Dividend Policy, Growth, and the Valuation of Shares

Author(s): Merton H. Miller and Franco Modigliani

Source: The Journal of Business, Vol. 34, No. 4 (Oct., 1961), pp. 411-433
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/2351143 .
Accessed: 11/09/2013 08:33

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The
Journal of Business.

http://www.jstor.org

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
THE JOURNAL OF BUSINESS
The GraduateSchool of Business of the University of Chicago

VOL.XXXIV OCTOBER1961 No. 4

DIVIDEND POLICY, GROWTH, AND THE

VALUATION OF SHARES*
MERTON H. MILLERt AND FRANCO MODIGLINIt

Tz i~xeffect of a firm'sdividendpolicy of dividendpolicy. Lackingsuch a state-

on the currentpriceof its sharesis a ment, investigators have not yet been
matter of considerableimportance, able to frame their tests with sufficient
not only to the corporate officials who precision to distinguish adequately be-
must set the policy, but to investors tween the various contending hypothe-
planning portfolios and to economists ses. Nor have they been able to give a
seeking to understandand appraise the convincingexplanationof what their test
functioning of the capital markets. Do results do imply about the underlying
companies with generous distribution process of valuation.
policies consistently sell at a premium In the hope that it may help to over-
over those with -niggardlypayouts? Is the come these obstacles to effective empiri-
reverseever true? If so, under what con- cal testing, this paper will attempt to fill
ditions? Is there an optimum payout the existing gap in the theoreticallitera-
ratio or range of ratios that maximizes ture on valuation.We shall begin, in Sec-
the currentworth of the shares? tion I, by examiningthe effects of differ-
Although these questions of fact have ences in dividend policy on the current
been the subject of many empiricalstud- price of sharesin an ideal economy char-
ies in recent years no consensushas yet acterizedby perfect capital markets, ra-
been achieved. One reasonappearsto be tional behavior, and perfect certainty.
the absence in the literature of a com- Still within this convenient analytical
plete and reasonablyrigorousstatement frameworkwe shall go on in Sections II
of those parts of the economic theory of and III to considercertaincloselyrelated
valuation bearingdirectly on the matter issues that appear to have been respon-
* The authors wish to express their thanks to all sible for considerablemisunderstanding
who read and commented on earlier versions of this of the role of dividendpolicy. In particu-
paper and especially to Charles C. Holt, now of the
University of Wisconsin, whose suggestions led to lar, Section II will focus on the long-
considerable simplification of a number of the proofs. standing debate about what investors
t Professor of finance and economics, University "really"capitalizewhen they buy shares;
of Chicago.
and SectionIII on the muchmooted rela-
t Professor of economics, Northwestern Univer-
sity. tions betweenprice, the rate of growth of
411

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
412 THE JOURNALOF BUSINESS
profits, and the rate of growth of divi- vestor as to the future investment pro-
dends per share. Once these fundamen- gram and the future profits of every cor-
tals have been established,we shall pro- poration. Because of this assurance,
ceed in Section IV to drop the assump- there is, among other things, no need to
tion of certainty and to see the extent to distinguishbetween stocks and bonds as
which the earlierconclusionsabout divi- sourcesof funds at this stage of the anal-
dend policy must be modified.Finally, in ysis. We can, therefore, proceed as if
Section V, we shall briefly examine the there were only a single type of financial
implications for the dividend policy instrument which, for convenience, we
problem of certain kinds of market im- shall refer to as shares of stock.
perfections. The fundamental principle of valua-
tion.-Under'these assumptionsthe valu-
I. EFFECT OF DIVIDEND POLICY WITH PER-
ation of all shares would be governedby
FECT MARKETS, RATIONAL BEHAVIOR,
the followingfundamentalprinciple:the
AND PERFECT CERTAINTY
price of each sharemust be such that the
The meaningof the basic assumptions. rate of return (dividends plus capital
-Although the terms "perfectmarkets," gains per dollarinvested) on every share
"rational behavior," and "perfect cer- will be the same throughout the market
tainty" are widely used throughouteco- over any given interval of time. That is,
nomic theory, it may be helpful to start if we let
by spelling out the precise meaning of dj(t) = dividendsper sharepaid by firmj
these assumptionsin the presentcontext. duringperiodt
1. In "perfect capital markets," no pj(t) = the price (ex any dividend in t - 1)
buyer or seller (or issuer) of securitiesis of a sharein firmj at the start of
large enoughfor his transactionsto have period t,
an appreciableimpact on the then ruling we must have
price. All tradershave equal and costless
access to information about the ruling dj(t) +pj(t+ 1) -pj(t)
priceand about all otherrelevant charac- pj(t) ~~~(1)
teristics of shares (to be detailed spe- = p ( t ) independentof j;
cificallylater). No brokeragefees, trans-
fer taxes, or other transaction costs are or, equivalently,
incurred when securities are bought,
sold, or issued, and there are no tax dif- pj( t)= [dj(t)+pj(t+)] (2)
ferentialseither between distributedand
undistributed profits or between divi- for eachj and for all t. Otherwise,holders
dends and capital gains. of low-return (high-priced)shares could
2. "Rational behavior" means that increase their terminal wealth by selling
investors always prefer more wealth to these shares and investing the proceeds
less and are indifferentas to whether a in shares offeringa higher rate of return.
given incrementto their wealth takes the This process would tend to drive down
form of cash payments or an increase in the prices of the low-return shares and
the market value of their holdings of drive up the prices of high-returnshares
shares. until the differential in rates of return
3. "Perfect certainty" implies com- had been eliminated.
plete assuranceon the part of every in- The effectof dividendpolicy.-The im-

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 413
plications of this principlefor our prob- able informationas to what that future
lem of dividend policy can be seen some- dividendpolicy wouldbe. The first possi-
what more easily if equation (2) is re- bility being the relevant one from the
stated in terms of the value of the enter- standpointof assessingthe effects of divi-
prise as a whole rather than in terms of dend policy, it will clarify matters to as-
the value of an individual share. Drop- sume, provisionally,that the future divi-
ping the firm subscriptj since this will dend policy of the firm is known and
lead to no ambiguity in the present con- given for t + 1 and all subsequent peri-
text and letting ods and is independentof the actual divi-
n(t) = the number of shares of record dend decision in t. Then V(t + 1) will
at thestart of t also be independent of the current divi-
m(t + 1) = the number of new shares (if dend decision, though it may very well
any) sold during t at the ex be affected by D(t + 1) and all subse-
dividendclosingpricep(t + 1), quent distributions.Finally, currentdiv-
so that idends can influence V(t) through the
n(t + 1) = n(t) + m(t + 1)
V(t) = n(t) p(t) = the total value of third term, -m(t + 1) p(t + 1), the val-
the enterpriseand ue of new sharessold to outsidersduring
D(t) = n(t) d(t) = the total dividends the period. For the higher the dividend
paid duringt to holdersof rec- payout in any period the more the new
ordat the start of t,
capital that must be raisedfrom external
we can rewrite (2) sources to maintain any desired level of
investment.
V(t l +,) 1[D(t)+n(t)p(t+1) I The fact that the dividend decision
1+0 effects price not in one but in these two
conflicting ways-directly via D(t) and
-1+ (t) [ D(t) + V(t+ 1)
inversely via -m(t) p(t + 1)-is, of
-m (t+ 1) p (t+ 1)I. (3) course,preciselywhy one speaks of there
being a dividend policy problem.If the
The advantage of restating the funda- firm raises its dividend in t, given its in-
mental rule in this form is that it brings vestment decision,will the increasein the
into sharper focus the three possible cash payments to the currentholdersbe
routes by which currentdividendsmight more or less than enough to offset their
affect the current market value of the lower shareof the terminalvalue? Which
firm V(t), or equivalently the price of its is the better strategy for the firm in
individual shares, p(t). Current divi- financingthe investment: to reducedivi-
dends will clearly affect V(t) via the first dends and rely on retainedearningsor to
term in the bracket, D(t). In principle, raise dividends but float more new
current dividends might also affect V(t) shares?
indirectly via the second term, V(t + 1), In our ideal world at least these and
the new ex dividend market value. Since related questions can be simply and im-
V(t + 1) must depend only on future mediately answered: the two dividend
and not on past events, such could be the effects must always exactly cancel out so
case, however, only if both (a) V(t + 1) that the payout policy to be followedin t
were a function of future dividendpolicy will have no effect on the price at t.
and (b) the current distribution D(t) We need only expressm(t+l)1 p(t+1)
servedto convey some otherwiseunavail- in terms of D(t) to show that such must

