AUTOMATIC DIVIDEND REINVESTMENT PLANS
OF NONFINANCIAL CORPORATIONS
RODNEY PHIL MALONE
A DISSERTATION PRESENTED TO THE GRADUATE
COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL
FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
I sincerely wish to express my appreciation to Dr. Richard
H. Pettway, chairman of the dissertation supervisory committee,
for his advice, suggestions, and considerations in reading and re-
turning rough drafts quickly. His efforts enabled me to make nec-
I also wish to thank Dr. Eugene Brigham for his knowledge-
able suggestions which have made this dissertation a more meaning-
ful research effort. Dr. Warren Menke and Dr. Jack Vernon were
helpful in reading the dissertation and in serving on my super-
visory committee. Dr. Fred Arditti's course in financial theory
was of great benefit in providing me with necessary background
information. Dr. C.A. Matthews was instrumental in guiding and
advising me throughout my graduate program. Dr. Al Wildt rendered
assistance with the mechanics and interpretation of discriminant
My wife, Charlotte, was helpful in reading, making correc-
tions, typing and providing needed encouragement. My mother's
patience and love have always helped me obtain my goals.
Special gratitude and thanks are extended to the banks and
firms for answering questionnaires and providing information.
TABLE OF CONTENTS
LIST OF TABLES v
I. INTRODUCTION 1
Purpose of the Study 2
Limitations of the Study 3
Organization of the Study 4
II. THE NATURE OF DIVIDENDS 5
Early Controversy in the Literature 5
The Miller and Modigliani Model 9
The Gordon Model 16
Empirical Studies 19
Early Regression Models 20
The Durand model 20
The Gordon model 21
The Brigham and Gordon model 22
Critique of Regression Models 24
Other Empirical Studies 26
III. METHOD OF RESEARCH 36
The Collection of Information 36
The Analysis of the Data 42
The Multiple Discriminant Model 43
Multiple discriminant analysis 43
The model 44
Selection of variables 45
The sample 47
The Multiple Regression Model 48
IV. CHARACTERISTIC PROFILES OF ADR USERS 52
The Collection of Information 53
The Generation of Samples 54
Empirical Findings 55
Test of the Difference Between Two Means 55
Multiple Discriminant Analysis 58
Problems and Limitations 61
V. SURVEY FINDINGS 70
The Nature of Automatic Dividend Reinvestment Plans 70
Operating Characteristics of ADR Plans 70
Legal Considerations 75
Advantages and Disadvantages 76
Questionnaire Results 78
Questionnaire I 79
Questionnaire II 88
VI. CHARACTERISTICS OF FIRMS WITH HIGH 95
Selection of Variables 95
Age of Plans 97
Market Conditions 99
Shareholder Characteristics 99
The Samples 100
Selection of Best Equations 101
Empirical Findings of the Multiple Regression Model 102
Problems and Limitations 106
VII. THEORETICAL IMPLICATIONS 112
The Clientele Theory 118
VIII. SUMMARY 121
Collection of Information 121
Profiles of ADR Users 122
Characteristics of Firms with High Shareholder 123
Applicability to the Theory of Dividends 124
Suggestions for Further Research 125
BIOGRAPHICAL SKETCH 135
LIST OF TABLES
I FREQUENCY OF RESPONSES OF EXECUTIVES TO FACTORS 33
INFLUENCING DIVIDEND POLICY
II Z VALUES OF INDUSTRIAL FIRMS CALCULATED FOR 57
III SUMMARY STATISTICS OF STEPWISE MULTIPLE 59
DISCRIMINANT ANALYSIS FOR INDUSTRIALS
IV CLASSIFICATION MATRIX FOR INDUSTRIAL HOLDOUT SAMPLE 59
V SUMMARY STATISTICS OF STEPWISE DISCRIMINANT 60
ANALYSIS FOR UTILITIES
VI CLASSIFICATION MATRIX FOR UTILITY HOLDOUT SAMPLE 60
VII COST TO PARTICIPANTS FOR VARIOUS TYPES OF 75
VIII YEARS IN WHICH FIRMS RESPONDING TO THE 80
QUESTIONNAIRE ESTABLISHED ADR PLANS
IX ESTIMATED PARTICIPATION BY PERCENTAGE OF 81
SHAREHOLDERS FOR 1973 and 1974
X THE ESTIMATED PERCENTAGE OF SHARES OUTSTANDING 82
PARTICIPATING IN ADR PLANS FOR 1973 and 1974
XI THE ESTIMATED ANNUAL DOLLAR VOLUME OF FUNDS 84
INVESTED THROUGH ESTABLISHED ADR PLANS
XII REASONS GIVEN FOR ESTABLISHING AN ADR PLAN 85
XIII RESPONSES AS TO PRESENT INCLINATIONS OF THE 87
POSSIBILITY OF USING NEW STOCK RATHER THAN
OUTSTANDING COMMON IN ADR PLANS
XIV CURRENT POSITIONS OF FIRMS WITH RESPECT TO 89
IMPLEMENTING ADR PLANS
XV REASONS GIVEN FOR NOT ESTABLISHING AN ADR PLAN 91
XVI RESPONSES AS TO WHETHER OR NOT IT WAS LIKELY THAT 93
THE FIRM WOULD IMPLEMENT AN ADR PLAN WITHIN 3 YEARS
XVII REGRESSION COEFFICIENTS AND t STATISTICS IN THE 102
FINAL EQUATION FOR 34 INDUSTRIAL FIRMS WHEN THE
DEPENDENT VARIABLE WAS THE PERCENTAGE OF SHARE-
XVIII REGRESSION COEFFICIENTS AND t STATISTICS IN THE 103
FINAL EQUATION FOR 28 INDUSTRIAL FIRMS WHEN THE
DEPENDENT VARIABLE WAS THE PERCENTAGE OF SHARES
XIX PARTIAL CORRELATION MATRIX FOR VARIABLES IN THE 104
FINAL EQUATIONS FOR INDUSTRIAL FIRMS WITH
COEFFICIENTS OF CORRELATION GREATER THAN 30
XX REGRESSION COEFFICIENTS AND t STATISTICS IN THE 105
FINAL EQUATION FOR 15 UTILITIES WHEN THE DEPENDENT
VARIABLE WAS THE PERCENTAGE OF SHAREHOLDERS
XXI REGRESSION COEFFICIENTS AND t STATISTICS IN THE 105
FINAL EQUATION FOR 15 UTILITIES WHEN THE DEPENDENT
VARIABLE WAS THE PERCENTAGE OF SHARES PARTICIPATING
XXII PARTIAL CORRELATION MATRIX FOR VARIABLES IN THE 107
FINAL EQUATION FOR PUBLIC UTILITIES WITH
COEFFICIENTS OF CORRELATION GREATER THAN 30
Abstract of Dissertation Presented to the
Graduate Council of the University of Florida in Partial
Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
AUTOMATIC DIVIDEND REINVESTMENT PLANS
OF NONFINANCIAL CORPORATIONS
Rodney Phil Malone
Chairman: Richard H. Pettway
Major Department: Finance, Insurance, Real Estate and
Urban Land Studies
While a feature allowing shareholders to automatically re-
invest dividends has long been an option available to share-
holders of savings and loan associations, mutual savings banks
as well as some mutual funds and life insurance companies, non-
financial corporations did not begin to provide such options
until the late 1960s. The automatic dividend reinvestment (ADR)
movement among nonfinancial firms had only a few users until the
early 1970s; however, in 1972 and 1973 the movement seemed to
pick up considerable momentum. In fact, this study was able to
identify 265 nonfinancial corporations that had implemented an
automatic dividend reinvestment plan as of October, 1973.
In identifying firms that were ADR users and in gathering
information on the cost and mechanics of ADR plans, data were
collected from Fortune's Fifty Largest Commercial Banks. Also, a
questionnaire survey was conducted of the Standard and Poor (S&P)
500. Data from the bank survey and questionnaire were presented
and analyzed in order to provide information to aid financial
executives in deciding whether or not their firm could benefit
from implementing an ADR plan. In doing so, profiles of the fi-
nancial characteristics of firms that already had implemented ADR
plans were developed with the aid of stepwise multiple discrimi-
nant analysis and statistical hypothesis testing. Profiles of
firms that had realized the greatest degree of utilization in
terms of shareholder participation were also developed. Multiple
regression analysis was used to develop these profiles. Profiles
were developed for the S&P 425 industrials and the 55 public util-
ities in each of the two models. In comparison to nonADR users,
firms with established ADR plans were found to have a higher di-
vidend payout ratio and lower growth rates, price earnings ratios,
and debt to net worth ratios. Also, there appeared to be a ten-
dency for users to concentrate within a few S&P industrial classi-
fications. Participation rates, on the other hand, appeared to be
greater for large firms with higher growth characteristics.
Responses from the survey provided further information from
firms that were users and nonusers of ADR plans. Information was
gathered from nonusers on the degree of consideration and the
probable decision that would be made in regard to whether or not
an ADR plan should be established. Firms which indicated that the
concept had been rejected, provided information on the reason for
rejection. Users provided information such as the reasons for
implementing and the degree of utilization that they had experienced.
Also, efforts were made to gauge their reaction to whether or not
the ADR experience could have bearing on the firm's dividend policy.
In addition, the study sought to ascertain the potential of innova-
tive changes from the traditional type of ADR plan offered such as
firm-administered rather than bank-administered plans and the issu-
ance of new shares of common stock through ADR plans rather than
purchasing already outstanding issues.
The data from the questionnaire indicated that considerable
potential exists for the growth of the traditional type ADR plan
which is bank-administered and deals in outstanding issues of common
stock. There appeared to be little acceptance to date of firm-
administered plans or plans dealing in new issues of common stock.
