Pettit (1972). Journal of business, Vol 27, p.1002.
15
“These findings of capital market reaction to dividend announcement strongly support the information content
of the dividend hypothesis, namely that changes in quarterly cash dividends do provide information about changes
in management’s assessment of future prospects of the firm”. Aharony and Swary (1980), the Journal of Finance,
Vol 35, p. 8.

14

Electronic copy available at: https://ssrn.com/abstract=3359144

significant relationship between the observed variation in dividends and the calculated
abnormal return.
Kane, Lee and Marcus (1984), based on 352 observations of earnings and dividend
announcements during the period 1979-1980, find that the reaction of stock prices is positive
when the event is an announcement of a dividend increase, even if the declared profits are below
those anticipated. While the reaction of stock prices is negative when the announcement is a
decline and even if the earnings announced is higher than expected. This result found leads the
authors to confirm that the announcement of dividends cannot be without impact on stock
prices.
In the French context, Dumontier (1985), based on the risk-adjusted returns model, found that
the market reacted significantly when a dividend decline was announced. The fall in prices was
in order of 2.05%. It should be noted that this same study did not highlight a price reaction
when announcing an increase in dividends.
In another study, largely inspired by that of Fama et al (1969), Grinbatt and Titman (1984) used
daily data and observed investor returns as at the date of the dividend announcement, than
during the previous period. These authors looked at a particular sample where there are no other
types of announcements made in the three-day period around the announcement date and chose
firms for which a distribution decision did not been pronounced for three years. For this sample
composed of 125 listed financial securities, the authors observe a statistically significant
relationship about 3.44% around the date of announcement. This result has been interpreted by
Grinbatt and Titman as a favorable signal giving information on the future cash flows of the
firm.
Another study, similar to that of Grinbatt and Titman (1984), Grinbatt, Masulis and Titman
(1984) confirmed the results of the previous studies done by Woolridge (1983) and prove that
the effect of the dividend announcement depends on the size of the sample chosen. Indeed, the
effect is larger and it is around 4.9% for a sample of 382 dividend announcements, while it is
significantly lower for a sample of 84 distribution announcements.
Michaely, Thaler and Womack (1995), through the observation of 561 cases of dividend
initiation and 887 cases of omissions over the period 1964-1988, show that the market reacts
severely. The excess abnormal return found is of the order of 3.4% in case of initiation and the
order of 7% in case of omission. These results make the market reaction for the case of omission
much more intense than that of initiation. Michaely et al. (1995) thus assume that the market
has an asymmetric16 response for changes of different dividend announcement [increase,
decrease, initiation, omission]. This implies that the low level of dividends (decrease) has some
more important informative content than a higher level of dividends (increase). Michaely et al
(1995) attribute this market reaction to the fact that the reduction is an event that is not very
frequent or, because the reductions are of a greater magnitude.
To verify this asymmetric market reaction, Michaely et al. (1995) examined this line of research
and showed that the effect of a unitary change has a much greater impact on stock prices in case
of omission, than in initiation. This explains one of the conclusions of Lintner which supposes
that the managers are very reluctant during the decrease (or absence) of distribution and they
try most of the time to smooth their dividend level.
Elfakhani (1995) conducted an empirical study that took into account other parameters when testing the
dividend signaling models, especially all the events that existed around the announcement date of a

“Over, the results show that the market reaction to dividend change is significantly related to the magnitude of
the change ( p 593); .the long term reaction to omission announcement is greater than to initiation announcement
( p606) ” Michaely, Thaler and Womack (1995), the Journal of Finance, vol 50.

15

Electronic copy available at: https://ssrn.com/abstract=3359144

dividend change. Indeed, according to Elfakhani, these events can have repercussions on the size and
direction of the change in share prices as a response to the announcement of the dividend change.
The study by Elfakhani (1995) is based in a world characterized by information asymmetry and
supposed the existence of two phases: the first phase transmits information through financial documents
(capital investment, capital structure and agency problems), during this first phase, information will be
communicated mainly through financial instruments and investors will try to evaluate the clarity of this
dedicated information (either it is clear or ambiguous). In the second phase, managers will try
to use other informational sources, the dividend, to be notified a change in the dividend level.
This level can either confirm or invalidate the good or bad news already advanced in the first
phase and so that the information on the situation of the firm would be clearly identified in this
second phase.
According to Elfakhani, the announcement of dividends can be ambiguous and unclear and
does not have significant signaling power, if so, it explains the limited role of the dividend as a
signal. But if the dividend plays its full role as a signal, the response of stock prices to a dividend
announcement should be determined primarily by three factors: the unanticipated content of the
dividend change, the sign of the change in the level of dividends and the signaling role of the
dividend.
If the market is able to interpret the financial records, and in particular the balance sheet in a
very simple way, a certain consensus on the value of the firm would be achieved. In this case
the investors will try to react on the basis of their appreciations of the change and the signal
emitted by the dividends would be a signal of confirmatory type.
On the other hand, changes in the balance sheet may contain several attributes in the sense that
there may be both good and bad news, in which case the balance sheet interpretation would not
be very clear. The signal that would be issued by the dividends is intended to clarify. Whereas
if the signal emitted fails to be explained and does not dispel uncertainty in the minds of
investors, according to Elfakhani, the role of the dividend as a signal would not be clear and in
this case the dividend does not have a certain informational content or that its informational
effect has no impact on the value of stock prices.
To verify each of the roles that the dividend can play as a signal, Elfakhani and on the basis of
a set of hypotheses and adopting the methodology of the events studies, tried to test the two
phases of transmissions of the information and its impact on the value of stock prices. From the
results found, it appears that among the three roles attached to the dividend (confirmatory,
clarifying or ambiguous) the role of clarification is the most important, the market response to
this type of dividend being in fact the most intense. It should also be noted that the confirmatory
dividend does not have any particular interest for the market, as long as it does not allow the
addition of other news or unanticipated information.
Elfakhani, in his study also examined the role, and thus the market reaction, following several
types of dividend announcements (maintain, increase or decrease). Another surprising result
was found from this study and that is all to contradictory to the predictions of the signaling
theory is that the market responds positively to the announcement of a decline in the dividend,
and the decline in this case is necessarily synonymous with bad news.
Dewenter and Warter (1998) have tried to compare the signaling power by dividends in two
different contexts, that of American firms (420 firms) and that of Japanese firms (194 firms)17.