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
414 THE JOURNALOF BUSINESS
indeed be the case. Specifically, if I(t) once you think of it." It is, after all,
is the given level of the firm's invest- merely one more instance of the general
ment or increasein its holdingof physical principle that there are no "financialil-
assets in t and if X(t) is the firm's total lusions"in a rationaland perfecteconom-
net profit for the period, we know that ic environment. Values there are deter-
the amount of outside capital required mined solely by "real" considerations-
will be in this case the earning power of the
m(t+1)p(t+1) = I(t) firm'sassets and its investment policy-
(4) and not by how the fruits of the earning
- [X (t) -D (t) ]. power are "packaged" for distribution.
Substituting expression (4) into (3), the Obvious as the proposition may be,
D(t) cancel and we obtain for the value however,one finds few referencesto it in
of the firm as of the start of t the extensive literature on the problem.'
It is true that the literatureaboundswith
V (t)-n (t) p (t) statements that in some "theoretical"
(5) sense, dividend policy ought not to
= +p(t)[X
1 (t)-I(t) + V(t+ 1) count; but either that sense is not clearly
specified or, more frequently and espe-
Since D(t) does not appear directly cially among economists, it is (wrongly)
among the arguments and since X(t), identified with a situation in which the
I(t), V(t + 1) and p(t) are all independ- firm's internal rate of return is the same
ent of D(t) (either by their nature or by as the external or market rate of re-
assumption) it follows that the current turn.2
value of the firmmust be independentof A major source of these and related
the current dividend decision. misunderstandingsof the role of the divi-
Having established that V(t) is unaf- dend policy has been the fruitlessconcern
fected by the current dividend decision and controversy over what investors
it is easy to go on to show that V(t) must "really"capitalizewhen they buy shares.
also be unaffectedby any futuredividend We say fruitless because as we shall now
decisions as well. Such future decisions proceedto show, it is actually possibleto
can influenceV(t) only via their effect on derive from the basic principleof valua-
V (t + 1). But we can repeat the reason- tion (1) not merely one, but severalvalu-
ing above and show that V(t + 1)-and ation formulaseach starting from one of
hence V(t)-is unaffected by dividend the "classical" views of what is being
policy in t + 1; that V(t + 2)-and capitalized by investors. Though differ-
hence V(t + 1) and V(t)-is unaffected ing somewhat in outward appearance,
by dividend policy in t + 2; and so on the various formulascan be shown to be
for as far into the future as we care to equivalent in all- essential respects in-
look. Thus, we may concludethat given a cluding, of course, their implicationthat
firm's investment policy, the dividend dividend policy is irrelevant. While the
payout policy it chooses to follow will af- 1 Apart from the referencesto it in our earlier
fect neither the currentprice of its shares papers, especially [16], the closest approximation
nor the total return to its shareholders. seemsto be that in Bodenborn[1, p. 4921,but even
his treatmentof the role of dividendpolicy is not
Like many other propositionsin eco- completelyexplicit. (The numbersin bracketsrefer
nomics, the irrelevanceof dividend pol- to referenceslisted below,pp. 432-33).
icy, given investment policy, is "obvious, 2 See belowp. 424.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 415
controveryitself thus turns out to be an as T approachesinfinity4so that (7) can
empty one, the different expressionsdo be expressedas
have some intrinsic interest since, by T-1
highlighting different combinations of v (O) = rnim (8)
variablesthey provideadditionalinsights
into the process of valuation and they X [X(t)-I(t)],
open alternative lines of attack on some
of the problemsof empiricaltesting. which we shall further abbreviateto
c
1
II. WHAT DOES THE MARKET "REALLY"
=2
V(O) (1-+I1 [X(t)-I(t)]. (9)
CAPITALIZE? t- (I+ P)t
In the literatureon valuation one can
The discounted cash flow approach.-
find at least the following four more or
Consider now the so-called discounted
less distinct approachesto the valuation
of shares: (1) the discounted cash flow cash flow approach familiar in discus-
approach; (2) the current earnings plus sions of capital budgeting.There, in val-
future investment opportunities ap- uing any specificmachinewe discount at
proach; (3) the stream of dividends ap- the market rate of interest the stream of
proach; and (4) the stream of earnings cash receipts generated by the machine;
approach.To demonstratethat these ap- plus any scrap or terminal value of the
proachesare, in fact, equivalentit will be machine; and minus the stream of cash
helpful to begin by first going back to outlays for direct labor, materials, re-
equation (5) and developing from it a pairs, and capital additions. The same
valuation formula to serve as a point of approach,of course, can also be applied
referenceand comparison.Specifically,if to the firm as a whole which may be
we assume, for simplicity, that the mar- thought of in this context as simply a
ket rate of yield p (t) = p for all t,3 then, large, composite machine.5 This ap-
setting t = 0, we can rewrite (5) as 3More generalformulasin which p(t) is allowed
V (O) 1 IX (O)-I (0) ] to vary with time can alwaysbe derivedfromthose
presentedhere merelyby substitutingthe cumber-
+ 1 +p ( someproduct
+-- V (1). (6)

1L [l+p(r)] for (1+p)t+'

Since (5) holds for all t, setting t = 1 per- TO0

mits us to express V(1) in terms of V(2)

4 The assumptionthat the remainder vanishesis
which in turn can be expressedin terms introducedfor the sake of simplicityof exposition
of V(3) and so on up to any arbitrary only and is in no way essential to the argument.
terminal period T. Carrying out these What is essential,of course,is that V(O),i.e., the
substitutions, we obtain sum of the two termsin (7), be finite, but this can
always be safely assumedin economicanalysis.See
T-1 below,n. 14.
V(O) = E(l+p)t+l[X(t)I(t)] 5 This is, in fact, the approachto valuationnor-

mally takenin economictheorywhen discussingthe

value of the assetsof an enterprise,but much more
V(T). rarely applied, unfortunately,to the value of the
+(1+p) liability side. One of the few to apply the approach
to the sharesas well as the assetsis Bodenhornin [1],
In general,the remainderterm (1 + P)-T. who uses it to derivea formulacloselysimilarto (9)
V(T) can be expected to approachzero above.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
416 THE JOURNALOF BUSINESS
proach amounts to defining the value of investments in real assets that will yield
the firm as more than the "normal"(market)rate of
T-1 return. The latter opportunities, fre-
V(O) = E quently termed the "good will" of the
(0 P)
t=O (10) business,may arise,in practice,from any
of a number of circumstances (ranging
X
[E (t-co() +(+p Tv (T),
all the way fromspeciallocationaladvan-
tages to patents or other monopolistic
where IR(t)representsthe stream of cash
advantages).
receipts and ()(t) of cash outlays, or,
To see how these opportunitiesaffect
abbreviating,as above, to
the value of the business assume that in
co
some future period I the firminvests 1(t)
v ( =
?) E,_O (1+p),+'(11
1+p teRW [st-(t I (11) dollars. Suppose, further, for simplicity,
.
that starting in the period immediately
But we also know, by definition, that following the investment of the funds,
[X(t) -I(t)] = [IR(t) -()(t)] since, X(t) the projectsproducenet profitsat a con-
differs from IR(t) and 1(t) differs from stant rate of p*(t) per cent of I (t) in each
CO(t)merely by the "cost of goods sold" period thereafter.6 Then the present
(and also by the depreciationexpense if worth as of t of the (perpetual)stream of
we wish to interpretX(t) and I(t) as net profitsgeneratedwill be I(t) p*(t)/p, and
rather than gross profits and invest- the "good will" of the projects (i.e., the
ment). Hence (11) is formallyequivalent differencebetween worth and cost) will
to (9), and the discounted cash flow ap- be
proach is thus seen to be an implication I(t)fP-22)-I(t)
P* =1(t) [P P*(t) P
of the valuation principle for perfect
markets given by equation (1). The present worth as of now of this fu-
The investmentopportunitiesapproach. ture "good will" is
-Consider next the approachto valua-
tion which would seem most natural It P* ( ) p] (1 + p)-+
from the standpoint of an investor pro-
posing to buy out and operate some al- and the present value of all such future
ready-going concern. In estimating how opportunitiesis simply the sum
much it would be worthwhileto pay for
the privilege of operating the firm, the
amount of dividendsto be paid is clearly to P
not relevant, since the new owner can, Adding in the present value of the (uni-
within wide limits, make the future divi- formperpetual)earnings,X(O),on the as-
dend stream whatever he pleases. For
8The assumptionthat I(t) yields a uniformper-
him the worth of the enterprise,as such, petuity is not restrictivein the present certainty
will depend only on: (a) the "normal" context since it is always possible by means of
rate of return he can earn by investing simple,present-valuecalculationsto findan equiva-
his capital in securities (i.e., the market lent uniformperpetuityfor any project, whatever
the time shape of its actual returns.Note also that
rate of return); (b) the earning power of p*(t) is the averagerate of return.If the managersof
the physical assets currentlyheld by the the firmarebehavingrationally,they will, of course,
firm; and (c) the opportunities, if any, use p as their cut-off criterion(cf. below p. 418).
In this event we would have p*(t) > p. The for-
that the firmoffersfor makingadditional mulasremainvalid, however,even wherep*(t) < p.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATION OF SHARES 417

sets currentlyheld, we get as an expres- The first expression is, of course,

sion for the value of the firm simply a geometricprogressionsumming
to X(O)/p, which is the first term of (12).
V(O) =(O) + E I (t) To simplify the second expression note
P t=O (12) that it can be rewrittenas
xP*
(t) - p
XP()----( 1 +P
+ p)-(t+l).
p
1:I (t)
tO0
[p*t E
T-=t+2
( 1+ P) -T
To show that the same formulacan be
derivedfrom (9) note first that ourdefini- ( 1 + p)(t+)] -

tion of p*(t) impliesthe followingrelation

Evaluating the summation within the
between the X(t):
brackets gives
X (1) = X (O) + p* (O) I (O),
....................
E .1(t)
X (t) = X(t -1) +p* (t -1) I(t -1) [p*(t)( + p) +
, I(t) -(t+l
t00 - (1+p)-(t+1)]
and by successive substitution