Finally, the study sought to determine whether or not informa-
tion gathered from the ADR experience had any applications to divi-
dend theory. In doing so, the models of Miller and Modigliani (M&M)
were contrasted with Gordon. Empirical studies were also presented
which had bearing upon the question of the "importance" or
"unimportance" of the dividend decision in maximizing the value of
the firm. In particular, "the clientele theory" of M&M and their
assumption of "rationality" on the part of investors were examined.
Neither of these premises was found to be contradicted by the ADR
In 1967 Allegheny Power System Incorporated implemented an auto-
matic dividend reinvestment (ADR) plan and allowed participating share-
holders to apply their dividend checks to the purchase of outstanding
shares of the firm's common stock. While dividend reinvestment had
long been available to shareholders of mutual funds, savings and loan
associations, mutual savings banks and some life insurance companies,
Allegheny Power became the first nonfinancial corporation to allow a
dividend reinvestment option. In 1969 American Telephone and Tele-
graph became the first of the corporate giants to provide its share-
holders this option. The American Telephone and Telegraph Plan origi-
nated when the firm accepted an offer by First National City Bank of
New York to act as administrative agent for ADR plans. Since 1969 an
increasing number of firms have provided the mechanism for the auto-
matic reinvestment of dividends by shareholders into the common stock
of the firm. This study was able to identify 265 nonfinancial firms1
which, as of October 1973, had established an ADR plan.
As a consequence of the rapid emergence of ADR plans and the lack
of adequate information many financial executives may be having diffi-
culty in evaluating answers to questions such as: Why should a dividend
reinvestment plan be established? What, in effect, constitutes a suc-
The figure was obtained from a questionnaire survey of the
Standard and Poor 500 and from correspondence with Fortune's Fifty
cessful ADR plan? Is there any inherent information in the degree of
success realized with an ADR plan that should be considered in making
other decisions related to dividend policy?
Purpose of the Study
The purpose of this study is to answer questions of this nature
and to draw insights from the birth and growth of the ADR movement by:
(1) describing the mechanism of these plans; (2) calculating the cost
to participating shareholders; (3) determining the motivation for firm
implementation; (4) ascertaining the rationale for shareholder partic-
ipation; and (5) reviewing implications of the findings from the ADR
movement to the theory of corporate dividend policy. Pertinent data
(i.e., number of firms, number of shareholders subscribing to the
service, the characteristics of firms that have adopted ADR plans, etc.)
were collected and analyzed. Based upon the data, the study attempted
to determine the effects of ADR plans upon the dividend policy
decisions of the firm. Also, profiles were developed of the type of
firm which most commonly implements ADR plans as well as the type of
firm which realizes the greatest degree of shareholder participation.
From the data and subsequent analysis, inferences were drawn as to the
consistency of the ADR movement with contemporary dividend practice
Several statistical techniques were employed in analyzing the
data generated for the study. Techniques of simple data description
were employed to summarize the results to a questionnaire which was
circulated to the Standard and Poor 500.2 Two models were developed
and tested to determine the respective profiles of firms which do and
do not have an established ADR plan as well as profiles of firms which
historically have realized the greatest degree of shareholder partici-
pation. A test of significance of the difference between two means
and multiple discriminant analysis were employed to develop profiles
of users and nonusers. Multiple regression analysis was the principal
tool utilized for developing profiles of firms with the highest degree
Many of the required data was obtained from a questionnaire
survey and personal communications. The Compustat Tapes provided sup-
plementary data related to the financial characteristics of individual
firms. Since the Standard and Poor 500 was selected as the population
of firms to be studied, the questionnaire survey was distributed to
firms belonging to that population. The more innovative firms, such
as American Telephone and Telegraph and Allegheny Power, and Fortune's
Fifty Largest Banks3 were contacted individually. The selection and
calculation of relevant financial variables taken from the Compustat
Tapes will be discussed in detail in the subsequent analysis.
Limitations of the Study
Several procedural problems and shortcomings were encountered in
the collection of information and subsequent statistical techniques
2Outlook, "Stocks in the S&P '500' Price Index" (Standard and
Poor, January, 1974).
Fortune, "The Fortune Directory of: The Fifty Largest Commer-
cial Banking Companies" (Vol. LXXXVIII); Time Inc., July, 1973), pp.
employed in the study. The fact that a substantial portion of the data
came from a questionnaire created concern over response bias. Also the
use of Compustat data is subjected to criticism due to the manner in
which some data items are defined. Other limitations resulted from de-
fining the population of firms offering ADR plans as the Standard and
Poor 500 and the population of banks acting as administrative agents as
the Fortune Fifty. Finally, statistical problems were encountered in
both the multiple discriminant and the multiple regression models. Mul-
ticollinearity was of particular concern in the multiple regression
model. Errors in the a priori classification of firms as ADR users and
nonusers constituted a potential problem area in the discriminant
Organization of the Study
The introduction is followed by Chapter II which summarizes the
theoretical and empirical research that has been published on the im-
portance of dividends as a policy decision. In particular, the work
of Gordon as well as Miller and Modigliani is presented and contrasted.
The third chapter deals with methodology such as the development of
the two models as well as the intent and design of'the questionnaire.
Chapters IV, V and VI contain the empirical results of the study. Mul-
tiple discriminant analysis and a test for statistical difference in
two sample means were employed in Chapter IV to develop profiles of
users and nonusers. Chapter V analyzes the results of the survey of
banks and firms. Profiles of firms having high ADR participation
rates are presented in Chapter VI. Theoretical implications of the
ADR experience are discussed in Chapter VII. The final chapter is a
THE NATURE OF DIVIDENDS
Two opposing viewpoints are evident in the financial literature
on the theoretical importance of dividends in answer to the question
of whether or not there is an optimal dividend policy decision which
will maximize the financial worth of the firm. Many scholars in eco-
nomics and finance have contributed to the controversy. This study
will not attempt to present all of the models and equations relevant
to the topic. Instead Myron Gordon's model will be contrasted with
that of Merton Miller and Franco Modigliani. Then other significant
contributions by Walter, Durand, and Lintner will be presented and the
statistical findings of several empirical studies will be examined.
Finally, a survey will be undertaken to draw upon the observed behav-
ior of the firm in this question of the importance or insignificance
of dividend policy in the framework of maximizing the value of the
Early Controversy in the Literature
Early studies and models of the theory of dividends relied
heavily upon the concept of present value. Some of these early
theorists drew upon the work of John Burr Williams1 and viewed the
worth of the firm as the present value of spinoffs from the investment
to the stockholder. Others concluded that earnings were the appropri-
1John Burr Williams, The Theory of Investment Value (Harvard: The
Harvard University Press, 1938).
ate stream to discount in arriving at a market price for the firm. The
major area of disagreement between these two rival groups of theorists
appeared to be related to the form of shareholder returns, dividends
or capital gains, and to timing in an uncertain world. One group whom
Lintner2 coined "pure earnings" theorists contended that the market
value of a firm's common stock is solely a function of earnings and
not influenced by dividends. Identified by Lintner as belonging to
this group were Durand, the Lutzes, Solomon, Roberts, Kuh, Dean,
Weston, Miller and Modigliani. The opposition featured Walter, Gordon,
Shapiro, Bierman, Fouraker, Jaedicke, and Lintner. These theorists
rallied around Williams' present value formulation. Lintner also
called attention to the fact that Williams' work probably drew upon
earlier writings of economists such as Irving Fisher, Samuelson, Hicks,
Tinbergen, and Preinreich.
One of the earliest theoretical attempts at determination of an
optimal dividend policy was established by Walter3 in 1956 with the
D + R (E D) E R R
c a c
(2-1) Vc R + R (E D)
c c c
where D is the cash dividend, E is earnings, Ra is the marginal return
on a dollar of investment in the firm, Rc is the market discount rate,
and Vc is market value (present value) of the firm's common stock.
According to Van Home:
John Lintner, "Dividends, Earnings, Leverage, Stock Prices, and
the Supply of Capital to Corporations," Review of Economics and Statis-
tics, XLIV (August, 1962), pp. 243-269.
3James E. Walter, "Dividend Policies and Common Stock Prices,"
The Journal of Finance, XI (March, 1956), p. 32.
The Walter formula implies that the optimal dividend
payout should be determined solely by the profitability of
investments. If the firm has an abundance of profitable in-
vestment opportunities, there should be no cash dividends,
for the earnings are needed to finance these opportunities.
On the other hand, if the firm has no profitable investment
opportunities, all earnings should be distributed to stock-
holders in the form of dividends. In this case, the funds
are not needed for financing.
The treatment of dividend policy as a passive residual
determined strictly by the availability of acceptable invest-
ment proposals implies that dividends are irrelevant; the in-
vestor is indifferent between dividends and capital gains. If
investment opportunities promise a return greater on the
equity-financed portion than the required return, k the in-
vestor would prefer to have the company retain earnings. If
the return is equal to the required return, he would be indif-
ferent between retention and dividends. Contrary, if the
return were less than the required return, he would prefer
Prior to the contention by Graham and Dodd in their third edition
of Security Analysis: Principles and Technique, that the market price
of stock was a function of both earnings and dividends, the predominant
view appeared to be that the price of the firm was solely a function of
dividends. The assertion by Graham and Dodd was unsubstantiated by
empirical analysis; nevertheless, their allegation intensified the con-
troversy that already existed. Several empirical studies were thus
stimulated. The "pure earnings" theorists, in particular, Walter,
Miller and Modigliani probably reacted in defense to two empirical
studies by Durand6 and Gordon7 in the late 1950s. Both studies employed
4James C. Van Home, Financial Management and Policy (2nd ed.;
Englewood Cliffs, N.J.: Prentice-Hall, Inc., 1971), pp. 244-245.