17
“Due to the institutional differences in the structure of corporate ownership and the nature of corporate group
interaction, we assume that Japanese firms are subject to less information asymmetry and fewer agency conflict
than U.S. firms”. Dewenter & Warter (1998), the Journal of Finance, vol 53, p. 902.

16

Electronic copy available at: https://ssrn.com/abstract=3359144

From the empirical results found, the authors prove that the influence of dividends, as a signaling
mechanism, is much lower in Japan. Indeed an abnormal return is observed, in the case of an omission
as well the day of the announcement of dividend than the day which follows. This abnormal return is of
the order of -2.53% for Japanese firms and -4.89% for American firms. While, when the event is an
initiation the abnormal return found is of the order of 0.03% and 2.28% respectively for the Japanese
and American firms. In addition, the Dewenter and Warter study showed an abnormal return of -6.48%
and -17.03% respectively for Japanese and American firms. This return, and contrary to that calculated
previously, was observed on a window of 62 days before the event date (an omission) and of the order
of 0.1% and 10.24% for the case of an initiation. All these results found lead Dewenter and Warter to
conclude first, that Japanese firms are less exposed to information asymmetry which is mainly due to a
system of governance and a different property structure between the two countries. In addition, the
authors demonstrate that information asymmetry and agency conflicts significantly affect the firm's
dividend policy.
Van Eaton (1999) examined abnormal stock market returns over a 3-year horizon around
dividend change announcements during the period from 1973 to 1990. The results are consistent
with the assumption of the dividend ICH. Indeed, abnormal and statistically significant returns
have been observed after the announcement of a dividend reduction or omission.
Benartzi, Michaely and Thaler (1997) also find an average abnormal return of around 8.6% in
the year preceding the dividend increase and -28% for firms that reduce their dividends.
Kao and Wu (1994), following the formulation of an extension of the Marsh and Merton model
(1987), examined the informative content of dividends. The proposed model is an adjustment
model that refers to signaling theory. In contrast to Watts' studies, the empirical results show a
positive relationship between unanticipated dividends and unanticipated profits.
Grullon Michaely and Swaminathan (2002), by advancing the maturity hypothesis18, observed
a very large number of firms (6284 increases and 1358 decreases) over the period 1967-1993.
The results found highlight a significant abnormal yield due to a dividend increase of around
1.34% and a significant market reaction of around -3.71%.
Amihud and Murgia (1997), based on a sample of 200 firms listed on the Frankfurt Stock
Exchange find results to confirm the dividend ICH. The authors examined the reaction of share
prices for several types of events (255 announcements of increase, 51 announcements of
decrease) during the period 1988-1992, the results found highlight a significant abnormal return
of the order of 0.965 when announcing an increase and -1.73 when the announcement is a
decrease in the dividend level.
Best and Best (2001), based on a sample of 6189 dividend increases and 330 decreases studied
over the period 1977-1998, found a statistically significant abnormal return about 0.6068 and -
3.6773 respectively in the case of an increase and decrease in dividends.
Using the Fama and French three-factor model, Grullon, Michaely, and Swaminathan (2002)
find an abnormal return of around 8.3% in the three years following the year of the increase,
but they did not detect abnormal performance for firms that have reduced their dividends.
Amihud and Murgia (1997) examined the dividend policy of German companies, where
dividends are taxed less than capital gains. In this context, and based on the signaling models
in the presence of signaling costs as adopted by John and Williams (1995) and Allen, Bernardo

18
The maturity hypothesis assumes that dividends convey information about the change in the phase in which the
firm operates (from a growth phase to a maturity phase). They consider that the increase in the level of dividends
is indicative that the firm has reached a state of maturity.