X (t) = X(O) + Yd p* X ()
t-1
= I(t (t)]*P +p -t)
Tr=O

t=1,2 ...owhich is precisely the second term of

.
(12).
Substituting the last expression for Formula (12) has a number of reveal-
X(t) in (9) yields ing features and deserves to be more
V(O) = [X(O)-I(O)] (1 + p) widely used in discussionsof valuation.7
For one thing, it throws considerable
light on the meaning of those much
+X X(O) +Ep*(r)I (r) abused terms "growth" and "growth
stocks."As can readilybe seen from (12),
a corporationdoes not becomea "growth
stock" with a high price-earningsratio
merely because its assets and earnings
are growing over time. To enter the
=X(O)-(O +1p)-1 (1 glamorcategory, it is also necessarythat
I
p*(t) > p. For if p*(t) = p, then how-
t1-1 (
ever large the growth in assets may be,
t =1
the second term in (12) will be zero and
t = T=O
the firm'sprice-earningsratio would not
X ( + p)-t)
rise above a humdrum i/p. The essence
CO
0 ___ of "growth,"in short, is not expansion,
p) -t
=X(O) f,
t =1
(I1+
but the existence of opportunitiesto in-
vest significant quantities of funds at
higher than "normal"rates of return.
+ Y.
t =1 T=O
*T) T-It1 7A valuationformulaanalogousto (12) though
derivedand interpretedin a slightly differentway
{12
is foundin Bodenhorn[1].Variantsof (12) forcertain
X (+ P) +5 )(t
specialcases are discussedin Walter[201.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 418 THE JOURNALOF BUSINESS
Notice also that if p*(t) < p, invest- valuation.
ment in real assets by the firm will ac- Even without raisingquestions of bias
tually reduce the current price of the in the coefficients,9it should be apparent
shares. This should help to make clear that such a conclusion is unwarranted
among other things, why the "cost of since formula (12) and the analysis un-
capital" to the firm is the same regard- derlyingit imply only that dividendswill
less of how the investments are financed not count given current earnings and
or how fast the firmis growing.The func- growthpotential.No generalpredictionis
tion of the cost of capital in capital made (or can be made) by the theory
budgetingis to providethe "cut-offrate" about what will happen to the dividend
in the sense of the minimum yield that coefficient if the crucial growth term is
investment projects must promise to be omitted."0
worth undertaking from the point of The stream of dividends approach.-
view of the current owners. Clearly, no From the earnings and earnings oppor-
proposedproject would be in the interest tunities approach we turn next to the
of the currentownersif its yield were ex- dividend approach,which has, for some
pected to be less than p since investing in reason, been by far the most popularone
such projects would reduce the value of in the literature of valuation. This ap-
their shares.In the other direction,every proach too, properly formulated, is an
project yielding more than p is just as entirely valid one though, of course, not
clearly worth undertaking since it will the only valid approachas its more en-
necessarily enhance the value of the en- thusiastic proponents frequently sug-
terprise.Hence, the cost of capital or cut- gest." It does, however, have the disad-
off criterion for investment decisions is vantage in contrast with previous ap-
simply p.8 proachesof obscuringthe role of dividend
Finally, formula (12) serves to em- policy. In particular,uncriticaluse of the
phasize an important deficiencyin many I The seriousbias problemin tests using current
recent statistical studies of the effects of reportedearningsas a measureof X(O)wasdiscussed
dividend policy (such as Walter [19] or brieflyby us in [161.
Durand [4, 5]). These studies typically 11In suggesting that recent statistical studies
involve fitting regression equations in have not controlledadequatelyfor growthwe do not
mean to exempt Gordonin [81or [9]. It is true that
which price is expressedas some function his tests containan explicit "growth"variable,but
of currentearningsand dividends.A find- it is essentiallynothing more than the ratio of re-
ing that the dividend coefficient is sig- tainedearningsto book value. This ratio wouldnot
in generalprovidean acceptableapproximationto
nificant-as is usually the case-is then the "growth"variableof (12) in any samplein which
interpretedas a rejectionof the hypothe- firms resortedto external financing.Furthermore,
even if by somechancea samplewas foundin which
sis that dividend policy does not affect all firms relied entirely on retained earnings,his
8 The same conclusion could also have been tests then couldnot settle the questionof dividend
reached, of course, by "costing" each particular policy.Forif all firmsfinancedinvestmentinternally
sourceof capitalfunds.That is, since p is the going (or used externalfinancingonly in strict proportion
market rate of return on equity any new shares to internalfinancingas Gordonassumesin [81)then
floated to finance investment must be priced to there would be no way to distinguishbetween the
yield p; and withholdingfunds from the stockhold- effectsof dividendpolicyand investmentpolicy (see
ers to financeinvestmentwoulddeprivethe holders belowp. 424).
of the chanceto earn p on these funds by investing 11See, e.g., the classic statementof the position
their dividendsin other shares. The advantageof in J. B. Williams[211.The equivalenceof the divi-
thinkingin termsof the cost of capitalas the cut-off dend approachto many of the other standardap-
criterionis that it minimizesthe dangerof confusing proachesis noted to our knowledgeonly in our [16]
"costs"with mere"outlays." and, by implication,in Bodenhorn[1].

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 419
dividend approachhas often led to the The summation term in the last expres-
unwarrantedinferencethat, since the in- sion can be written as the differencebe-
vestor is buying dividendsand since div- tween the stream of dividends accruing
idend policy affects the amount of divi- to all the sharesof recordas of t + 1 and
dends, then dividend policy must also that portion of the stream that will ac-
affect the currentprice. crueto the sharesnewlyissuedin t, that is,
Properly formulated,the dividend ap-
proach defines the current worth of a 1:Dt (t+ r+ 1) I m (t+ 1)0
share as the discounted value of the
stream of dividends to be paid on the c
(16)
share in perpetuity. That is Dt+l (t+T+ 1)
X
co
( I1 + p)rl+
1 d
p (t) = +(13)
-r=o( 1 + P) But from (14) we know that the second
7+1

To see the equivalencebetween this ap- summation in (16) is precisely V(t + 1)

proach and previous ones, let us first so that (15) can be reduced to
restate (13) in terms of total market
value as V(t) =_l [D (t)

V (t)-
V(t)= 2. (t + ) 14)
(+I p)+1'(4
< (t+ 1)p (t+ 1)
where Dt(t + r) denotes that portion of [D(o+l)V(t+ 1)> (17)
the total dividendsD(t + r) paid during X V(t+ 1)]
periodt + r, that accruesto the sharesof
record as of the start of period t (indi- =+[D(t) + V(t+ 1)
cated by the subscript). That equation
(14) is equivalent to (9) and hence also -m(t+ 1) p(t+ 1)],
to (12) is immediately apparent for the which is (3) and which has already been
specialcase in whichno outside financing shown to imply both (9) and (12).12
is undertakenafter period t, for in that There are, of course, other ways in
case which the equivalence of the dividend
approachto the other approachesmight
-I(t+=r).
-X(t+r)
12The statement that equations (9), (12), and
To allow for outside financing,note that (14) are equivalentmust be qualifiedto allow for
we can rewrite (14) as certainpathologicalextremecases,fortunatelyof no
real economicsignificance.An obvious exampleof
such a case is the legendarycompanythat is expect-
V(t) D(t) ed never to pay a dividend.If this wereliterallytrue
then the valueof the firmby (14) wouldbe zero;by
(9) it wouldbe zero (or possiblynegativesince zero
dividends rule out X(t) > I(t) but not X(t) < I(t));
1 +P [ ( whileby (12) the valuemightstill be positive.What
is involvedhere, of course,is nothing more than a
discontinuityat zero since the value under(14) and
(9) would be positive and the equivalenceof both
+E ( 1 +p)7~~~~~~ with (12) wouldhold if that valuewerealsopositive
as long as there was some period T, howeverfar in
the future, beyond which the firm would pay out
+ E DtcoD(t +1 e > 0 per cent of its earnings,howeversmall the
value of e.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
420 THE JOURNAL OF BUSINESS