Benjamin Graham and David L. Dodd, Security Analysis: Principles
and Technique (3rd ed.; New York: McGraw-Hill Book Co., Inc., 1951),
6David Durand, Bank Stock Prices and the Bank Capital Problem, Oc-
casional Paper No. 54 (New York: National Bureau of Economic Research,
7Myron J. Gordon, "Dividends, Earnings, and Stock Prices," Review
of Economics and Statistics, XLI (May, 1959), pp. 99-105.
multiple regression analysis of cross-sector data and yielded the con-
clusion that dividends were a significant factor in influencing the
market price of common shares and the value of the firm. In retalia-
tion to the studies and in defense of the position of the "pure
earnings" theorists Miller and Modigliani advanced the irrelevance
position. The essence of the argument was not entirely new; among
others, Walter, and the Lutzes10 had earlier taken a similar stance.
The Miller and Modigliani contribution was, however, fully developed
and precisely stated. The arbitrage methodology and explicit statement
of assumptions along with the element of completeness resulted in an
important contribution to valuation theory. The irrelevance position
taken by Miller and Modigliani stood in sharp contrast to the position
of Gordon and others who opposed the "pure earnings" theorists. Gordon,
for example, asserted that dividends do matter and that, in general,
investors prefer dividends now to capital gains in the future or
greater future dividends for the same reason that "a bird in the hand
is more valuable than two in the bush." In the Gordon framework
dividends increase the value of the firm more than capital gains which
result from the reinvestment of earnings. Miller and Modigliani even-
tually reached the reverse conclusion that if there is any preference
Merton H. Miller and Franco Modigliani, "Dividend Policy, Growth,
and the Valuation of Shares," Journal of Business, XXXIV (October, 1961),
9Walter, pp. 29-41.
1Friedrich August Lutz and Vera Lutz, The Theory of Investment of
the Firm (Princeton, N.J.: Princeton University Press, 1951), pp. 213-
1Myron J. Gordon, "Optimal Investment and Financing Policy," The
Journal of Finance, XVIII (May, 1963), pp. 264-272.
for dividends or capital gains that preference results only because of
tax laws and favors the retention of earnings. The Miller and
Modigliani (M&M) arguments and their model will now be presented.
The Miller and Modigliani Model
The initial M&M model made the following three assumptions:
(1) Rationality Investors prefer more wealth to less
and are indifferent as to the form in
which it is received.
(2) Perfect Capital Markets Investors have costless and
equal access to relevant share informa-
tion; there are no taxes, no transfer
charges, and no single investor can af-
fect the price of a security.
(3) Perfect Certainty Not only are the returns on all
ventures known, but they are also equal
since there is no elemenp of risk where
events are predictable.
Under these assumptions the value of the firm at time zero V(0)
is the present value of the dividend paid at the end of that period
D(0) plus the present value of the price that would be received at the
beginning of the next period if the firm were sold V(1). The value of
the firm for any time period V(t) is equal to the price of a share of
stock in that time period t times the number of shares of common stock
outstanding n(t) at the beginning of the period. The following rela-
tionship between share price and market value of the firm should hold
where k is the appropriate discount rate:
(2-2) V(t) = p(t)n(t)
(2-3) V(0) = 1 [D(0) + V(1)].
1 + k
Equation (2-3) holds only if no new shares are issued during the period.
1Miller and Modigliani, p. 412.
If new shares are offered at the end of a period, the value must be ad-
justed for the new shares offered. If m(t) is equal to the new shares
offered at the end of period t-l, the value of the firm becomes the
price of a share times the total shares outstanding. Thus equation
(2-2) can be rewritten:
(2-4) V(t) = p(t)n(t-l) + p(t)m(t-l)
(2-5) p(l)n(l) = p(l)n(0) + p(l)m(0)
equation (2-3) could then be restated:
(2-6) V(O) = [D(0) + p(l)n(l) m(O)p(l)]
1 + k
(2-7) V(O) = 1 [D(0) + V(1) p(l)m(0)].
Equation (2-7) simply states that value is the present worth of divi-
dends and capital gains. It also differentiates between the value of
the firm to all shareholders V(1) and the value to new shareholders
p(l)m(0). The difference, V(1) p(l)m(0), is of course the value
that is attributable to shareholders from the previous period.
The net effect of equation (2-7) is to show that the value of the
firm in period t is affected only through the dividend in period t and
is independent of the dividend in periods beyond t. M&M then develop
an arbitrage argument based upon a given investment for two firms iden-
tical in all respects except for dividend policies. If a firm is trying
to maintain a given investment level I(t), then the greater the divi-
dend D(t) the greater the amount of new shares m(t+l) that must be is-
sued. The question is then, "Does the increased dividend D(t) more than
offset the loss in value that results from issuing m(t+l) new shares of
common stock?" If X(t) is the income generated by the firm in period t,
the retention of earnings for investment by the firm is income minus the
dividend or X(t) D(t). Provided that investment is at least as great
as retention, then new stock sales must account for any difference
between investment and retention. This relationship is depicted by the
(2-8) I(t) (x(t) D(t)) = p(t + l)m(t).
That is, investment minus retention equal new stock financing required
to maintain the desired level of investment I(t). Substituting in equa-
tion (2-7) the following results:
(2-9) V() = 1 [D(0) + V(l) I(0) + X(0) D(0)]
1 + k
(2-10) V(0) = V(l) 1(0) + X(0)]
and the general form of the equation becomes:
(2-11) V(t) = 1 [V(t + 1) I(t) + X(t)].
Since all of these terms are independent of dividend policy, the value
of the firms, V(t), is independent of present dividend policy. What
about future dividend policy? That is, "Can D(t+l) affect the present
value of the firm?" The answer is apparent when equation (2-11) is
solved for the value of the firm in period t + 1 since dividends must
affect V(t + 1) in order to affect V(t). This reformulation is as fol-
(2-12) V(t + 1) = [V(t + 2) l(t + 1) + X(t + 1)].
1 + k
Consequently, dividend policy in period t + 1 has no affect upon the
value of the firm in that period, and future dividend policy can exert
no influence except through the value of the firm in period t + 2.
Equation (2-11) can likewise be reformulated to show that the value of
the firm in all future periods is independent of dividend policy in
that period. Of course, M&M's results are only valid when the assump-
tions of their model are realistic. In essence, the results stem from
the assumption of a given level of investment I(t). In maintaining
that level of investment at any time, the firm that paid a dividend
found it necessary to sell additional shares of stock. The new share-
holders now own that share of the firm that was distributed as divi-
dends. The old shareholders received a dividend that was of value to
them but lost a share of ownership in the firm equal to the assets
distributed to pay the dividends. The result is that the total value
of the firm to old shareholders, V(t), is independent of the firm's
Now M&M relax the initial assumption of uncertainty and allow
variables to be random elements. They assume that the firm has no debt
in its capital structure and that dividends do not operate in a
bayesian sense of providing additional information about either
dividends or earnings in future time periods. By considering the ef-
fect of dividend policy upon the value of two firms which are identi-
cal in all respects except for dividend policy, M&M are able to reach
the same conclusion they did in a world of certainty. The assumption
of a given investment policy is once again paramount. To account for
the fact that there are two firms, each variable will now have a
double subscript. The first subscript j will denote the particular
firm and the second subscript t will again denote the period of time.
Equation (2-11) can be reformulated to differentiate between the two
firms as follows:
(2-13) V(jt) = + [V(jt+l) I(jt) + X(jt)].
The assumptions imply in an uncertain world (with o denoting uncertain-
(2-14) X(lt) = X(2t) for all t
(2-15) I(lt) = I(2t) for all t.
That is, investments and income are random variables and are the same
for both firms. M&M then allow one of the firms to pay all its earn-
ings out as dividends the first period; otherwise, the two firms will
have the same dividend policy in all future periods (n).
(2-16) D(lt) = D(2t) for all t except t = 0.
Thus, from the end of period 0 to n, the firms are identical in all
respects. The total return for present shareholders of firm j at the
end of period t is also a random variable R(jt). Also this return
variable, though random, is comparable to V(jt) in (2-13) except R is
used to denote an anticipated return to current shareholders. Equation
(2-13) thus becomes:
(2-17) R(jt) 1 [V(jt+l) I(jt) + X(jt)].
1 + k
Solving for period zero for both firms:
(2-18a) R1(0) = Vl(1) I1(0) + Xl(0)
(2-18b) R2(0) = V2(0) 12(0) + X2(0).
Equations (2-14) and (2-15) have already established the fact that the
investment and income for the two firms are identical by assumption.
Since after the first period the two firms are identical in all respects
(2-19) Vl(1) = V2(1)
(2-20) ) = 2().
(2-20) R1(0) = R2(0).
With wealth or gross return for the two firms identical after the first
period, the value of equation (2-17) is equal for both and is independ-
ent of dividend policy in all future periods. Thus both firms should
command the same price in a rational market. A recursive proof is also
applicable when the dividend policy is allowed to differ in periods
beyond time zero. The conclusion is still that dividends do not matter
even when the assumption of certainty is relaxed.
Of course with the valuation model developed by M&M13 to prove
that the cost of capital is independent of capital structure, the inclu-
sion of debt into the dividend framework presents no special problem.
Since the cost of capital is independent of capital structure, it is
irrelevant within the M&M framework whether debt or stock sales are
used to finance investments necessitated by paying dividends.