17

Electronic copy available at: https://ssrn.com/abstract=3359144

and Welch (2000), Amihud and Murgia find no informational power derived from a change in
the dividends of German companies. Indeed no reaction of the prices was observed around a
variation of the dividends. However, despite this disappointing result for the signaling hypothesis,
Amihud and Murgia (1997) find that the change in dividends in Germany generates a financial market
reaction that is similar to other empirical studies in the US context.
In the Japanese context, Fukuda (2000) finds that firms' performance tends to decrease
following a drop in distribution rates, it also highlights a positive (negative) reaction to share
prices following an announcement of increases (decrease). In the same context, Harada and
Nguyen (2005) based on a sample of 13708 observations studied over the period 1992-2001
and divided into five categories of dividend announcements (increase, decrease, initiation,
omission and no change) and by using the Logit model regression, find that an improvement in
performance is observed following an increase in dividends (+ 0.6%) whereas in a case of
decrease a negative performance is founded (-0.18%).

4. Dividend signaling and stock price forecasting

Theoretically, the ICH is initially developed by Lintner (1956) who assumes that firms increase
their dividend rate only when the company's managers estimate a certain increase in the
projected profits. MoMi (1961) also, suggest that dividends can convey information about the
future value of cash flows, but this is only possible when markets are incomplete. Bhattacharaya
(1979), Miller and Rock (1985), John and Williams (1985), in their information asymmetry
models, also find that dividends can signal the future value of profits.
Watts (1973) was the first to test the proposition that the current dividend level can be used as
a predictor of stock prices. To verify this hypothesis, Watts used a sample of 310 firms observed
over the period 1945-1968, the test conducted consists in verifying in which circumstances the
profit of the following year ( t  1 ) can be explained both, by the profit and the dividend of the
current year ( t ) and the previous year ( t  1). The model developed by Watts is of the form:

E i ,t 1   i   1 E i ,t   2 ,i E i ,t 1   3,i Di ,t   4 Di ,t 1   it (31)

Where, Ei ,t and Di ,t denote respectively the income and the dividend of the company i realized
during the period t.
Based on the results found, Watts demonstrates that there is a positive relationship between
current income and the current dividend, except that this relationship is not statistically
significant19. From this result, it appears that even if a relationship exists between the dividend
and the profit it is not very important. But it should also be noted that if a relationship between
profit and dividend was not found, the results of this same study show that the anticipated
income depends closely on the current income.
Watts expanded his tests to study the relationship between the change in dividends and profits
of the previous year. The model, as estimated by Watts by the OLS method, largely inspired by
Lintner's basic model and Fama and Babiak (1968), shows the existence of a positive
relationship between unanticipated change in dividends and change in earnings, but this

19
“The preliminary test of the information hypothesis suggest that while the relationship between current dividends
and future earnings implied by the hypothesis might exist, it probably is not very strong”. Watts (1973), the journal
of business, vol 46(2), p.198.

18

Electronic copy available at: https://ssrn.com/abstract=3359144

relationship is not statistically significant. Of all these results found, Watts concludes that the
informational content of dividends is not very important to predict future earnings.20
In a very relevant study, and unlike most previous studies conducted, Benartzi et al., used to
test the information content hypothesis of the dividends of a very large number of firms (7186
observations) studied during the period 1979- 1991. In this study Benartzi et al. (1997) studied
in particular the relationship between the observed variation of dividends and future earnings.
The model developed by Benartzi et al. (1997) is as follows:

E i , t Divi , 0 Divi ,0
   1   2 I i,0   Dummy i , 0   it (32)
Pt , 1 Divi , 1 Divi , 1

E i , t Divi ,0 Div i , 0
   1   2 I i,0  X it 1   Dummy i , 0   it (33)
Pt , 1 Divi , 1 Div i , 1

In the study by Benartzi et al., (1997) two relevant results were found: (1) First, there is a
correlation relationship between dividend change and benefit change. Indeed, according to the
results found, when there is an increase in the level of dividends, a similar variation has been
observed in terms of past profits. (2) But a similar relationship between the change in dividends
and future earnings has not been demonstrated through this same study. Indeed, in the two years
following the increase in dividends, the change in profits is uncorrelated neither to the sign nor
to the magnitude of the change in dividends.21
De Angelo, De Angelo and Skinner (1996), through a study conducted on growth, dividend
policy and its signaling power with external investors, have also tried to empirically validate
the hypothesis of the informative content of dividends and to test the predictive power through
dividends, as supported previously by MoMi (1961), Bhattacharaya (1979), Miller and Rock
(1985) and John and Williams (1985). Their test is essentially based on the decision taken
during a reference year, which is supposed to have some informative content for investors, and
to ensure that this information content conveyed through the dividends is transient or
permanent. The tests were conducted on 145 firms on the New York Stock Exchange with the
following characteristics: they experienced a growth in their annual profits for nine consecutive
years followed by a year of declining annual profits. The year of decline in annual profits (years
0) represents the year of transition from a period of positive growth to a period of zero growth.
De Angelo et al., (1996) were particularly interested in the dividend policy for the reference
year (year 0), since at this point in the decline in annual profits, investors should be more
interested in the forecasts of the managers of the company on growth opportunities.

20
“However, all of the tests also suggest that the average absolute size of the future earning changes which might
be conveyed by unexpected dividend changes is very small”. Watts (1973), the journal of business, vol 46(2), p.
211.