have been established, but the method often suggested, that it overlooks the
presented has the advantage perhaps of fact that the corporationis a separateen-
providing some further insight into the tity and that these profits cannot freely
reason for the irrelevance of dividend be withdrawn by the shareholders;but
policy. An increasein currentdividends, ratherthat it neglects the fact that addi-
given the firm'sinvestment policy, must tional capital must be acquiredat some
necessarily reduce the terminal value of cost to maintain the future earnings
existing sharesbecausepart of the future stream at its specifiedlevel. The capital
dividend stream that would otherwise to be raised in any future period is, of
have accruedto the existing sharesmust course, I(t) and its opportunity cost, no
be diverted to attract the outside capital matter how financed, is p per cent per
from which, in effect, the higher current period thereafter. Hence, the current
dividends are paid. Under our basic as- value of the firm under the earnings ap-
sumptions,however,p must be the same proach must be stated as
for all investors,new as well as old. Con- co

sequently the market value of the divi- V (0) = f +w+

dends diverted to the outsiders,which is
(18)
both the value of their contributionand
the reductionin terminalvalue of the ex- X [X(t) - pI(r)].
isting shares, must always be precisely
the same as the increase in current divi-
That this version of the earnings ap-
dends.
proachis indeedconsistentwith ourbasic
The stream of earnings approach.-
assumptions and equivalent to the pre-
Contraryto widely held views, it is also
vious approachescan be seen by regroup-
possible to develop a meaningful and
ing terms and rewritingequation (18) as
consistentapproachto valuationrunning
in terms of the stream of earningsgener-
ated by the corporationrather than of V(0) So (lp+ X(t)
the dividend distributionsactually made 00 00
to the shareholders.Unfortunately, it is pI (t)
also extremely easy to mistate or mis- t=oVS (I +p)7+12

interpretthe earningsapproachas would

be the case if the value of the firm were
to be defined as simply the discounted
sum of future total earnings.'3 The 00

troublewith such a definitionis not, as is y

(
+ p ) t+1
13 In fairness,we shouldpoint out that thereis no
one, to our knowledge,who has seriouslyadvanced 00
this view. It is a view whosemainfunctionseemsto PI (t)
be to serve as a "strawman" to be demolishedby
thosesupportingthe dividendview. See,e.g., Gordon
(9, esp. pp. 102-31.Other writers take as the sup- Since the last inclosed summation re-
posed earnings counter-viewto the dividend ap-
proachnot a relationrunningin termsof the stream duces simply to I(t), the expression(19)
of earningsbut simply the propositionthat price is in turn reduces to simply
proportional to current earnings, i.e., V(O)= 0c
X(O)/p. The probable origins of this widespread
misconceptionabout the earningsapproachare dis- V(0) = E (- 1 t_+1[X(t)-I (t)], (20)
cussedfurtherbelow (p. 424).

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 421
which is precisely our earlier equation and kp* is the (constant) rate of growth
(9). of total earnings. Substituting from (21)
Note that the version of the earnings into (12) for 1(t) we obtain
approachpresentedhere does not depend
for its validity upon any special assump- V(O) +_E
tions about the time shape of the stream
of total profitsor the stream of dividends X kX(O) [ 1 + kp*] t
per share. Clearly, however, the time
paths of the two streams are closely re- X ( 1 + p)-(t+I) (2 2)
lated to each other (via financialpolicy) _x(o) r k (p* -P)
and to the stream of returns derived by - L 1 +
holders of the shares. Since these rela- co
tions are of some interest in their own 1 +k P*>t
X Sk -J
right and since misunderstandingsabout
them have contributedto the confusion
Evaluating the infinite sum and simpli-
over the role of dividend policy, it may
fying, we finally obtain14
be worthwhile to examine them briefly
before moving on to relax the basic as- V(O) =-(?) [1 + k(p* p)]
sumptions. p p -k p
(23)
III. EARNINGS, DIVIDENDS, AND
_ X(O) (1 -k)
GROWTH RATES

The convenientcase of constantgrowth whichexpressesthe value of the firm as a

rates.-The relation between the stream function of its currentearnings, the rate
of earningsof the firm and the stream of of growthof earnings,the internalrate of
dividends and of returns to the stock- return, and the market rate of return.15
holders can be brought out most clearly 14One advantageof the specialization(23) is that
by specializing (12) to the case in which it makesit easy to see what is reallyinvolvedin the
investment opportunitiesare such as to assumptionhere and throughoutthe paperthat the
V(O)given by any of our summationformulasis
generate a constant rate of growth of necessarilyfinite (cf. above, n. 4). In terms of (23)
profits in perpetuity. Admittedly, this the conditionis clearlykp* < p, i.e., that the rate of
case has little empiricalsignificance,but growthof the firmbe less than marketrate of dis-
count.Althoughthe caseof (perpetual)growthrates
it is convenient for illustrative purposes greater than the discount factor is the much-dis-
and has received much attention in the cussed"growthstock praradox"(e.g. [6]), it has no
literature. real economicsignificanceas we pointed out in [16,
esp.n. 17,p. 664].This willbe apparentwhenone re-
Specifically, suppose that in each pe- calls that the discountrate p, though treated as a
riod t the firm has the opportunityto in- constant in partial equilibrium (relative price)
vest in real assets a sum 1(t) that is k per analysis of the kind presentedhere, is actually a
variable from the standpoint of the system as a
cent as large as its total earningsfor the whole.That is, if the assumptionof finite value for
period; and that this investment pro- all sharesdid not hold, becausefor some shareskp*
duces a perpetual yield of p* beginning was (perpetually)greater than p, then p would
necessarilyrise until an over-allequilibriumin the
with the next period.Then, by definition capitalmarketshad been restored.
X(t) = X(t- 1) + p*I(t- 1) 15 An interesting and more realistic variant of
(22), whichalsohas a numberof convenientfeatures
=X(t-) [I+kp*] (21) from the standpointof developingempiricaltests,
-X(O) [I + kp*] can be obtainedby assumingthat the specialinvest-

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
422 THE TOURNALOF BUSINESS
Note that (23) holds not just for period the price per share? Clearly, the answer
0, but for every t. Hence if X(t) is grow- will vary depending on whether or not
ing at the rate kp*, it follows that the the firm is paying out a high percentage
value of the enterprise, V(t), also grows of its earnings and thus relying heavily
at that rate. on outside financing. We can show the
The growthof dividendsand the growth nature of this dependence explicitly by
of total profits.-Given that total earn- making use of the fact that whatever the
ings (and the total value of the firm) are rate of growth of dividendsper share the
growing at the rate kp* what is the rate present value of the firm by the dividend
of growth of dividends per share and of approach must be the same as by the
earningsapproach.Thus let
ment opportunitiesare availablenot in perpetuity
but only over some finite intervalof T periods.To g = the rate of growthof divi-
exhibit the value of the firmfor this case, we need dends per share, or, what
only replacethe infinitesummationin (22) with a amountsto the same thing,
summation running from t = 0 to t = T - 1. Eval- the rate of growthof divi-
uating the resultingexpression,we obtain dendsaccruingto the shares
of the currentholders(i.e.,
V(O) X(O) x +k _(p*_ p) Do(t) = Do(O)[1+ g]t);
p p - kp kr= the fractionof total profits
(22a) retainedin each period (so
X[il(P + *)T]p that D(t) = X(O)[1 -kr]);
ke k - kr = the amountof externalcapi-
Note that (22a) holds even if kp* > p, so that the tal raised per period, ex-
so-calledgrowthparadoxdisappearsaltogether.If, pressed as a fraction of
as we should generallyexpect, (1 + kp*)/(l + p) profitsin the period.
is close to one, and if T is not too large, the right
hand side of (22a) admits of a very convenientap- Then the present value of the stream of
proximation.In this case in fact we can write dividends to the original owners will be
1I+P ] _I +T(kp* - p) (1O+g)t D (O)
Do EO p) p g
(24)
the approximationholding,if, as we shouldexpect,
(1 + kp*) and (I + p) are both close to unity. X(0O)[ 1-kr]
Substitutingthis approximationinto (22a) and sim- P-g
plifying,finallyyields
By virtue of the dividend approach we
V (0 X(O) [1 + k ( p* P)
p P-kP* know that (24) must be equal to V(O).
If, therefore,we equate it to the right-
XT(P -kp*) ( hand side of (23), we obtain
=X( )+ kX (O) (22b X (0)[1 1-kr] X ( O)[ 1-( kr+ ke)]
p P-g P-~kP*
from which it follows that the rate of
growth of dividends per share and the
The commonsense of (22b)is easy to see. The cur-
rent valueof a firmis givenby the valueof the earn- rate of growth of the price of a share
ing powerof the currentlyheld assetsplus the mar- must bel6
ket value of the specialearningopportunitymulti-
plied by the numberof yearsfor whichit is expected 16 That g is the rate of priceincreaseper shareas
to last. well as the rate of growthof dividendsper sharefol-