Several potential shortcomings may exist with the M&M model. One
argument is that dividends may provide additional information in a
bayesian context about future dividends and earnings. In this frame-
work D(t) provides information about V(t + 1) which in turn affects
V(t). Secondly, M&M assume away brokerage fees and reason that if an
individual has a preference for dividends, he need only sell enough
shares of his stock to satisfy that preference. Likewise an investor
preferring capital gains could simply reinvest the dividend. Even if
brokerage fees, etc. may prevent such activity, M&M contend that it is
still possible that the market behavior might be little different than
it would be in absence of all transfer charges. M&M argue:
13Franco Modigliani and Merton H. Miller, "The Cost of Capital,
Corporation Finance, and The Theory of Investment," American Economic
Review, XLVIII (June, 1958), pp. 261-297.
If, for example, the frequency distribution of corporate-
payout ratios happens to correspond exactly with the dis-
tribution of investor preferences for payout ratios, then
the existence of these preferences would clearly lead
ultimately to a situation whose implications were differ-
ent in no fundamental respect from the perfect market
case. Each corporation would tend to attract to itself a
"clientele" consisting of those preferring its particular
payout ratio, but one clientele would be entirely as good
as another in terms of the valuation it would imply for
Taxes may constitute another problem area for the M&M model. The
irrelevance argument assumed no taxes. The problem area is not the
mere existence of taxes but the existing differential between taxes on
personal income (dividend income) and capital gains. M&M admit that
the existence of taxes or at least the differential creates some prob-
lems for them and, in fact, concede that shareholders may have a
preference for income from capital gains over dividend income because
of the lower tax rate that some investors may pay on capital gains.
Also the rationality assumption may not be valid and people may not be
wealth maximizers. Finally, the assumption of a given investment
policy is not consistent with the advocates of the residual theory of
dividends15 who claim that the market value of the firm is maximized
when the capital budgeting expenditure, optimal capital structure and
dividend policy are simultaneously determined.
In essence, M&M argue that either the assumptions that they made
are valid or that the market behaves in such a manner that the effects
are not altered. Thus, with the exception of the tax criticism, the
irrelevance position is valid and investors are indifferent between
1Miller and Modigliani, p. 431.
1Advocates of the residual theory of dividends include Weston and
Brigham as well as Lintner.
dividend income and capital gains. There is no optimal dividend policy
for the firms just as there is no optimal financial policy. When taxes
are considered, investors should exert a preference for capital gains.
The Gordon Model
Gordon16 presents a slightly different model from M&M. He makes
two critical assumptions and draws upon valuation models which discount
cash flows. The basic assumptions are that individuals are risk averse
and that risk increases over time. Working under conditions of
uncertainty, Gordon equates the price of a share of stock, Po, to the
present value of its future dividend spinoffs.
Yo Yo Yo Yo
(2-21) Po = + + (+k)2 + (+k)3 + + ( k)n
1 + k (1 + k) (1 + k) (1 + k)
In effect, Gordon assumes that if no additional investments are made
beyond asset replacements that the firm will earn the same return, Yo,
in all future periods. If a corporation elects to retain a portion of
its dividend equal to Yo at time zero, then that firm will pay no divi-
dend at the end of the zero period. Instead, the investment of Yo will
yield a return in all future periods. In the Gordon model, k is not
only the discount rate but it is also the return that investors require
on a share of stock. If the reinvested earnings are expected to yield
a return of k = Yo/Po, the resulting valuation equation takes the fol-
SYo +kYo + Yo +kYo + Yo +kYo
1 + k (1 + k) ( + k) (1 + k)n
Simplifying one gets:
16Gordon, "Optimal Investment and Financing Policy," pp. 264-272.
kYo 1 1 1 Yo
(2-23) Po = 2 [1+ 1 k l+ 2 + ++ + nI + +k)2 +
(1+k) 1+k (l+k) (1+k)
(1+k)3 + + (+k)n
kYo 1 1 1 Yo
(2-24) kYo2 [1+ 2 + *** + 1 ] Yo
(1+k)2 [1+ l+k + (1+k) (l+k)n l+k
Yo Yo Yo Yo
(1 + k) + ( + k)2 + (1 + k)3 + (1 + k)
Therefore, equation (2-21) is equal to equation (2-22) and the value of
the firm is unaffected by the retention of earnings. The present value
of the dividend forgone is exactly offset by the present value of in-
creased future dividends resulting from the additional investment in
period zero. Gordon now takes issue with the policy of using a constant
discount rate in the valuation framework. Instead he offers the premise
that individuals may tend to discount the increased dividend resulting
from earnings retention at a higher rate than the dividend that would
have been paid otherwise, Yo. Therefore, he adjusts equation (2-22) in
such a manner that the incremental dividend resulting from earnings
retention is discounted at k' with k' being a greater discount rate than
k. The following equation results:
(2-26) Po Yo 2 + Yo 3 + Yo n+ kYo 2+ kYo +
(1+k) (1+k) (1+k)" (1+k') (1+k')3
(2-27) Po kYo [1+ i 1 2 + 1 n + Yo2+
(2-27) P ')2 l+k' (1+k') (l+k') (1+k)
(l+k)3 + ... + (+k)n
kYo 1 1 1 kYo
(2-28) 211 + + 1 2 + + n] k *
S (1+k') l+k' (+k') (1+k') k'(l+k')
(2-29) kYo Yo
k'(l + k') 1 + k
Then Po' is less than Po. That is, the present value of the future cash
flows for a firm which retained earnings and reinvested those earnings
at such a rate as to earn the required rate of return, k, would be less
than that for an otherwise identical firm that paid out a dividend Yo in
the initial period. Gordon admits, however, that he can not verify that
investors require a higher return on reinvested earnings. Instead, he
offers the possibility that investors' discount rates increase with time
with k < k2 < k < ... < k Gordon's rationale is that the greater
1 2 3 n
the lapse of time between investment and receipt the greater the risk
involved with a venture; therefore, the higher the discount rate should
be. Equation (2-21) thus modified becomes:
(2-30) Po = Yo + Yo 2 + Yo 3 Yo
1+k1 (l+k2) (1+k-)3 (l+kn
Gordon concludes that the k in equation (2-21) could be a weighted aver-
age of kt 's in equation (2-30) with the weights being determined by the
cash flows in the respective periods. Now if the dividend is retained
in the initial period the following represents the present value of the
(2 ) P' Yo+kYo 2 Yo+kYo++ Yo + kYo
(2-31) Po' = 2 ++ k3 + + n *
(1 + k2) (1 + k3) (1 + k)
However, if k' is the weighted average of all the kt 's then equation
(2-31) is equal to equation (2-26). Likewise it can be shown that k' is
greater than k. Therefore, Po in equation (2-30) is greater than Po' in
equation (2-31). The results indicate that investors value the paying
out of earnings in dividends more than the retention of earnings.
A few assumptions of the Gordon model have been questioned.
Perhaps the most important problem area is that he assumed no taxes.
In particular, the tax differential between the rate that individuals
may have to pay on capital gains and dividend income may negate his
conclusion in a world of taxes. Also his resolution of uncertainty by
increasing the discount rate when dividends are retained and reinvested,
while generally accepted, may still be questioned.
The unfortunate fact is that the Gordon and M&M models are incompa-
rable on purely theoretical grounds. That is, Gordon and the irrelevance
position are not talking about the same case. M&M assumed a given
investment policy; Gordon did not. Also Gordon did not provide for the
possibility of an investor selling a portion of his stock. The inabili-
ty to compare the two models because of conflicting assumptions encour-
aged studies which sought to uncover the pragmatic importance of
In summary, studies supporting Gordon have verified the importance
of dividends in regression models. Contrary studies in support of the
irrelevance position have shown evidence of the existence of a clientele
effect or have presented a case for retention of earnings due to the
existence of a tax advantage of incomes from capital gains over dividend
income. Neither side has conceded the argument.
The empirical studies attempting to test the various positions on
valuation with respect to earnings and dividends can be classified into
two categories. First, there are the multiple regression studies from
the late 1950s through the 1960s. The bulk of evidence from these
studies is at best circumstantial due to inherent statistical problems.
The second group of studies is a very diverse collection of data which
attempts to determine the incidence of tax policy, the existence of
clienteles, the propensities of firms to pay out dividends, and the pro-
jection of a dividend payout ratio for a firm from historical and
financial data. The evidence presented by this group is probably the
stronger; yet, the total information provided by all empirical studies
may well be inconclusive.
Early Regression Models
In surveying the first group of empirical studies, three multiple
17 18 19
regression models by Durand, Gordon,8 and Brigham and Gordon will
be presented. The statistical problems of multiple regression analysis
as applied to the valuation controversy will be discussed.
The Durand model
Durand's model was concerned with explanation of share valuation
in terms of financial variables. The particular variables that he chose
were the book value per share, B, earnings per share, E, and the
dividend payout ratio, D/E. The particular function that he chose was
the log linear function that follows:
(2-32) P = log(a) + b1log(B) + b2log(E) + b3log (D/E).
1Durand, pp. 52-59.
18Gordon, "Dividends, Earnings, and Stock Prices," pp. 99-105.
19Eugene F. Brigham and Myron J. Gordon, "Leverage, Dividend Policy,
and the Cost of Capital," Journal of Finance, XXIII (March, 1968), pp.
That is, share price, P, is a log linear function of a constant, book
value per share, earnings per share and the dividend payout ratio.
Durand then tested the statistical significance of adding the dividend
payout ratio variable to an equation in which that one variable had
been omitted. He found that the addition of the dividend payout ratio
to an equation in which price was a log-linear function of a constant,
book value per share, and earnings per share was significant at the .01
level when an F-test was performed on the entering variable. He
concluded that prices are influenced by dividends.