21
“Consistent with the earlier finding of Watts (1973), we are unable to find any evidence to support the view that
changes in dividends have information content about future earning while there is a strong past and current link
between earning and dividend changes, the predictive value of changes in dividends seems minimal. Indeed, the
only strong predictive power we can find is that dividend cuts reliably, signal an increase in future earning …”
Benartzi et al., (1997), the Journal of Finance, p. 1031.

19

Electronic copy available at: https://ssrn.com/abstract=3359144

Three main models were used to test the signaling effect by dividends: (1) the first model
measures the future profits of companies that increased their dividends in the year 0; (2) the
second model checks whether dividends are used by managers to differentiate themselves and
achieve an equilibrium, (3) the third model tests the dividend effect on the return of securities
at the base year (year 0) in a long-term relationship. In this study of De Angelo et al., (1996),
firms were divided into three categories, those that increase the dividends in year 0, those that
decrease them and the one that leaves constant the level of dividends. The results found show
that most firms (99 companies) increased their dividends to the reference year (year 0),
suggesting that managers do not see this reference year as a good strategic time to reduce
dividends, a decline could decrease the price of the shares. Of the 99 firms in the study that
increased their dividends in year 0, 67 firms increased this level one year before. The managers
of these companies, despite a decline in profits, seem at least as confident of their companies'
prospects in year 0 as in year -1.
To test the hypothesis that an increase in dividends is considered good news, De Angelo et al.,
(1996) conducted tests on the 99 firms that increased their dividends in year 0. Using two
different models, the first is a random walk, the second is the growth adjustment model. The
results found show that for the random walk model, the benefits of the years 1, 2 and 3 are not
very different from the profits of the reference year (year 0) which proves that the market does
not react significantly. For the growth adjustment model, the abnormal profits in years 1, 2 and
3 are negative, which implies that the market interprets this message as bad news.
To test the separating equilibrium, De Angelo et al. (1996) performed a cross-sectional analysis
of dividends in the reference year, the dividend being used by managers to differentiate
themselves from other firms in the same situation but with weaker future growth prospects.
Thus, if dividends are used by firms to differentiate, firms that increase dividends should have
higher abnormal future profits.
Four regressions were performed with the dependent variable as the abnormal benefits while
the explanatory variables are the benefits of year 0 and year -1 in addition to four dummy
variables:
i. A dummy variable equal to 1 if the firm increases the dividends to year 0 and equal to 0
otherwise;
ii. A dummy variable equal to 1 if the change in dividends in year 0 is greater than the change
in dividends in year -1 and equal to 0 otherwise;
iii. A variable representing the percentage change in dividends at the base year (year 0);
iv. A variable representing the difference between the percentage change of dividends in
year 0 and year -1.
Based on the results found, none of the coefficients are statistically significant, so the separating
equilibrium hypothesis cannot hold. In addition the coefficients on the variables representing
the dividends and those of the variables dummy, are very close to zero. These results are
contradictory to what is normally predicted by the separating equilibrium hypothesis, which
assumes that firms that increase their dividends in year 0 are the ones for whom normally the
future profits are higher.
To test the signaling effect of dividends, De Angelo et al. (1996) also observed the payment of
dividends in years subsequent to the reference year (year 0), namely years 1, 2 and 3. So from
the full sample, the results found show that there is about the same number of firms that increase
their dividends at least three times or more than they do that do not increase them at all. Most

20

Electronic copy available at: https://ssrn.com/abstract=3359144

firms that increase dividends three or more times increased in year 0, and two-thirds of firms
that did not increase dividends in year 0 did not increase year 1 to 3.
To test again whether dividends can be considered a good signaling tool, De Angelo et al.,
(1996) repeated the previous regressions only that abnormal profits are calculated with year 1
as a reference year. In these regressions, De Angelo et al., (1996) compared the abnormal
returns of firms that increased their dividend in year 0 and year 1 with the abnormal profits of
all firms with the abnormal profits of firms that did not increase their dividends, neither the year 0
nor the year 1. From the results, it appears that no significant difference was found, beyond the
authors conclude that the increase in dividends provides no information about profits future.
These results are similar to those of Benartzi et al., (1997).
To test the impact of dividends on stock prices, De Angelo et al. (1996) subdivided the horizon
into two parts, a short-term horizon (the day of the announcement of the dividend increase) and
other long term. In the short term the securities associated with a rise in dividends have a better
return. In fact, a positive and low abnormal return was found (0.66% in year 0 and 0.55% in
year 1).In the long run, it appears that firms that increased their dividend in year 0 have
cumulative abnormal returns that are less negative than those that did not increase their dividend
(-10.17% versus -22.03%).
Nissim and Ziv (2001) have attempted to study the relationship between a change in future
profitability, measured in terms of expected profit and abnormal profit. This study supports the
hypothesis of the existence of an information content of dividends. In this study, Nissim and
Ziv find that dividends lead to information about the level of profitability for subsequent years.
They also find that the change in the level of dividends is related to the change in profits for the
two years following the change in the dividend.
The studies of Nissim and Ziv, based on the methodology adopted by Benartzi et al., examined
the correlation between the rate of change of dividend per share in a reference year, the year 0,
corrected by the market value of shares. This assumption makes the profits follow a random
relationship and in such a way that the change in profit makes it possible to measure any
unanticipated change in the profitability of the firm. To check the results found by Benartzi et
al, Nissim and Ziv (2001) used the following regression:

(Et  Et1) / P1  0 1RDIV0 t (34)

With, E t refers to the profit of the current year t ; P1 the market value of the shares of the firm
at the beginning of the year in which there was a change in the dividend; RDIV0 represents
the rate of change per share of the dividend in year 0.
The results of the OLS estimates found by Nissim and Ziv (2001) confirm those of Benartzi et
al., (1997). Indeed, the coefficient of the variable that reflects the change in dividends (  1 ) is
positive and is statistically significant for year 0 but it is not significant for years and 2.
Nissim and Ziv continued their study and assume that the model specification can make the
coefficients of negative sign for years 1 and 2, this may be due to model specification errors,
especially the fact that the variable dependent is strongly correlated with any change in dividend
or the omission of a significant control variable that is correlated with the change in the
dividend.
To try to take into account model specification errors, Nissim and Ziv tried to specify the
measurement error in the independent variable, afterwards they tried to add other variables to

21

Electronic copy available at: https://ssrn.com/abstract=3359144

improve the specification of the model. In another formulation, Nissim and Ziv consider that
the price-to-earnings ratio ( ROE ), may contain a certain informative attribute regarding future
profits. Thus this variable has been added as an additional variable in their base model, so that the
estimated model is of the form:

(Et  Et 1 ) / B1  0 1 RDIV0 2 ROEt 1   t (35)

This equation is estimated for years 1 and 2, the ratio ROEt 1 is measured by the ratio [
E t 1 / Bt 1 ] between the profit of the year t  1 ( E t 1 ) and the book value of the shares ( Bt 1 ).

Nissim and Ziv used two types of regressions: the first is an estimation by the OLS while in the
second the authors tried to take into account both the heteroskedasticity and the autocorrelation
between the variables [Fama and MacBteh (1973)].
The results found by Nissim and Ziv show that for years t  1 and t  2 , the coefficient of the
variable reflecting the change in dividends (  1 ) is positive and is statistically significant and
that the coefficient of the return on equity (  2 ) is also statistically significant but it is of
negative sign. These results indicate that the change in dividends has some information on
future earnings for the two years following the year in which there is a change in the level of
dividends.
The results found also prove that the variation observed in terms of dividends is strongly
correlated with the variation observed in the profit level of the current year. As a result, the
positive relationship between the dividend and the earnings change for the next two years may
be due to a possible correlation in the series of earnings changes.
To examine whether the change in dividends informs about changes in future earnings, Nissim
and Ziv added another control variable, namely the relationship between the observed variation
between future earnings (in year 1) and the profit realized during the current year and the book
value of the shares [ ( E 0  E1 ) / B1 ]. Adding this new variable, the model is presented as
follows:
( E t  E t 1 ) / B 1   0   1 p DPC 0  RDIV 0   1n DNC 0  RDIV 0   2 ROE t 1   3 ( E 0  E 1 ) / B 1   t (36)

This model is estimated for years 1 and 2, given that variables DPC and DNC are dummy
variables which take the value 1 for the case of a dividend increase (decrease) and the zero
value otherwise.
The results of the estimation of this model prove that for year 1, the coefficient for the case of
dividend increases and decreases are both significant and of positive sign. But the coefficient
of the dividend increase is greater than that of the decrease. While for year 2, the coefficient of
the dividend increase remains positive and statistically significant, but the coefficient of the
decrease is almost equal to zero.
Nissim and Ziv tried to examine the relationship between the dividend change and the profit
level for the five years following the year in which there was a change in the dividend. For this
purpose the authors used two other alternative measures of profit: normal profit and abnormal
profit. Normal profit measures the return allocated to each share while the abnormal return is
measured by the difference between the normal and the return required by the shareholders
given a known cost of capital. The model tested for the case of an ordinary profit is the
following:

22

Electronic copy available at: https://ssrn.com/abstract=3359144

 E t   0t   1t DPC 0  DIV0   2t DNC 0  DIV0   3t E 1   4t B1   5t P1
(37)
 6t DIV1   7t DPC 1  DIV1   8t DNC 1  DIV1   9t E 0   t

While the model tested for the case of an abnormal profit is as follows:

AE t   0t   1t DPC 0  DIV0   2t DNC 0  DIV0   3t AE 1   4t B1   5t P1

(38)
 6t DIV1   7t DPC 1  DIV1   8t DNC 1  DIV1   9t AE 0   t

These two models have been estimated for years t  1,2,...,5 , with the year t , is the year in
which there was a change in the dividend and DIV0 is the change in the dividend.