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 423

g=kp* 1_kr_ kep 1k . (25) tive kp*, if p* < p and if the firm pays
out a large fraction of its income in divi-
Notice that in the extreme case in which dends. In the other direction, we see
all financing is internal (ke = 0 and k = from (25) that even if a firm is a
kr), the second term drops out and the "growth"corporation(p* > p) then the
first becomes simply kp*. Hence the stream of dividends and price per share
growth rate of dividends in that special must grow over time even though kr =

In X(O)[II

FiG. 1.-Growth of dividendsper sharein relationto growthin total earnings:

A. Total earnings:ln X(t) = ln X(O) + kp*t;
B. Total earningsminus capital invested: ln [X(t) - I(t)] = In X(O) [1 - k] + kp*t;
Dividendsper share (all financinginternal):ln Do(t) = In D(O)+ gt = In X(O) [1 - k] + kp*t;
C. Dividendsper share (somefinancingexternal):ln Do(t) = In D(O)+ gt;
D. Dividendsper share (all financingexternal):In Do(t) = In X(O)+ [(k/i - k) (p* - p)]t.

case is exactly the same as that of total 0, that is, even though it pays out all its
profits and total value and is propor- earningsin dividends.
tional to the rate of retention kr. In all The relation between the growth rate
other cases, g is necessarilyless than kp* of the firm and the growth rate of divi-
and may even be negative, despite a posi- dends under various dividend policies is
illustrated graphically in Figure 1 in
lows from the fact that by (13) and the definition which for maximum clarity the natural
of g
logarithm of profits and dividends have
been plotted against time.'7
E
T (1 + p)T+ Line A shows the total earningsof the
firm growing through time at the con-
stant rate kp*, the slope of A. Line B
T=-O ( + P ) shows the growth of (1) the stream of
total earningsminus capital outlays and
d(r) 17 That is, we replaceeach discretecompounding
expressionsuch as X(t) = X(O) [1 + kp*]t with its
counterpartunder continuousdiscountingX(t) =
X(O)ekP*t which, of course, yields the convenient
=p(O) [1 + t
linear relationIn X(t) = In X(O)+ kp*t.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
424 THE JOURNALOF BUSINESS
(2) the streamof dividendsto the original optimum dividend policy for the firm
owners (or dividends per share) in the that depends on the internal rate of re-
special case in which all financingis in- turn. Such a conclusion is almost in-
ternal. The slope of B is, of course, evitable if one worksexclusivelywith the
the same as that of A and the (constant) assumption, explicit or implicit, that
differencebetween the curves is simply funds for investment come only from re-
ln(l - k), the ratio of dividends to tained earnings.For in that case dividend
profits. Line C shows the growth of divi- policy is indistinguishable from invest-
dends per share when the firm uses both ment policy; and there is an optimal in-
internal and external financing.As com- vestment policy which does in general
pared with the pure retention case, the depend on the rate of return.
line starts higher but grows more slowly Notice also from (23) that if p* = p
at the rate g given by (25). The higher and k = kr, the term [1 - kr] can be
the payout policy, the higherthe starting canceled from both the numerator and
position and the slowerthe growth up to the denominator.The value of the firm
the other limiting case of complete ex- becomes simply X(O)/p, the capitalized
ternal financing,Line D, which starts at value of current earnings. Lacking a
ln X(O)and growsat a rate of (k/I - k) . standardmodel for valuation more gen-
(P* -P). eral than the retainedearningscase it has
The special case of exclusivelyinternal been all too easy for many to conclude
financing.-As noted above the growth that this droppingout of the payout ratio
rate of dividends per share is not the [1 - kr] when p* = p must be what is
same as the growth rate of the firm ex- meant by the irrelevance of dividend
cept in the special case in which all policy and that V(O) = X(O)/p must
financingis internal. This is merely one constitute the "earnings"approach.
of a numberof peculiaritiesof this special Still another example of the pitfalls in
case on which, unfortunately, many basing argumentson this special case is
writers have based their entire analysis. provided by the recent and extensive
The reason for the preoccupation with work on valuation by M. Gordon.'8Gor-
this special case is far from clear to us. don argues, in essense, that because of
Certainlyno one would suggest that it is increasinguncertainty the discount rate
the only empiricallyrelevant case. Even p$(t) applied by an investor to a future
if the case were in fact the most common, dividend payment will rise with t, where
the theorist would still be under an obli- t denotes not a specific date but rather
gation to consider alternative assump- the distance from the period in which
tions. We suspect that in the last analy- the investor performs the discounting.'9
sis, the popularityof the internal financ- 18 See esp. [8]. Gordon's views represent the most
ing model will be found to reflect little explicit and sophisticated formulation of what might
more than its ease of manipulationcom- be called the "bird-in-the-hand" fallacy. For other,
elaborate, statements of essentially the same
bined with the failure to push the analy- less
position see, among others, Graham and Dodd [11,
sis far enough to disclosehow special and p. 433] and Clendenin and Van Cleave [3].
how treacherousa case it really is. 19 We use the notation Ap(t) to avoid any confusion

In particular, concentration on this between Gordon's purely subjective discount rate

special case appearsto be largely respon- and the objective, market-given yields p(t) in Sec. I
above. To attempt to derive valuation formulas
sible for the widely held view that, even under uncertainty from these purely subjective dis-
underperfect capital markets,there is an count factors involves, of course, an error essentially

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 425
Hence, when we use a single uniformdis- ally, such a change cannot affect market
count rate p as in (22) or (23), this rate valuations. Indeed, if they valued shares
shouldbe thought of as really an average according to the Gordon approach and
of the "true"rates p(t) each weighted by thus paid a premium for higher payout
the size of the expected dividend pay- ratios, then holders of the low payout
ment at time t. If the dividend stream is shareswouldactually realizeconsistently
growing exponentially then such a higher returns on their investment over
weighted average p would, of course, be any stated interval of time.20
higher the greater the rate of growth of Corporateearningsand investorreturns.
dividendsg since the greaterwill then be -Knowing the relation of g to kp* we
the portion of the dividend stream aris- can answera question of considerablein-
ing in the distant as opposedto the near terest to economic theorists, namely:
future. But if all financingis assumed to What is the precise relation between the
be internal, then g = krp* so that given earningsof the corporationin any period
p*, the weighted average discount factor X(t) and the total return to the owners
p will be an increasing function of the of the stock during that period?2'If we
rate of retention kr which would run let Gt(t) be the capital gains to the
counter to our conclusionthat dividend owners during t, we know that
policy has no effect on the currentvalue Dt (t) +Gt (t) = X(t) 26
of the firm or its cost of capital. X(1 - kr)+U V( )
For all its ingenuity, however, and its
20 This is not to deny that growthstocks (in our
seeming foundation in uncertainty, the sense)maywellbe "riskier"thannon-growthstocks.
argument clearly suffers fundamentally But to the extent that this is true, it will be due to
from the typical confoundingof dividend the possibly greater uncertaintyattaching to the
policy with investment policy that so size anddurationof futuregrowthopportunitiesand
henceto the size of the futurestreamof total returns
frequently accompanies use of the in- quite apart from any questionsof dividendpolicy.
ternal financingmodel. Had Gordonnot 21 Note also that the aboveanalysisenablesus to

confinedhis attention to this special case deal very easily with the familiarissue of whethera
(or its equivalent variants), he would firm'scost of equity capitalis measuredby its earn-
ings/priceratioor by its dividend/priceratio.Clear-
have seen that while a change in divi- ly, the answeris that it is measuredby neither,ex-
dend policy will necessarily affect the cept undervery specialcircumstances.For from(23)
size of the expected dividendpayment on we have for the earnings/priceratio
the sharein any futureperiod,it need not, X(O) _p-kp*
in the general case, affect either the size, V (O) 1-k
of the total return that the investor ex- which is equal to the cost of capital p, only if the
pects duringthat period or the degree of firm has no growth potential (i.e., p* = p). And from
(24) we have for the dividend/priceratio
uncertainty attaching to that total re-
turn. As should be abundantly clear by D(O) g
now, a change in dividend policy, given V (O)
investment policy, implies a change only whichis equalto p only wheng = 0; i.e., from(25),
in the distributionof the total return in either when k = 0; or, if k > 0, when p* < p and
the amountof externalfinancingis precisely
any period as between dividends and
capital gains. If investors behave ration- kg=pp k [I1-kr] X

analogousto that of attemptingto developthe cer- so that the gainfromthe retentionof earningsexact-
tainty formulasfrom "marginalrates of time pref- ly offsetsthe loss that wouldotherwisebe occasioned
erence"ratherthan objectivemarketopportunities. by the unprofitableinvestment.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
426 THE JOURNALOF BUSINESS