The Gordon model
Gordon's empirical study in the late 1950s was based upon a cross-
section of the Chemical, Foods, Steel, and Machine Tools industries for
the years 1951 and 1954. In essence, the study was a multiple regres-
sion model of three hypotheses and two equations. Gordon contended that
there were three plausible hypotheses for the valuation of the common
stock of a firm. The first he associated with the statement by Graham
and Dodd that stock prices were, in fact, a function of both dividends
and earnings. The second was the position which he supported and which
had evolved from the earlier work of John Burr Williams, that price was
solely determined by dividends. The third represented the position of
the "pure earnings" theorists. To test the Graham and Dodd contention,
Gordon considered the following equation appropriate:
(2-33) P = ao + aD + a2Y
where P is the year-end share price, D is the annual dividend, and Y is
annual income. To test the second hypothesis Gordon advanced the equa-
tion that price was a function of dividends and retained earnings (Y-D)
which he considered a surrogate for growth. That equation was:
(2-34) P = ao + a D + a2(Y-D).
It should be noted that the two equations will give identical coeffi-
cients of determination. This is more apparent when equation (2-33)
is rewritten as follows:
(2-35) P = ao + alD + a2(Y-D) + a2D.
That is, the al in equation (2-34) will be equal to the sum of al + a2
in equation (2-33). In either case the regression coefficients of ao
and a2 will be the same for equations (2-33) and (2-34). Gordon felt
that some of the signs were wrong for equation (2-33) and was disturbed
by the fact that the change between the two years in regression coeffi-
cients for the same industry was sometimes drastic. Consequently, he
believed that equation (2-34) probably was more appropriate. In addi-
tion, he reasoned that equation (2-34) not only supported his position
but also failed to support that of the "pure earnings" theorists. The
evidence that Gordon cited here was that the coefficient for equation
(2-34) was greater for D than for (Y-D). His logic was that dividends
play a greater role in the price mechanism than growth (i.e., earnings).
The Brigham and Gordon model
The Miller and Modigliani0 study of the electrical utilities
industry was published in 1966. The empirical study concluded that
dividends had a negligible effect upon share prices for the electrical
utility sector in the 1954-1957 period. Perhaps in response to the
M&M conclusions, Brigham and Gordon's cross-section multiple regression
20Merton H. Miller and Franco Modigliani, "Some Estimates of the
Cost of Capital to the Electric Utility Industry," American Economic
Review, LVI (June, 1966), pp. 333-391.
study of the electrical utilities industry was published in 1968. The
findings of the latter were inconsistent with the "pure earnings"
theorists. The principal equation fitted by their study was:
(2-36) Po = a0 + alg.
That is, the current dividend per share divided by current market price
per share is a function of a constant and the historical growth rate of
retained earnings. Brigham and Gordon subsequently added additional
explanatory variables in order to reduce specification bias and to test
for other effects such as leverage. To interpret the findings of the
regression analysis one only needs to recall that the present value of
a stock with a constant growth rate in the Williams' evaluation model21
(2-37) Po = Do
Equation (2-37) simply states that share price is equal to the dividend
per share divided by the difference between the discount rate applied
to equity stream, ke and the growth rate. Solving for Do/Po or the
dividend yield the equation becomes:
(2-38) pD = ke g.
Brigham and Gordon use br the product of the retention rate (b) and the
rate of return on equity investment (r) as a surrogate for growth so:
(2-39) g = br.
Consequently, they note that if al is equal to -1 then k is unaffected
by dividend policy and, in effect, the investor is indifferent between
dividends and capital gains or growth. Otherwise, the investor has a
21Williams, p. 88.
preference for either dividends or growth. If al is greater than -1
then investors have a preference for dividends over capital gains. If
al is less than -1 the reverse is true. In most of their linear
models with from two to five additional explanatory variables added
for the reasons described previously, the coefficient for al ranged
from about -.5 to -.2. Brigham and Gordon interpret this as evidence
of the importance of dividends as a policy decision.
Critique of Regression Models
In a 1964 article Irwin Friend and Marshall Puckett22 dealt ex-
tensively with statistical problems in multiple regression valuation
equations, in general, and the findings of Gordon and Durand, in
particular. In essence, five distinguishable problems were presented
in their study.
Seemingly the most serious is that of omitted variables or speci-
fication bias. In regard to conclusions such as those reached by
Gordon from equations similar to (2-34), Friend and Puckett took issue.
Two possible omissions troubled the two researchers. The more critical
was perhaps the omission of an appropriate risk variable. For example,
if the price of stock is an inverse function of risk and high risk
firms, out of concern for maintaining their dividends over time, tend
to pay smaller dividends than comparable low risk firms, then the coef-
ficient of the dividend variable will be biased upward in equation
(2-34). What appears to show that dividends are more prominent in
valuation of shares than retention may be simply the fact that more
risky shares do not command the same market price as low risk shares.
221rwin Friend and Marshall Puckett, "Dividends and Stock Prices,"
American Economic Review, LIV (September, 1964), pp. 656-682.
Also Friend and Puckett were reluctant to concede that growth is deter-
mined exclusively by retention of earnings. Additional specification
bias may result from the omission of a growth variable although the
direction of the bias was viewed as indeterminant by the study.
Regression weights may constitute another trouble area for mul-
tiple regression studies. Each observation's contribution to the final
equation is relative to the deviation from the average of that observa-
tion for each variable. As a result, extreme values have a greater
relative contribution than values which cluster around the mean. If it
is evident that high and low quality stocks exhibit high and low
dividend payouts respectively, then the regression coefficient of divi-
dends in equation (2-34) may exhibit an upward bias.
A third potential problem may exist if investors consider part
of the fluctuation in earnings as random. Since dividends fluctuate
considerably less than earnings then multiple regression analysis would
tend to attribute greater prominence to the dividend variable and less
to the earning variable.
An additional area of concern may be the fact that dividend
figures can be precisely measured and earnings figures can not. A down-
ward bias in retention of earnings may also be present in equation
(2-34) in relation to the dividend variable. As a result, the earnings
figure reported is not the appropriate figure that investors consider
due to accounting principles that necessitate adjustment in earnings
figures before meaningful earning figures are obtainable.
Finally, what Friend and Puckett identify as a least-squares bias
may exist. That is, causation is assumed to run from dividends to the
price. However, if investor and management expectations are not
comparable then the price of shares may influence management's decision
to pay dividends. This bias would also be upward for the dividend
It would appear that the Brigham-Gordon regression model sought
to minimize the shortcomings pointed out by the Friend and Puckett
study. The mere fact that the electrical utilities industry was se-
lected probably minimized the specification bias of omitting a risk
variable and failure to take account of growth not influenced by re-
tention. For one thing, risk and growth rates within that particular
industry do not vary greatly from firm to firm as one would have ex-
pected in the Gordon study. Even earnings exhibit a greater stability.
The Brigham-Gordon study did create some new statistical problems when
additional variables such as debt to net worth were added to equation
(2-36). While, in fact, this study probably has fewer serious
statistical problems than the studies by Gordon and Durand, it may well
have more minor problems because of the intercorrelation among inde-
pendent variables which were added to reduce specification bias. At
any rate, statistical shortcomings make the evidence seem less
conclusive than the figures would otherwise seem to indicate.
Other Empirical Studies
Edwin J. Elton and Martin J. Gruber23 sought to provide empirical
substantiation of the "clientele effect" and "rationality" by studying
marginal stockholder tax rates. To test the hypothesis that the
marginal tax rate of marginal investors decreased as the dividend pay-
23Edwin J. Elton and Martin J. Gruber, "Marginal Stockholder Tax
Rates and the Clientele Effect," Review of Economics and Statistics,
LII (February, 1970), pp. 68-74.
out ratio increased, the study focused on equilibrium price movements
in share prices between prices on stock sold with the dividend and
that same share sold ex-dividend one day later. If Pa is the price of
a share sold ex-dividend, Pb is the price of a share sold with divi-
dends on the day before that share goes ex-dividend, Pc is the price
at which the stock was purchased, to is the tax rate on dividend income,
tc is the tax rate on capital gains, and D is the amount of the divi-
dend, then the following equation dictates an equilibrium situation:
Pb Pa 1 to
D 1 tc
In equilibrium the investor at the margin will be indifferent between
the after-tax wealth he receives from selling the stock or not selling.
(2-41) Pb tc(Pb Pc) = Pa tc(Pa Pc) + D(l to).
Elton and Gruber, consequently, used (Pb Pa)/D as a surrogate for
the marginal tax bracket of marginal investors. To test the hypothesis,
the study utilized closing prices on stocks on the New York Stock
Exchange traded on the ex-dividend and the previous day. Using a
Spearman rank correlation coefficient to discern if there was a rela-
tionship between the surrogate for the marginal tax rate and both the
dividend yield and the dividend payout ratio, the results were reported
to be significant at the .01 level and positive. Elton and Gruber
concluded that the statistical evidence supported the clientele effect
of Miller and Modigliani and provided some evidence of market rational-
ity in that shareholders in higher income tax brackets preferred
capital gains to dividend income. The researchers admitted a few
statistical problems. They seemed particularly concerned about the
bias that could result from the use of dividends as the denominator of
the dependent variable and the numerator of the independent variable.
Perhaps there is cause for concern. Yet it is indeed questionable
whether the findings should be totally invalidated by the presence of
spurious correlation between the independent and dependent variable.
Additional empirical studies of marginal tax rates of investors were
made by Weston and Brigham,24 Holland,25 and Jolivet.26 The Elton and
Gruber estimate of the marginal tax rate was about 36 percent; whereas,
the respective estimates of the others were 46, 55.6, and 39 percent.