The results of the estimation of this model show that an increase in dividends is positively
related to the income of the four years following the year of the dividend change, and that a
decrease in dividends is not related to future income. The absence of a correlation between the
dividend cut and future income does not necessarily mean that the dividend cut has no
information content for future income. The informative content of this decline can be captured
by the income of the current year.
Of all the results found, Nissim and Ziv conclude that the future profitability of the firm is
related to the variation (and the magnitude of the variation) observed in terms of dividends and
that the reaction of the market differs depending on whether it is an announcement of increase
or decrease of dividend (asymmetrical reaction). Only in the event of a dividend increase does
an improvement in performance occur during the four years following the announcement of an
increase, but no abnormal profitability is observed in the case of an announcement dividend
decline.22
Ofer and Siegel (1987) used a sample of 781 observed changes in dividends to examine how
financial analysts are changing their current earnings guidance as a response to the change in
dividends. Ofer and Siegel find that analysts react to the change in dividends and revise their
forecasts by a certain amount that is positively correlated with the size of the dividend change.
In addition, they pointed out that the revision of the forecasts is positively correlated with the
market's reaction to the announcement of the dividend.
Healy and Palepu (1988), through a sample of 131 companies that practice dividend initiation,
find that profits are growing very fast in previous years and continue to grow for at least the
next two years. However, in their sample of 172 companies that did not perform a distribution,
the result is totally contradictory to what the signaling theory assumes. In fact, the profits
decrease during the year of the omission, but increase very significantly during the following
years.
Fukuda (2000), on the Japanese context and based on a sample of 223 companies, tested the
dividend signaling hypothesis and the relationship between the dividend announcement and the

22
“We document that, after controlling the expected change in future earnings, dividend changes are positively
related to earning changes of two years following the dividend change. We also show that dividend changes are
positively related to the level of future profitability … the findings are not symmetric for dividend increases and
decreases. For full sample, dividend increases are associated with future profitability for at least four years after
the dividend change, while dividend decreases are not related to future profitability after controlling for current
and expected profitability”. Nissim & Ziv (2001), journal of Finance, vol 56, p.2131.

23

Electronic copy available at: https://ssrn.com/abstract=3359144

current and future value of stock prices. The results found show that the coefficient of the
dividend changes is positive, but it is not statistically significant, the authors conclude that the ICH is
not verifiable on the Japanese context.
Grullon, Michaely and Swaminathan (2002) find a result similar to that of Benartzi et al (1997).
The signaling model developed highlights a relationship between the change in the level of
dividends and the change in the characteristics of the firm, particularly its risk class. Using a
sample of companies that change their dividend by more than 10%, Grullon et al. (2002) show
that, not only the profits in subsequent years do not increase, but above all that, the level of
profitability of firms decreases in the year following the announcement of the increase.
Benarzti, Grullon, Michaely, and Thaler (2002), using the partial adjustment model of Fama
and French (2000) to try to control the predictable portion of future earnings and based on past
profits and changes that have taken place in previous years, re-examined the relationship
between the level of dividends and the change in profits. The results found confirm that the
observed changes in dividends do not contain any informative content concerning future profits.
Indeed the coefficient relative to the variation of the dividends is statistically insignificant when
the dependent variables are the changes of the profits of year 1 and year 2.

5. Conclusion
In this article we have tried to present a review of the literature dealing with the problem of
dividends under the hypothesis of signal theory. Inspired by the work of Spence (1974) and
Riley (1975), several theoretical models of dividend signaling are developed [Bhattacharaya
(1979), Miller and Rock (1985), John and Williams (1985) and Allen Bernardo and Welch
(2000)], all the theoretical models have tried to formalize the idea that the dividends have a
certain informative content that is not insignificant to inform both the current and future
profitability of the firm.
The dividend signaling models, which have all the merit of providing a new framework for
studying the issue of dividends, are based on obvious intuitions: firms that generally increase
their dividends are companies that are undervalued by the market and vice versa. As a result,
most of these models show that the dividend can be a vehicle for transmitting good information
about the quality of the firm and in particular its current and future profitability. Empirically,
several studies have attempted to test this hypothesis of the informational content of dividends.
From this empirical work it follows that:
i. Empirical evidence does not allow to fully validate the theoretical models of signaling by
dividends;
ii. When announcing a dividend change, this change usually leads to a similar change in
share prices. Thus, any increase (decrease) in the dividend induces positive (negative)
abnormal returns and this variation is perceived by the market as good (bad) news;
. iii. The importance of the reaction of share prices depends on the importance (the
magnitude) of the changes observed in dividends;
iv. The reaction of the market when announcing a change in dividends is not symmetrical in
case of increase and decrease. The announcement of a reduction has a greater impact than
the announcement of a decline.