since the rate of growth of price is the share; or we can think in terms of the
same as that of dividends per share. total value of the enterprise, total earn-
Using (25) and (26) to substitute for g ings, and the rate of growth of total earn-
and V(t) and simplifying,we find that ings. Our own preferencehappens to be
for the second approach primarily be-
De(t)+Gt (t) = X (t) [P(_ k)] (2 7) cause certain additional variables of in-
terest-such as dividendpolicy, leverage,
The relation between the investors' re- and size of firm-can be incorporated
turn and the corporation'sprofits is thus more easily and meaningfully into test
seen to depend entirely on the relation equationsin whichthe growth term is the
between p* and p. If p* = p (i.e., the growth of total earnings. But this can
firm has no special "growth"opportuni- wait. For present purposes,the thing to
ties), then the expressionin brackets be- be stressed is simply that two ap-
comes 1 and the investor returnsare pre- proaches, properly carried through, are
cisely the same as the corporateprofits. in no sense opposingviews of the valua-
If p* < p, however,the investors' return tion process;but ratherequivalentviews,
will be less than the corporateearnings; with the choice between them largely a
and, in the case of growth corporations matter of taste and convenience.
the investors' return will actually be
IV. THE EFFECTS OF DIVIDEND POLICY
greater than the flow of corporateprofits
UNDER UNCERTAINTY
over the interval.22
Some implicationsfor constructingem- Uncertaintyand the general theoryof
pirical tests.-Finally the fact that we valuation.-In turning now from the
have two different(thoughnot independ- ideal world of certainty to one of uncer-
ent) measuresof growthin kp*and g and tainty our first step, alas, must be to jet-
two correspondingfamilies of valuation tison the fundamental valuation prin-
formulas means, among other things, ciple as given, say, in our equation (3)
that we can proceed by either of two
routes in empirical studies of valuation. V(t) - [D(t)+n(t) p(t+ 1)I
1+ p(t)
We can follow the standard practice of
the security analyst and think in terms and from which the irrelevanceproposi-
of price per share, dividends per share, tion as well as all the subsequentvalua-
and the rate of growth of dividends per capitalgain will alwaysbe greaterthan the retained
22The above relationbetweenearningsper share earnings(and there will be a capitalgain of
and dividendsplus capital gains also means that
therewill be a systematicrelationbetweenretained kX( P)[
earningsand capitalgains. The "marginal"relation
is easy to see and is alwayspreciselyone for one re-evenwhenall earningsarepaidout). Fornon-growth
gardlessof growthor financialpolicy.That is, taking corporationsthe relation between gain and reten-
a dollar away from dividendsand adding it to re- tions is reversed.Note also that the absolutediffer-
tained earnings(all other things equal) means an ence betweenthe total capitalgain and the total re-
increasein capitalgainsof one dollar(or a reduction tained earningsis a constant (given, p, k and p*)
in capitalloss of one dollar).The "average"relation unaffectedby dividendpolicy. Hence the ratio of
is somewhatmorecomplex.From (26) and (27) we capital gain to retainedearningswill vary directly
can see that with the payout ratio for growthcorporations(and
p -p vice versafornon-growthcorporations).This means,
Gt(t) =krX(t) +kX(t) p p*. amongother things, that it is dangerousto attempt
p -kp* to drawinferencesabout the relativegrowthpoten-
Hence, if p* = p the total capitalgain receivedwill tial or relativemanagerialefficiencyof corporations
be exactly the same as the total retainedearnings solely on the basisof the ratioof capitalgainsto re-
per share. For growth corporations,however, the tainedearnings(cf. Harkavy[12,esp. pp. 289-94]).

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 427
tion formulasin Sections II and III were fairly complex and space-consuming
derived.For the terms in the bracketcan task. Fortunately, however, this task
no longer be regardedas given numbers, need not be undertaken in this paper
but must be recognizedas "randomvari- which is concernedprimarilywith the ef-
ables" from the point of view of the in- fects of dividendpolicy on marketvalua-
vestor as of the start of periodt. Nor is it tion. For even without a full-fledgedthe-
at all clear what meaning can be at- ory of what doesdeterminemarket value
tached to the discount factor 1/[1 + underuncertaintywe can show that divi-
p(t)] since what is being discountedis not dend policy at least is not one of the de-
a given return,but at best only a proba- terminants. To establish this particular
bility distributionof possiblereturns.We generalizationof the previous certainty
can, of course,deludeourselvesinto think- resultswe need only invoke a correspond-
ing that we arepreservingequation(3) by ing generalizationof the original postu-
the simple and popular expedient of late of rationalbehavior to allow for the
drawinga bar over each term and refer- fact that, under uncertainty, choices de-
ring to it thereafteras the mathematical pend on expectationsas well as tastes.
expectation of the randomvariable. But "Imputedrationality"and "symmetric
except for the trivial case of universal marketrationality."-This generalization
linear utility functions we know that can be formulatedin two steps as follows.
V(t) would also be affected, and mate- First, we shall say that an individual
rially so, by the higherordermoments of trader "imputes rationality to the mar-
the distributionof returns. Hence there ket" or satisfies the postulate of "im-
is no reason to believe that the discount puted rationality"if, in formingexpecta-
factor for expected values, 1/[1 + p(t)], tions, he assumesthat every other trader
would in fact be the same for any two in the marketis (a) rationalin the previ-
firms chosen arbitrarily,not to mention ous sense of preferringmore wealth to
that the expected values themselvesmay less regardlessof the form an increment
well be different for different investors. in wealth may take, and (b) imputes ra-
All this is not to say, of course, that tionality to all other traders. Second, we
there are insuperable difficulties in the shall say that a market as a whole satis-
way of developing a testable theory of fies the postulate of "symmetricmarket
rational market valuation under uncer- rationality"if every traderboth behaves
tainty.23On the contrary, our investiga- rationallyand imputes rationality to the
tions of the problem to date have con- market.24
vinced us that it is indeedpossibleto con- Notice that this postulate of sym-
struct such a theory-though the con- 24We offerthe term"symmetricmarketrationali-
struction, as can well be imagined, is a ty" with considerablediffidenceand only afterhav-
ing been assuredby game theoriststhat thereis no
23 Nor doesit meanthat all the previouscertainty accepted term for this conceptin the literatureof
analysishas no relevancewhateverin the presence that subject even tbough the postulate itself (or
of uncertainty.Thereare manyissues,suchas those close parallelsto it) does appearfrequently.In the
discussedin Sec. I and II, that reallyrelateonly to literatureof economicsa closelyrelated,but not ex-
what has beencalledthe pure"futurity"component act counterpartis Muth's "hypothesisof rational
in valuation.Here, the valuationformulascan still expectations" [18]. Among the more euphonic,
be extremelyusefulin maintainingthe internalcon- though we feel somewhat less revealing,alterna-
sistencyof the reasoningand in suggesting(or criti- tives that have been suggested to us are "puta-
cizing) empirical tests of certain classes of hy- tive rationality"(by T. J. Koopmans),"bi-ration-
pothesesabout valuation,even thoughthe formulas ality" (by G. L. Thompson), "empathetic ra-
themselvescannot be used to grindout precisenu- tionality" (by Andrea Modigliani), and "pan-
mericalvaluesfor specificreal-worldshares. rationality"(by A. Ando).