While there is some disagreement due to differences of measurement
techniques and to the fact that the studies were conducted over a span
of some twenty years,27 the argument remains that, after the basic
dividend exemption, the marginal tax rate on dividends is greater than
the rate applied to capital gains.
Another strong case for retention of earnings is evident if the
following statement by West and Bierman predominates:
J. Fred Weston and Eugene F. Brigham, Managerial Finance (2nd
ed.; New York: Holt, Rinehart and Winston, 1966), p. 309.
25Daniel M. Holland, Dividends Under the Income Tax (New York:
National Bureau of Economic Research, 1962), p. 120.
26Vincent Jolivet, "The Weighted Average Marginal Tax Rate on
Dividends Received by Individuals in the U.S.," American Economic
Review, LVI (June, 1966), pp. 473-477.
27Jolivet attempted to determine the marginal tax rate from his-
torical tax returns for 1961 and 1962. Weston and Brigham weighted
the marginal tax rates of dividend recipients. The weights were
calculated according to the total dollar dividends received by share-
holders in each tax bracket. The Weston and Brigham study was for
1965. Holland's figure was a weighted average marginal tax rate on
dividends received by individuals for 1952. The Elton and Gruber
study was based on April 1, 1966 to March 31, 1967 data in the manner
Dividend policy merely influences the magnitude and
timing of the shareholders' tax payments. In particular,
decreases in present dividends tend to reduce and put off
the payment of taxes .... Under uncertainty, this case
is, if anything, strengthened: a decrease in dividends
results in a known reduction in current tax payments which
is only partially offset by a smaller increase in the
present value of a less certain future tax liability.28
Baumol also made an interesting refutation of a statement to a
declaration by Miller and Modigliani that in the long run one clientele
would prove equally as beneficial to the valuation process as another
for the firm. Baumol concluded:
As is the case with any potential supplier of a differentiated
product, good product design strategy consists in looking for
segments of the market whose wants are relatively poorly supplied,
and in turning out a product calculated to satisfy these desires,
for such a product will command a price which is comparatively
high in relation to those which serve the more saturated segments
of the market. Thus, choice of dividend policy may well be a
matter of considerable moment for the corporation, one that merits
most careful consideration and planning, and for which optimality
calculations are of the utmost relevance.29
No empirical evidence is cited in support of the above statement. Theo-
retically, the existence of overabundance in a segmented market would
seem to violate the assumptions of rationality and perfect markets by
Miller and Modigliani. Pragmatically, the existence of such conditions
may be cause for concern to the firm if there is any way of discerning
Another argument directed against the "pure earnings" theorists
is offered by Lintner.30 The focus of his argument is that firms in
the M&M model who are forced to enter the market in order to pay divi-
28Richard R. West and Harold Bierman, Jr., "Corporate Dividend
Policy and Pre-emptive Security Issues," Journal of Business, XLII
(January, 1968), p. 75.
William J. Baumol, "On Dividend Policy and Market Imperfection,"
Journal of Business, XXXVI (January, 1963), p. 115.
30John Lintner, pp. 256-259.
dends and still meet a given investment level may find that they have
downward sloping demand curves for their common stock. When underpric-
ing of this nature exists it will be necessary for firms to sell a
greater number of shares to replace the distribution of earnings to the
shareholder. Consequently, present stockholders would be better off if
earnings were retained and the market price of the shares were maintain-
ed and not depressed by stock issues. Theoretically, this is a viola-
tion of the assumption of perfect capital markets. On this matter Friend
and Puckett surmise:
We do not normally witness perceptible drops in the market
price level when the aggregate supply of corporate stock
is increased by new issues, requiring for their absorption
the substitution of current for future income and potenti-
ally raising the risk premium demanded by investors; nor do
we typically witness sharp drops in the per-share price
when the supply of an individual company's shares is
increased. It is possible to infer of course that these
increases in supply are precisely timed so as to be auto-
matically offset by upward shifts in investor expectations,
but this seems completely unrealistic.31
Other empirical studies took the alternate approach of looking at
what firms are doing and inferring that the behavior of those firms
should approach optimality. One such study was advanced by Lintner32
in the late 1950s. The basic equation utilized multiple regression
and was of the following form:
(2-42) Dit = ait + b Pit + b2 Di(t-l) + uit
The dividends of firm (i) in period (t) were then regressed against a
constant, in period t, and dividends in the previous period. An error
term u. for the ith firm in period t was also included. Lintner's
Friend and Puckett, p. 659.
32John Lintner, "Distribution of Incomes of Corporations Among
Dividends, Retained Earnings, and Taxes," American Economic Review,
XLVI (May, 1956), pp. 97-113.
contention was that the firm behaves very conservatively and is reluc-
tant to raise its dividend payment until it is reasonably sure that
the increase dividend level can be maintained. More precisely, there
is a substantial degree of inertia or resistance to departure from
what is apparently considered an optimal dividend payment. Reformula-
tion of equation (2-42) for a particular firm reveals:
(2-43) Dt Dt1 = at + c(rPt Dtl) + ut .
Brittain offers an excellent interpretation of this equation. He con-
Corporations were pictured as pursuing a "target" payout
ratio r which they apply to current earnings [P 1. If
dividends were fully adjusted each year to achieve the
target level [rP ], they would be changed by [rPt Dt-l]
between times t 1 and t. However, a conservative bias
against large revision leads them to move only a fraction
of the way, given by the speed of adjustment coefficient
c. The constant term a -- expected to be positive -- was
added to allow for a presumed greater reluctance to cut
dividends than to raise them. If a is positive, the model
says that dividends will tend to drift up even if the rest
of the model calls for no change. Thus, the target level
of D must be enough below D to offset the constant term
in order for a cut in dividends to be indicated.33
Lintner tested equation (2-43) by fitting it to 1946-1954 data of Amer-
ican Corporations. Brittain34 showed that the predictive power of the
Lintner model could be increased with the substitution of cash flow
Keith Smith,35 in an effort to test for the empirical existence
of a in equation (2-43), studied the "increasing stream" hypothesis.
3John A. Brittain, Corporate Dividend Policy (Washington, D.C.:
The Brookings Institution, 1966), p. 19.
3bid., p. 52-73.
35Keith V. Smith, "Increasing Stream Hypothesis of Corporate Divi-
dend Policy," California Management Review, XIV (Fall, 1971), pp. 56-
Smith's statement of that hypothesis is perhaps best illustrated by his
The increasing stream hypothesis of corporate divi-
dend policy is that the board of directors deliberately
avoids dividend cuts if at all possible and attempts to
construct over time an increasing, or at least a nondecreas-
ing, record of cash dividend payments.36
Working with a large sample of firms from the Standard and Poor's
Compustat Service, Smith found that a large number of firms did not cut
dividends during the 1949 to 1966 period. He also classified firms ac-
cording to the number of consecutive years that dividends had been
increased and noted that the majority of firms had increased dividends
over a number of years. His evidence would seem to present some strong
empirical support of the reluctance to decrease the dividend level. One
criticism is that it is a very short-run approach and fails to signal
or explain drastic alterations in dividend payments which do occur from
time to time.
A more recent and probably more pragmatic study is one by Edwin P.
Harkins and Francis J. Walsh, Jr.37 The study was a survey of finan-
cial executives of 166 firms. Of the 155 responses to a question as to
which is more important cash flows or net income in determination of
dividends, 70 felt the two were equally important, 57 felt that net in-
come was more important, and 28 felt that cash flow was more important.38
The figures would not seem to substantiate Brittain's refinement of
Lintner's regression equation (2-43). Also of interest are frequency
3Ibid., p. 57.
37Edwin P. Harkins and Francis J. Walsh, Jr., Dividend Policies and
Practices (New York: The Conference Board, Inc., 1971), pp. 1-27.
8Ibid., p. 10.
of response tabulations to a question designed to reveal the considera-
tions that influence boards of directors in making dividend decisions.
A partial list of the most important items in that study is presented
in Table I:
FREQUENCY OF RESPONSES OF EXECUTIVES
TO FACTORS INFLUENCING DIVIDEND POLICY*
Earnings record and future prospect 94
Continuity or regularity of dividend payment 76
Stability of rate per share 47
Cash flow, position and future needs 44
*Edwin P. Harkins and Francis J. Walsh, Jr., Dividend Policies and
Practices (New York: The Conference Board, Inc., 1971), p. 4.
This information would seem to contain support for conflicting
viewpoints on the question of dividend policy. The second and third
responses would seem to support Lintner's contention of an "inertia",
a positive value for the constant, a, in equation (2-43), and the speed
of adjustment coefficient. The first response would seem to substan-
tiate the "pure earnings" theorists. Yet the addition of "future
prospect" makes this more uncertain. In fact the "pure earnings" the-
orists themselves are divided over whether historical earnings or ex-
pected earnings is the appropriate value to discount in valuation
models. Miller and Modigliani use expected earnings. It is possible
that the frequency response might have been quite different if the two
concepts had been differentiated. Of course, one could argue that
future prospects are quite likely to be a projection of historical
earnings. This may be true. However, there is a possibility that
"future prospect" includes a subjective consideration of the capital
budgeting outlook. If so, then perhaps proponents of the residual
theory have some indirect support. The fourth response directly sup-
ports Brittain's substitution of cash flow for net income. On the
surface it would appear that the first response contradicts this same
suggestion. Indirectly, the fourth response may support the residual
dividend policy if future need includes capital budgeting considera-
tions. And to the extent that cash flow and earnings do not move in
harmony, the position of the "pure earnings" theorists is weakened.