Bibliography

24

Electronic copy available at: https://ssrn.com/abstract=3359144

[1] Aharony J. and Swary I., (1980) “Quarterly Dividend and Earning Announcements and
Stockholder’s Returns” The Journal of Finance, 35, 1-12. https://doi.org/10.1111/j.1540-
6261.1980.tb03466.x
[2] Akerlof G., (1970) “The market for Lemon: Quality uncertainty and the market Mechanism”
Quarterly Journal of Economics 89, 488-500. https://doi.org/10.2307/1879431
[3] Allen F. Bernardo A. and Welch I., (2000), “A theory of dividend based on tax clientele” The
Journal of Finance 55 (6), 2499-2536. https://doi.org/10.1111/0022-1082.00298
[4] Allen F. and Michaely R., (2001) “Payout Policy” Wharton school Working Paper.
https://doi.org/10.2139/ssrn.309589
[5] Ambarish R., John K. and Williams J., (1987) “Efficient signalling with dividends and
investments” The Journal of Finance, 42, 321-343. https://doi.org/10.1111/j.1540-
6261.1987.tb02570.x
[6] Amihud Y. and Murgia M., (1997) “Dividends, Taxes and Signaling Evidence from Germany”
The Journal of Finance, 52, 397-408. https://doi.org/10.1111/j.1540-6261.1997.tb03822.x
[7] Asquith P. and Mullins D., (1983) “The impact of initiating dividend payment on shareholder’s
wealth” The Journal of Business, 56, 77-96. https://doi.org/10.1086/296187
[9] Bagwell L., Simon L. and Shoven J., (1989) “Cash Distributions to Shareholders,” Journal of
Economic Perspectives, 47, 129-140. https://doi.org/10.1257/jep.3.3.129
[10] Ball R, Brown Ph., Finn F.J. and Officer R.R., (1979) “Dividend and The Value of The firm:
Evidence from the Australian Equity Market” Australian Journal of Management, 4, 13-26.
https://doi.org/10.1177/031289627900400102
[11] Ball R., Brown P., (1968) “An Empirical Evaluation of Accounting Numbers” The Journal of
Accounting Research, 6, 159-178. https://doi.org/10.2307/2490232
[12] Barclay M.J., (1987) “Dividends, Taxes and Common Stock Price: The Ex-Dividend Day
Behavior of Common Stock Prices Before the Income Tax” The Journal of Financial Economics,
19, 31-44. https://doi.org/10.1016/0304-405x(87)90027-4
[13] Bar-Yosef S. and Huffman C., (1988) “The Information Content of Dividends: A signalling
approach” The Journal of Finance and Quantitative Analysis, 21, 47-58.
https://doi.org/10.2307/2330990
[14] Bar-Yosef S. and Kolodny R., (1976) “Dividend policy and capital market theory” The Review
of Economics and Statistics, Vol 58, 181-190. https://doi.org/10.2307/1924024
[15] Beaver W., (1968) “The Information Content of Annual Earnings Announcements” The Journal
of Accounting Research, 6, 67-92. https://doi.org/10.2307/2490070
[16] Benartzi S., Michaely R. and Thaler R., (1997) “Do Changes in Dividends signal the Future or
the Past” The Journal of Finance, 52, 1007-1043. https://doi.org/10.2307/2329514
[17] Bernheim D. and Wantz A., (1995) “A Taxed-Based Test of Dividend Signaling Hypothesis” The
American Economics Review 85, 532-551. https://doi.org/10.3386/w4244
[18] Bernheim D., (1991) “Tax policy and the Dividend Puzzle” Rand Journal of Economics 2, 455-
476. https://doi.org/10.3386/w3434
[19] Best R.J. and Best R.W., (2001) “Prior Information and Market Reaction to Dividend Changes”
The Review of Quantitative Finance and Accounting, 17, 361-376.
[20] Bhattacharaya S., (1979) “Imperfect Information, Dividend policy, and” the Bird in the Hand
Fallacy” Bell Journal of Economics 10, 259-270. https://doi.org/10.2307/3003330