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
428 THE JOURNALOF BUSINESS
metric market rationality differs from The irrelevanceof dividend policy de-
the usual postulate of rational behavior spite uncertainty.-In Section I we were
in severalimportantrespects.In the first able to show that, given a firm's invest-
place, the new postulate covers not only ment policy, its dividend policy was ir-
the choice behavior of individuals but relevant to its currentmarket valuation.
also their expectations of the choice be- We shallnow show that this fundamental
havior of others. Second, the postulate is conclusionneed not be modified merely
a statement about the market as a whole because of the presence of uncertainty
and not just about individual behavior. about the future courseof profits,invest-
Finally, though by no means least, sym- ment, or dividends (assuming again, as
metric market rationality cannot be de- we have throughout, that investment
duced from individual rational behavior policy can be regardedas separablefrom
in the usual sense since that sense does dividend policy). To see that uncer-
not imply imputingrationalityto others. tainty about these elements changes
It may, in fact, imply a choice behavior nothing essential, consider a case in
inconsistent with imputed rationality which currentinvestors believe that the
unless the individual actually believes futurestreamsof total earningsand total
the market to be symmetricallyrational. investment whatever actual values they
For if an ordinarilyrationalinvestor had may assume at different points in time
good reason to believe that other inves- will be identical for two firms, 1 and 2.26
tors would not behave rationally,then it Suppose further, provisionally, that the
might well be rationalfor him to adopt a same is believed to be true of future total
strategy he would otherwise have re- dividendpayments fromperiodone on so
jected as irrational. Our postulate thus that the only way in which the two firms
rules out, among other things, the possi- differ is possibly with respect to the
bility of speculative "bubbles" wherein prospective dividend in the current pe-
an individually rational investor buys a riod, period0. In terms of previousnota-
security he knows to be overpriced (i.e.,
tion we are thus assumingthat
too expensive in relation to its expected
long-runreturn to be attractive as a per- X1(t)= X2(t) t= . . .
manent addition to his portfolio) in the
expectation that he can resell it at a still A(l() =I(l) t =IO. .. oo
more inflated price before the bubble
bursts.25
21We recognize,of course,that such speculative capital markets.Needless to say, whetherour con-
bubbleshave actually arisenin the past (and will fidencein the postulateis justifiedis somethingthat
will have to be determinedby empiricaltests of its
probablycontinueto do so in the future),so that our implications(such as, of course,the irrelevanceof
postulatecan certainlynot be taken to be of univer- dividendpolicy).
sal applicability.We feel, however,that it is alsonot
of universalinapplicabilitysince from our observa- 26The assumptionof two identicalfirmsis intro-
tion, speculative bubbles, though well publicized duced for convenienceof expositiononly, since it
whenthey occur,do not seemto us to be a dominant, usuallyis easierto see the implicationsof rationality
or even a fundamental,featureof actual marketbe- when there is an explicit arbitragemechanism,in
haviorunderuncertainty.That is, we wouldbe pre- this case, switches between the shares of the two
pared to argue that, as a rule and on the average, firms. The assumption,however, is not necessary
marketsdo not behavein wayswhichdo not obvious- andwe can,if we like, thinkof the two firmsas really
ly contradictthe postulateso that the postulatemay correspondingto two states of the same firmfor an
still be useful, at least as a first approximation,for investorperforminga seriesof "mentalexperiments"
the analysis of long-run tendencies in organized on the subjectof dividendpolicy.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 429
the subscripts indicating the firms and all periods and of equal D(t) in period 2
the tildes being added to the variablesto and beyond. Clearly, the only way dif-
indicate that these are to be regarded ferences in dividends in period,1 can ef-
from the standpoint of current period, fect Rj(O)and hence Vj(O)is via Vi(l).
not as known numbers but as numbers But, by the assumption of symmetric
that will be drawnin the future from the market rationality, current investors
appropriate probability distributions. know that as of the start of period 1 the
We may now ask: "What will be the re- then investors will value the two firms
turn, k1(O)to the currentshareholdersin rationally and we have already shown
firm 1 during the current period?" that differencesin the current dividend
Clearly, it will be do not affect current value. Thus we
R1(O) = Db(O)+ VI(l) - hi1(1)phl(1). (28) must have f11(l) = 172(1)-and hence
V1(O)= V2(0)-regardless of any pos-
But the relation between D1(O) and sible difference in dividend payments
mi(1) pl(l) is necessarily still given by duringperiod 1. By an obvious extension
equation (4) which is merely an account- of the reasoning to Vi(2), fj(3), and so
ing identity so that we can write on, it must follow that the currentvalua-
AI(1) pil(l) = Ii(O) - [X1(?)- f)(0)], (29) tion is unaffected by differencesin divi-
dend payments in any future period and
and, on substituting in (28), we obtain thus that dividendpolicy is irrelevantfor
A(O) = X1(O)- Ii(O) + PO(i) (30) the determination of market prices,
given investment policy.27
for firm 1. By an exactly parallelprocess Dividendpolicy and leverage.-A study
we can obtain an equivalent expression of the above line of proof will show it to
for P2(O). be essentially analogousto the proof for
Let us now compareR1(0) with P2(O). the certaintyworld,in whichas we know,
Note first that, by assumption,X1(O)= firms can have, in effect, only two alter-
X2(O) and Il(O) = 12(0). Furthermore, native sources of investment funds: re-
with symmetric market rationality, the tained earnings or stock issues. In an
terminalvalues Vi(1) can dependonly on uncertain world, however, there is the
prospective future earnings, investment additional financing possibility of debt
and dividendsfromperiod 1 on and these issues. The question naturally arises,
too, by assumption,are identical for the therefore, as to whether the conclusion
two companies.Thus symmetric ration- about irrelevanceremains valid even in
ality implies that every investor must the presence of debt financing, particu-
expect fl(l) = V2(1) and hence finally larly since there may very well be inter-
L1(O) = R2(O). But if the return to the
27 We mightnote that the assumptionof symmet-
investors is the same in the two cases,
ric marketrationalityis sufficientto derivethis con-
rationality requires that the two firms clusionbut not strictly necessaryif we are willing
commandthe same currentvalue so that to weakenthe irrelevancepropositionto one running
V1(0)must equal V2(0) regardlessof any in termsof long-run,averagetendenciesin the mar-
ket. Individualrationalityalone could conceivably
differencein dividend payments during bringaboutthe latter,for over the longpull rational
period 0. Suppose now that we allow investors could enforce this result by buying and
dividends to differ not just in period 0 holding"undervalued" securitiesbecausethis would
insurethemhigherlong-runreturnswheneventually
but in period 1 as well, but still retain the pricesbecamethe same. They might, however,
the assumptionof equal ?$(t) and 11(t)in have a long, long wait.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
430 THE JOURNALOF BUSINESS
actionsbetween debt policy and dividend policy under uncertainty,we might take
policy. The answer is that it does, and note brieflyof a commonconfusionabout
while a complete demonstration would the meaning of the irrelevance proposi-
perhapsbe too tedious and repetitiousat tion occasioned by the fact that in the
this point, we can at least readily sketch real world a change in the dividend rate
out the main outlines of how the proof is often followedby a changein the mar-
proceeds. We begin, as above, by estab- ket price (sometimes spectacularly so).
lishing the conditions from period 1 on Such a phenomenonwould not be incom-
that lead to a situation in which f1(l) patible with irelevanceto the extent that
must be broughtinto equality with fV2(1) it was merely a reflectionof what might
where the V, following the approachin be called the "informationalcontent" of
our earlierpaper [17],is now to be inter- dividends,an attribute of particulardivi-
preted as the total market value of the dend payments hitherto excluded by as-
firm, debt plus equity, not merely equity sumptionfrom the discussionand proofs.
alone. The return to the original inves- That is, where a firm has adopted a pol-
tors taken as a whole-and remember icy of dividend stabilizationwith a long-
that any individualalwayshas the option established and generally appreciated
of buying a proportionalshare of both "target payout ratio," investors are
the equity and the debt-must corre- likely to (and have good reasonto) inter-
spondinglybe broadenedto allow for the pret a change in the dividend rate as a
interest on the debt. There will also be a change in management'sviews of future
corresponding broadening of the ac- profit prospects for the firm.29The divi-
countingidentity (4) to allow, on the one dend change, in other words, provides
hand, for the interest return and, on the the occasionfor the price change though
other, for any debt funds used to finance not its cause, the price still being solely a
the investment in whole or in part. The reflectionof future earnings and growth
net result is that both the dividend com- opportunities.In any particularinstance,
ponent and the interest component of of course, the investors might well be
total earningswill cancel out making the mistaken in placing this interpretation
relevant (total) return, as before, on the dividend change, since the man-
[li(O) - i(O)+ fi(l)] which is clearly agement might really only be changing
independent of the current dividend. It ing its payout target or possibly even
follows, then, that the value of the firm attempting to "manipulate" the price.
must also therefore be independent of But this wouldinvolve no particularcon-
dividend policy given investment pol- flict with the irrelevanceproposition,un-
icy.28 less, of course, the price changes in such
The informationalcontentof dividends. cases were not reversedwhen the unfold-
To concludeour discussionof dividend ing of events had made clear the true
28 This same conclusion must also hold for the nature of the situation.A0
current market value of all the shares (and hence 29 For evidence on the prevalence of dividend
for the current price per share), which is equal to
the total market value minus the given initially stabilization and target ratios see Lintner [15].
outstanding debt. Needless to say, however, the 30 For a further discussiQn of the subject of the

price per share and the value of the equity at future informational content of dividends, including its im-
points in time will not be independent of dividend plications for empirical tests of the irrelevance prop-
and debt policies in the interim. osition, see Modigliani and Miller [16, pp. 666-68].