Empirical studies and a few arguments with little or no support-
ing data have been advanced to settle the argument between the "pure
earnings" theorists and followers of John Burr Williams who insist
that dividends are the appropriate stream to discount. In particular,
empirical evidence of varying degrees of merit and reliability has
been presented in support of the positions of Miller and Modigliani as
well as Gordon. The preponderance of the empirical evidence supports
the conclusion of neither, in that Miller and Modigliani advocated no
dividends for income tax reasons and Gordon advocated no retention due
to the higher discount rate associated with uncertainty that is applied
when dividends are withheld. Perhaps the most overwhelming evidence is
the simple observation that very few firms pay either all or no divi-
dends. The study of Harkins and Walsh found the majority of firms'
payout ratio ranges from 30 percent to 59 percent.3 Either this
means that the greater majority of the firms are suboptimizing and
could increase their market value by either paying out more or less
dividends, or else the models of M&M and Gordon are inapplicable to
the real world.40 Perhaps the assumptions imbedded in cetaris
paribus are unrealistic. Theoretically, the Miller and Modigliani
as well as the Gordon position are irrefutable granted the different
assumptions of the two positions. The question then is why do the
two conflicting models fail to depict the real world situation?
This paper does not attempt to resolve the conflict between
M&M and Gordon. Instead, an effort is made to analyze an additional
source of empirical evidence that may aid in ascertaining the role
that dividends play in determining the market value of the firm.
39Ibid., p. 4
40Intermediate payout ratios would not appear to be inconsistent
with the residual theory and the related simultaneous approach of
Lerner and Carleton. The contention of the simultaneous approach is
that the declining rate of return on investments realized by firms
which move along their downward-sloping marginal efficiency of
capital schedule may be responsible for different optimal payout ra-
tios. Once a firm's marginal return on investment falls below its
marginal cost of capital, the firm no longer finds it profitable to
invest earnings and should pay dividends to shareholders.
See Eugene M. Lerner and Williard T. Carleton, A Theory of
Financial Analysis (New York: Harcourt Brace Jovanovich, Inc., 1966).
METHOD OF RESEARCH
Several problems became apparent in the process of screening
automatic dividend reinvestment plans as a possible area for empirical
research. Very little work of an academic nature was available for
background research as the development is one of recent history, since
1967. Also, the process itself seems to be picking up momentum and
there is a possibility that the study might be outdated before it
could be completed. When dealing with such a dynamic process, timing
should be a critical issue and any survey must deal with a specific
and relatively narrow time horizon or information from different
sources may not be comparable due to changes that might take place.
Subsequently, the decision was made to conduct a preliminary survey in
the spring of 1972. Even then, it was obvious that the result of that
survey would be mainly informational and that it would be necessary to
follow up with a similar survey at a later date.
The Collection of Information
The fifty largest banks as defined by Fortune were contacted in
the spring of 1972 on a very informal basis and a list of bank-managed
plans in existence was collected from the 39 banks that responded to
the initial survey. Plan descriptions were also generously supplied
by several of these banks, along with bits and pieces of miscellaneous
information. Several pragmatic articles, mostly in the business media
were helpful in providing background information. Mr. Bill Paul of the
Wall Street Journal, Mr. Francis J. Walsh of the Conference Board, and
the American Society of Corporate Secretaries were kind enough to
provide needed information. Also a copy of a paper presented at the
ABA Trust Operations and Automation Workshop in Atlanta, Georgia, by Joe
L. Bookout1 of North Carolina National Bank was extremely helpful. Since
pertinent information needed to be collected simultaneously from several
sources, it appeared that three populations should be surveyed to gather
the type of information that the study required. Unfortunately, it soon
became apparent that a survey of shareholders would be infeasible due to
cost considerations and to the attitude which one financial executive so
eloquently expressed with the following statement: "I would not give a
list of our stockholders to my mother." Thus, two populations
constituted the data sources for the study. The first of these, Fortune's
Fifty Largest Banks had already been contacted. The second population
was nonfinancial firms. Unable to deal with all nonfinancial firms, the
decision was made to survey only the Standard and Poor 500. Financial
data on these firms were obtained from the Compustat Tapes.
Several problems were also inherent in the population chosen to
represent both the banks and the firms. Actually, not all bank-managed
plans are handled by the fifty banks chosen, but the vast majority of
them are. Consequently, some plans may be administered by banks not in
the Fortune Fifty and could go unidentified. Yet errors in classifying
firms as users or nonusers are much more likely to result from failure
of the banks to respond to the survey than from defining the population
of banks handling ADR plans as the Fortune Fifty.
Joe L. Bookout, Dividend Reinvestment, unpublished paper (North
Carolina National Bank, March, 1973).
The necessity of defining the population of firms with ADR plans
as the Standard and Poor 500 is a more serious problem. While the
survey revealed that 376 financial and nonfinancial firms had estab-
lished ADR plans, this study was limited to nonfinancial firms. A sub-
sequent decision to restrict the population to members of the S&P 500
was made for several reasons. First of all, the computational burden
of dealing with all financial and nonfinancial firms would be simply
overwhelming. Also, the availability of data would surely be a problem
and excluding the firms for which sufficient datawerenot available
could well bias the results. It was felt that sufficient data would
be available from the Compustat Tapes for most of the 500 and very
little bias would result in that fewer firms would be excluded because
of insufficient data. It is admitted that problems do exist in the
use of the S&P 500 as a sample for all nonfinancial firms; however,
the alternative of biasing the results by omitting a considerable num-
ber of firms for which data might not be available appeared less
appealing. At least, it is hard to dispute that the sample chosen
contains most of the productive power in our economy and occupies a
pivotal position in the productive process.
A final survey of the fifty banks was conducted in September and
October of 1973. At the same time a questionnaire was mailed to the
S&P 500. Most of the responses were received within a one-month
period. The respective response rates were 86 and 67 percent. It was
fortunate not only that most of the banks responded but also that
practically all of the twenty largest did. The corporate response rate
was also considered high especially in light of the fact that some
firms have a written policy against answering questionnaires.
The banks provided information on number, cost, and identity of
the various plans that they administered. They also provided consider-
able literature that was being mailed to shareholders to solicit partic-
ipation if they handled any such plans. Banks not administering ADR
plans sometimes cited reasons for not offering their corporate deposi-
tors this service and often gave some indication of how they felt about
the future possibility of administering these plans.
All correspondence between the banks was again on an informal and
individual basis. To a large degree, requests were comparable among
banks but to some degree differences existed. The differences in in-
formation required were to a large degree determined by several factors
such as whether or not the bank responded to a similar request in the
spring, whether or not the bank had identified itself as active in
administering ADR plans, etc.
For the survey of S&P 500 firms, a different technique was
employed. Some of the firms which had been prominent or innovative in
the ADR movement such as American Telegraph and Telephone, Long Island
Lighting Company,2 General Motors, and Allegheny Power Company were
contacted long before the questionnaire was mailed. Since the question-
naire itself needed to be short and easy to answer, two questionnaires
were employed (see Exhibits I and II). The first questionnaire was
designed to be answered only by firms with established ADR plans. The
second questionnaire was, consequently, designed to be answered by
those firms without an established ADR plan. Since a few of the firms
had not been identified as users or nonusers, both questionnaires were
Long Island Lighting Company does not belong to the S&P 500.
Although data on Long Island Lighting Company were not used in the sub-
sequent analysis, information that was obtained did prove useful.
AUTOMATIC DIVIDEND REINVESTMENT PLANS
(Only for firms with established ADR plans)
1. When was your plan established? Month Year
2. Is the administrative function of your plan handled by a commer-
3. Approximately what percentage of your stockholders use the plan?
Please make a "best guess" estimate.
a. Percentage of number
b. Percentage of shares
4. What is the approximate annual dollar volume of funds being rein-
vested in your stock via the ADR plan?
5. What were your primary reasons for establishing the plan? If more
than one reason is applicable, please indicate the most important
by 1, the second most important by 2, etc.
a. To provide a service for those shareholders who want to
save and not use current dividend income for consumption.
b. To create a broader market for our common stock, thus
holding up the price of our stock to some extent.
c. To encourage small shareholders to own a larger portion
of the company and increase ownership dispersion.
d. To raise new capital for the company.
e. Other, please specify.
6. Does your present plan allow you to issue new shares of
common stock to participating shareholders?
If not, what are your present inclinations about such a feature?
a. Still relatively uncertain.
b. Have definitely rejected issuing new shares of common
at least for some time.
c. Expect to change our plan to include this feature
sometime in the not so distant future.
d. Have really never seriously considered the possibility
7. If a "significant" number of your stockholders elected to use the
plan, might this influence your firm's dividend payout policy?
AUTOMATIC DIVIDEND REINVESTMENT PLANS
(Only for firms without established ADR plans)
1. Have you considered the idea of establishing such a plan?
a. The idea has simply never come up.
b. The idea has been considered in some detail and it was
decided that a plan would not be feasible for our
company at this time.
c. The idea is presently being explored
1. and no decision has been made to date.
2. and it appears quite likely that a plan may
be established in the not too distant future.
3. and it appears quite likely that the company
will not establish a plan at this time.
2. If you have considered or are exploring an ADR plan at
this time, does the plan involve new equity issues?
3. If you considered establishing a plan but rejected the idea,
what were your reasons? If more than one reason is applicable,
please indicate the most important by 1, the second most
important by 2, etc.
a. The dollar volume of our dividend payments are so low
as to make such a plan impractical.
b. The cost to the company would be too high.
c. The cost to participating stockholders would be too
d. We are convinced that our stockholders want current
income, hence the actual number of stockholders that
would want to participate would not be great enough
to justify setting up an ADR plan.
e. Other, please specify.