25

Electronic copy available at: https://ssrn.com/abstract=3359144

[21] Black F., (1976) “The Dividend Puzzle” Journal of Portfolio management, Vol 2, 5-8.
https://doi.org/10.3905/jpm.1976.408558
[22] Brav A., Graham J., Harvey C. and Michaely R., (2003) “Payout Policy in the 21st century” The
Journal of Financial Economics 77, 486-527. https://doi.org/10.2139/ssrn.571046
[23] Brigham E. and Gordon M., (1968) “Leverage, Dividends Policy and the Cost of Capital” The
Journal of Finance 23, 85-103. https://doi.org/10.1111/j.1540-6261.1968.tb02999.x
[24] Brown S. and Warner J., (1985) “Using daily stock returns: The Case of event Studies” The
Journal of Financial Economics, 14, 3-31. https://doi.org/10.1016/0304-405x(85)90042-x
[25] Charest G., (1978) “Dividend Information, Stock Return and Market Efficiency” The Journal of
Financial Economics, 6, 297-330. https://doi.org/10.1016/0304-405x(78)90033-8
[26] DeAngelo H. and DeAngelo L., (2006) “The irrelevance of the MM Dividend Irrelevance
Theorem” The Journal of Financial Economics, 79, 293-315. https://doi.org/10.2139/ssrn.680855
[27] DeAngelo H., DeAngelo L. and Skinner D., (1996) “Reversal of fortune, dividend signaling and
the disappearance of sustained earning growth” The Journal of Financial Economics, 40, 341-371.
https://doi.org/10.1016/0304-405x(95)00850-e
[28] DeAngelo H., and DeAngelo L., (1990) “Dividend Policy and Financial Distress: An Empirical
Investigation of Troubled NYSE firms” The Journal of Finance, Vol 45, 1415-143.
https://doi.org/10.2307/2328743
[29] Denis J., Denis K., and Sarin A., (1994) “The information content of dividend changes: Cash-
flow signaling over investment and dividend clienteles” The Journal of Financial and Quantitative
Analysis 29, 567-587. https://doi.org/10.2307/2331110
[30] Eades K., (1982) “Empirical evidence on dividend as a signal of firms value” The Journal of
Financial and Quantitative Analysis Vol 17(4), 471-500. https://doi.org/10.2139/ssrn.2243505
[31] Elfakhani S., (1995) “An empirical examination of the information content of balance sheet and
dividend announcements: A signaling Approach” The Journal of Financial and Strategic decision,
Vol 8, 65-76
[32] Fama E. and French K., (1988) “Dividend yield and expectated stock returns” The Journal of
Financial Economics, Vol 22, 3-25. https://doi.org/10.1016/0304-405x(88)90020-7
[33] Harada K. and Nguyen P., (2005) “Dividend changes context and signaling efficiency in Japan”
Pacific-Basin Finance Journal, 13, 504-522. https://doi.org/10.1016/j.pacfin.2005.01.001
[34] Healy P. and Palepu K., (1988) “Earning information conveyed by dividend initiation and
omission” The Journal of Financial Economics, 21, 149-176. https://doi.org/10.1016/0304-
405x(88)90059-1
[35] Jensen M. and Meckling W., (1976) “Theory of the firm: managerial behavior, agency costs and
ownership structure” The Journal of Financial Economics, 3, 305-360.
https://doi.org/10.2139/ssrn.94043
[36] John K. and Kalay A., (1982) “Costly contracting and optimal payout constraints” The journal of
Finance, 37, 457-470. https://doi.org/10.2307/2327348
[37] John K. and Williams J., (1985) “Dividend, Dilution and Taxes: A signaling Equilibrium” The
journal of Finance, 40, 1053-1070. https://doi.org/10.2307/2328394
[38] Kalay A., (1980) “Signaling Information Content and the Reluctance to Cut Dividends” The
Journal of Financial and Quantitative Analysis, 15, 855-869. https://doi.org/10.2307/2330564

26

Electronic copy available at: https://ssrn.com/abstract=3359144

 [39] Lintner J., (1956) “ Distribution of Incomes of Corporations among Dividend, retained earnings
and taxes” The American Economic Review , 46, 429-443
[40] Lintner J., (1964) “Optimal Dividends and Corporate Growth under uncertainty” The Quarterly
Journal of Economics, 78, 49-95. https://doi.org/10.2307/1880545
[41] Michaely R., Thaler R. and Womack K., (1995) “Prices reaction to dividend initiation and
omission: Overreaction or drift?” The Journal of Finance, 50, 573-608.
https://doi.org/10.3386/w4778
[42] Miller M. and Modigliani F., (1961) “Dividend Policy, Growth and the Valuation of Shares” The
Journal of Business, 34, 411-43. https://doi.org/10.1086/294442
[43] Miller M. and Rock K. (1985) “Dividend Policy under Asymmetric Information” The Journal of
Finance, 40, 1118-1141. https://doi.org/10.2307/2328393
[44] Nissim D. and Ziv A., (2001) “Dividend Changes and Future Profitability” The Journal of Finance,
56, 2111-2133. https://doi.org/10.1111/0022-1082.00400
[45] Ofer A. and Siegel D., (1987) “Corporate Financial policy Information and Market Expectations:
An Empirical investigation of Dividends” The Journal of Finance, 42, 889-911.
https://doi.org/10.2307/2328297
[46] Pettit R., (1972) “Dividend Announcements, Security Performance, and Capital Market
Efficiency” The Journal of Finance, 27, 993-1007. https://doi.org/10.1111/j.1540-
6261.1972.tb03018.x
[47] Pettit R., (1976) “The impact of Dividend announcements: A reconciliation” The Journal of
Business, 49, 86-96. https://doi.org/10.1086/295813
[48] Taleb L., (2015). Politique de Dividende des Entreprises Tunisiennes : Une Explication par une
Etude de Cas et les Modèles de Base. SSRN Working Paper.
http://dx.doi.org/10.2139/ssrn.3145218
[49] Taleb L., (2018). “Dividend Policy in Tunisia: A Signalling Approach” The International Journal
of Economics and Finance, Vol10, 4, 84-94. doi:10.5539/ijef.v10n4p84
[50] Walter J., (1956) “Dividend Policy and Common Stock Prices” The Journal Finance, 11, 29-41.
https://doi.org/10.1111/j.1540-6261.1956.tb00684.x
[51] Walter J., (1963) “Dividend Policy: it’s Influence on the Value of the Enterprise” The Journal of
Finance, 18, 280-291. https://doi.org/10.2307/2977909
[52] Watts R., (1973) “The information Content of Dividend” The Journal of Business, 46, 191-211.
https://doi.org/10.1086/295525
[53] Williams J., (1988) “Efficient Signaling with Dividend, Investments and Stock Repurchases” The
Journal of Finance, 43, 737-747. https://doi.org/10.1111/j.1540-6261.1988.tb04605.x.

27

View publication stats Electronic copy available at: https://ssrn.com/abstract=3359144