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 431
V. DIVIDEND POLICY AND MARKET ket. If, for example, the frequency dis-
IMPERFECTIONS tribution of corporatepayout ratios hap-
To complete the analysis of dividend pened to correspondexactly with the dis-
policy, the logical next step would pre- tribution of investor preferencesfor pay-
sumably be to abandon the assumption out ratios, then the existence of these
of perfect capital markets. This is, how- preferenceswould clearlylead ultimately
ever, a good deal easier to say than to do to a situation whose implications were
principallybecausethere is no unique set differentin no fundamentalrespect from
of circumstances that constitutes "im- the perfect market case. Each corpora-
perfection."We can describenot one but tion would tend to attract to itself a
a multitude of possible departuresfrom "clientele" consisting of those preferring
strict perfection,singly and in combina- its particularpayout ratio, but one clien-
tions. Clearly, to attempt to pursue the tele would be entirely as good as another
implicationsof each of these would only in terms of the valuation it would imply
serve to add inordinately to an already for the firm.Nor, of course,is it necessary
overlong discussion. We shall instead, for the distributionsto match exactly for
therefore,limit ourselvesin this conclud- this result to occur. Even if there were a
ing section to a few brief and generalob- "shortage" of some particular payout
servations about imperfect markets that ratio, investorswould still normallyhave
we hope may prove helpful to those tak- the option of achieving their particular
ing up the task of extending the theory saving objectives without paying a pre-
of valuation in this direction. mium for the stocks in short supply
First, it is important to keep in mind simply by buying appropriatelyweighted
that fromthe standpointof dividendpol- combinationsof the more plentiful pay-
icy, what counts is not imperfectionper out ratios. In fact, given the great range
se but only imperfectionthat might lead of corporatepayout ratios known to be
an investor to have a systematic prefer- available, this process would fail to
ence as between a dollar of currentdivi- eliminate permanent premiums and dis-
dends and a dollar of current capital counts only if the distributionof investor
gains. Where no such systematic prefer- preferenceswere heavily concentratedat
ence is produced, we can subsume the either of the extreme ends of the payout
imperfectionin the (random)errorterm scale.3"
alwayscarriedalongwhen applyingprop- Of all the many market imperfections
ositionsderivedfromideal modelsto real- that might be detailed, the only one that
world events. would seem to be even remotely capable
Second, even where we do find imper- of producingsuch a concentrationis the
fections that bias individual preferences substantial advantage accorded to capi-
-such as the existence of brokeragefees tal gains as comparedwith dividendsun-
which tend to make young "accumula- 31The above discussion should explain why,
tors" prefer low-payout shares and re- among other reasons, it would not be possible to
draw any valid inference about the relative pre-
tired persons lean toward "income ponderance of "accumulators" as opposed to "in-
stocks"-such imperfectionsare at best come" buyers or the strength of their preferences
only necessary but not sufficient condi- merely from the weight attaching to dividends in a
simple cross-sectional regression between value and
tions for certain payout policies to com- payouts (as is attempted in Clendenin [2, p. 50] or
mand a permanentpremiumin the mar- Durand [5, p. 651]).

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
432 THE JOURNALOF BUSINESS
der the personal income tax. Strong as measure or even to detect any premium
this tax push toward capital gains may for low-payout shares on the basis of
be for high-incomeindividuals,however, standardstatistical techniques.
it should be rememberedthat a substan- Finally, we may note that since the
tial (and growing)fractionof total shares tax differentialin favor of capital gains is
outstanding is currently held by inves- undoubtedlythe majorsystematicimper-
tors for whom there is either no tax dif- fection in the market, one clearly cannot
ferential (charitableand educational in- invoke "imperfections"to account for
stitutions, foundations, pension trusts, the difference between our irrelevance
and low-income retired individuals) or propositionand the standard view as to
where the tax advantage is, if anything, the role of dividend policy found in the
in favor of dividends (casualty insurance literature of finance. For the standard
companiesand taxable corporationsgen- view is not that low-payout companies
erally). Hence, again, the "clientele ef- command a premium; but that, in gen-
fect" will be at work. Furthermore,ex- eral, they will sell at a discount!33If such
cept for taxable individuals in the very indeed were the case-and we, at least,
top brackets, the required differencein are not preparedto concedethat this has
before-taxyields to produceequal after- been established-then the analysis pre-
sented in this paper suggests there would
tax yields is not particularlystriking, at
be only one way to account for it; name-
least for moderatevariationsin the com-
ly, as the result of systematic irrational-
position of returns.32
All this is not to say, ity on the part of the investing public.34
of course,that differencesin yields (mar- To say that an observedpositive pre-
ket values) caused by differencesin pay- mium on high payouts was due to irra-
out policies should be ignored by man- tionality would not, of course,make the
agements or investors merely because phenomenonany less real. But it would
they may be relatively small. But it may at least suggest the need for a certain
help to keep investigatorsfrom being too measureof caution by long-rangepolicy-
surprised if it turns out to be hard to makers. For investors, however naive
32For example,if a taxpayeris subjectto a mar- they may be when they enter the market,
ginal rate of 40 per cent on dividendsand half that do sometimeslearn from experience;and
or 20 per cent on long-termcapitalgains, then a be- perhaps,occasionally,even from reading
fore-taxyield of 6 per cent consistingof 40 per cent articles such as this.
dividendsand 60 per cent capitalgainsproducesan
after-taxyieldof 4.32percent.To net the sameafter- 33 See, amongmany, many others,Gordon[8, 91,

tax yield on a stockwith 60 per cent of the returnin Grahamand Dodd [11,esp. chaps.xxxivand xxxvi],
dividendsand only40 percent in capitalgainswould Durand [4, 5], Hunt, Williams,and Donaldson[13,
requirea before-taxyield of 6.37percent. The differ- pp. 647-49], Fisher [7], Gordonand Shapiro[10],
ence would be somewhatsmallerif we allowedfor Harkavy[12],Clendenin[2],Johnson,Shapiro,and
the presentdividendcredit,thoughit shouldalso be O'Meara[14],and Walter[19].
kept in mind that the tax on capitalgains may be 4 Or, less plausibly,that there is a systematic
avoided entirelyunderpresentarrangementsif the tendencyfor externalfunds to be used more pro-
gains are not realizedduringthe holder'slifetime. -ductively than internalfunds.

REFERENCES
1. BODENHORN,DIRAN. "On the Problem of 2. CLENDENIN,JOHN. "What Do Stockholders
Capital Budgeting," Journal of Finance, Like?" California Management Review, I
XIV (December, 1959), 473-92. (Fall, 1958), 47-55.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions
 THE VALUATIONOF SHARES 433
3. CLENDENIN, JOHN, and VAN CLEAVE, M. 13. HUNT, PEARSON, WILLIAMS, CHARLES, and
"Growth and Common Stock Values," DONALDSON, GORDON.Basic Business Fi-
Journal of Finance, IX (September, 1954), nance. Homewood, Ill.: Richard D. Irwin,
365-76. 1958.
4. DURAND, DAVID. Bank Stock Prices and the 14. JOHNSON, L. R., SHAPIRO, ELI, and
Bank Capital Problem. ("Occasional Pa- O'MEARA, J. "Valuation of Closely Held
per," No. 54.) New York: National Bureau Stock for Federal Tax Purposes: Approach
of Economic Research, 1957. to an Objective Method," University of
5. . "The Cost of Capital and the The- Pennsylvania Law Review, C, 166-95.
ory of Investment: Comment," American 15. LINTNER, JOHN. "Distribution of Incomes
EconomicReview,XLIX (September, 1959), of Corporations among Dividends, Re-
639-54. tained Earnings and Taxes," American
6. . "Growth Stocks and the Peters- Economic Review, XLVI (May, 1956), 97-
burg Paradox," Journal of Finance, XII 113.
(September, 1957), 348-63. 16. MODIGLIANI, FRANCO, and MILLER, MER-
7. FISHER, G. R. "Some Factors Influencing TON. "'The Cost of Capital, Corporation
Share Prices," Economic Journal, LXXI, Finance and the Theory of Investment,':
No. 281 (March, 1961), 121-41. Reply," American Economic Review,XLIX
8. GORDON, MYRON. "Corporate Saving, In- (September, 1959), 655-69.
vestment and Share Prices," Reviewof Eco- 17. . "The Cost of Capital, Corpora-
nomics and Statistics (forthcoming). tion Finance and the Theory of Invest-
9. . "Dividends, Earnings and Stock ment," ibid., XLVIII (1958), 261-97.
Prices," ibid., XLI, No. 2, Part I (May, 18. MTUTH, JOHN F. "Rational Expectations
1959), 99-105. and the Theory of Price Movements,"
10. GORDON,MYRON, and SHAPIRO,ELI. "Capi- Econometrica(forthcoming).
tal Equipment Analysis: The Required 19. WALTER, JAmES E. "A Discriminant Func-
Rate of Profit," Management Science, III, tion for Earnings-Price Ratios of Large In-
1956, 102-10. dustrial Corporations," Review of Econom-
11. GRAHAmi, BENJAMIN, and DODD, DAVID. ics and Statistics, XLI (February, 1959),
Security Analysis. 3d ed. New York: Mc- 44-52.
Graw-Hill Book Co., 1951. 20. . "Dividend Policies and Common
12. HARKAVY, OSCAR, "The Relation between Stock Prices," Journal of Finance, XI
Retained Earnings and Common Stock (March, 1956), 29-41.
Prices for Large Listed Corporations," 21. WILLIAMS, JOHN B. The Theory of Invest-
Journal of Finance, VIII (September, 1953), ment Value. Cambridge, Mass.: Harvard
283-97. University Press, 1938.

This content downloaded from 130.226.87.142 on Wed, 11 Sep 2013 08:33:39 AM

All use subject to JSTOR Terms and Conditions