4. At the present time, do you think that your company will
set up a plan within the next three years if:
a. The plan allows only for reinvestment in out-
standing issues of common stock.
b. The plan allows only for the issuance of new
shares of common stock to participating share-
sent to each firm along with a cover letter explaining the nature of
the study and assuring respondents that information on the question-
naires would be kept confidential.
The questionnaire was addressed to the financial officer of the
firm who appeared most likely to be directly involved in the opera-
tions should the firm have an ADR plan. Names and addresses were
obtained from the Standard and Poor Corporation Directory. The first
questionnaire was designed to ascertain the nature of the ADR plan of
that particular firm, the rationale for establishing the plan, whether
or not the success of the plan might have any effect upon dividend
policy, and the degree of participation that the firm has experienced
in that particular plan. Likewise, the second questionnaire attempted
to determine whether or not ADR plans had been discussed. If so,
several likely reasons for rejection were listed for the participant to
designate. Finally, the questionnaire sought to determine the firmness
of the rejection if there had been one. It was felt that this line of
questioning would give information that might be useful in forecasting
future growth as well as reflect the feelings of those managers who
had rejected the ADR option.
The Analysis of the Data
Other than the tabulation of the frequency of various responses
to the questionnaire and the calculation of simple averages, two other
statistical techniques were employed. Multiple Discriminant Analysis
and Multiple Regression Analysis were selected to develop profiles and
3Standard and Poor Register of Corporations, Directors and Execu-
tives (New York: Standard and Poor's Corporation, 1974).
to notice relationships of firms that were utilizing automatic dividend
The Multiple Discriminant Model
The statistical tool employed to test for hypothesized differences
in the profiles of ADR users and nonusers was multiple discriminant
analysis. MDA is cited by Sheth4 as being the appropriate statistical
technique to employ when the independent variable consists of discerni-
ble classifications. It can be used as a predictive tool to classify
observations into groups or as a means of developing profiles that
distinguish characteristics between groups.
Multiple discriminant analysis
Discriminant analysis seeks to reduce a multivariate data group
to a single variable with its linear transformation function selected
in a manner which maximizes the ratio of variances between groups to
inter-group variance. Several advantages are recognized. The problem
is reduced in complexity fron an I dimensional space to an I 1 dimen-
sional space. Also simultaneous consideration is given to all J vari-
ables rather than taking each into account on a univariate basis.
The underlying assumptions that a researcher makes in applying
linear discriminant analysis are: (1) that all J variables are from a
multivariate normal distribution in each of I exhaustive populations,
(2) that the I populations are discrete and readily identifiable, (3)
that each of the I populations can be characterized by J variables, and
(4) that a linear relationship is the appropriate one to employ.
Jagdish N. Sheth, "The Multivariate Revolution in Marketing Re-
search," Journal of Marketing, XXXV (January, 1971), pp. 13-19.
Applications of MDA to problems in economics and business are
numerous with Durand5 providing one of the earliest adaptations. Other
contributors to financial analysis include Williams and Goodman,6
7 8 9
Altman, Pinches and Mingo, Simkowitz and Monroe, Klemkosky and
10 11 12
Petty,1 Walter, and Smith.
This particular statistical technique was employed to determine
if ADR users display certain financial characteristics which differ
from nonusers. The goal then is the development of profiles for firms
that have implemented ADR plans. Since there are only two groups,users
5David Durand, Risk Elements in Consumer Installment Financing
(New York, National Bureau of Economic Research, 1941).
6W.H. Williams and M.L. Goodman, "A Statistical Grouping of Corpo-
rations by Their Financial Characteristics," Journal of Financial and
Quantitative Analysis, VI (September, 1971), pp. 1095-1104.
Edward I. Altman, "Financial Ratios, Discriminant Analysis and
the Prediction of Corporate Bankruptcy," Journal of Finance, XXIII
(September, 1968), pp. 589-609.
8George E. Pinches and Kent A. Mingo, "A Multivariate Analysis of
Industrial Bond Ratings," Journal of Finance, XXVIII (March, 1973),
9Michael Simkowitz and Robert J. Monroe, "A Discriminant Analysis
Function for Conglomerate Targets," Southern Journal of Business, VI
(November, 1971), pp. 1-16.
10Robert C. Klemkosky and J. William Petty, "A Multivariate
Analysis of Stock Price Variability," Journal of Business Research,
I, No. 1 (Summer, 1973), pp. 1-10.
J1ames E. Walter, "A Discriminant Function Earnings-Price Ratios
of Large Industrial Corporations," Review of Economics and Statistics,
XLI (February, 1959), pp. 42-52.
12Keith V. Smith, Classification of Investment Securities Using
Multiple Discriminant Analysis, Institute Paper No. 101 (Purdue
University Institute for Research in the Behavior, Economic and Manage-
ment Sciences, 1965).
and nonusers, the problem is reduced from a two-dimensional to the one-
dimensional form that follows:
Z = B.X.
Where Z is the transformation value of the discriminant function for an
observation, the Xi's are the independent variables, and the B 's are
the weights or coefficients assigned to the independent variables. The
linear transformation of the multivariate function, Z score, is thus a
function of the independent variables for that observation and the
weights. Classification of a firm as a user or nonuser is subsequently
related to the B 's and the cutting score, Z*, used to differentiate
between users and nonusers. Observations in which the Z score is
greater than Z* are classified as belonging to one group. If the Z
score is less than Z* the observation is classified as belonging to the
other group. Optimal weighting coefficients are determined in a manner
which maximizes the differences between means for the two groups rela-
tive to the variation within groups for that particular variable. Both
Z* and the B.'s were determined by a stepwise discriminant program
(BioMedical Computer Program 07M).
Selection of variables
Seven categories of variables were selected for the study:
X1 Three-year weighted average yield (change in market price
plus dividend income divided by initial market price) for
the years 1970-1972.13
13The respective weights used in the study for 1970-1972 were .2,
.3 and .5.
X2 Average price earnings ratio for 1972 (high + low share
price divided by EPS).
X3 Growth rate in EPS, the arithmetic mean, based on yearly
X Growth rate in EPS, the geometric mean, based on yearly
X5 Current ratio for 1972.
X6 Debt over net worth for 1972.
X7 Total dollar dividends paid in 1972.
X8 Total number of shareholders as of the end of 1972.
Xg Total assets as of the end of 1972.
X10- Average dividend payout ratio for 10 years, 1963-1972, after
throwing out the high and the low.
X11- Industry utilization factor based on S&P industry classifica-
tion and equal to the number of firms in an industry using
ADR's divided by the number of firms in that industry.
Variables were selected according to two criteria. First,
variables which were considered to add to the predictive power of the
model were included. Secondly, variables which might have otherwise
proven to be good predictive variables were omitted from the study if
their inclusion would have sufficiently reduced the sample size of the
multiple discriminant model. In fact, this consideration was respon-
sible for the omission of a risk category. Actually a sample using a
risk variable, the beta coefficient calculated on quaterly data, was
run in a preliminary screening effort. With a sample of 30 firms, the
beta coefficient was not found to be statistically significant. Even
with these results, the beta coefficient would have been used in the
model if its computation had not drastically.reduced the number of
firms having the required data.
Since the current ratio for utilities is subjected to a different
interpretation, it was dropped for all but industrial firms. Return on
book value was substituted for the current ratio in the analysis of
public utilities since their profits are regulated. This additional
return variable is more relevant for utilities. The resulting analysis,
therefore, employed eleven independent variables for both utilities and
Since the coefficients of correlation between the size variables,
total dollar dividends, number of shareholders outstanding, and total
assets, were .83, .83 and .89, only the first of the three variables
to enter the stepwise procedure in the MDA model was included in the
analysis. The remaining two were omitted. Likewise, one growth
variable was permitted in the final discriminant and regression equa-
tions. In the utility model only the most significant of the two
return variables was allowed in the final equation.
The multiple discriminant analysis study of profiles of users
versus nonusers had a sample of 205 industrial firms and 41 utilities.
No railroads were used in either analysis. The great reduction in in-
dustrial firms from 425 to 205 was the result of two considerations.
If data items were missing or if EPS were negative in one of the
years used to compute the geometric mean, that firm was omitted from
the study. It is admitted that the decision to compute the geometric
mean excluded some 15 firms from the study. The preclusion of these
firms contains in itself an element of bias in that the geometric
mean could not be calculated without using imaginary numbers if the
earnings per share were negative. Thus some of the weaker firms were
probably excluded. In some cases the negative figure for EPS was ap-
parently a freak occurrence and happened only once in the time span.
In others it was not, and reflected the weakness of the firm. At any
rate, the decision to use the variable was responsible for the preclu-
sion of several firms. Since only a small number of firms were
eliminated due to the inclusion of the geometric mean, the problem ap-
pears to be minor. The other alternative or the omission of the
geometric mean as a variable would seem a more questionable approach.
Also a large number of firms were eliminated because they were
on a different calendar year than the sample. It was felt that this
step was necessary in a cross-sectorial study to insure that the firms
were being compared for the same time period. The samples for the
industrials and utilities were then broken down into users and non-
users. Consequently, there were 72 users and 133 nonusers for
industrial firms. The utility breakdown was 26 and 15 respectively.
The Multiple Regression Model
Multiple regression analysis was then employed to see if partic-
ipation by shareholders in ADR plans followed from any financial
characteristics of the firm. All the independent variables employed
in multiple discriminant analysis were used except for the industrial
utilization figure which was discarded. It was felt that the share-
holders would not be guided by the concept of an industry norm as
management might. Two measures of a dependent variable were obtained
from the questionnaire responses. Either the percent of total share-
holders or the percentage of shares outstanding could serve as an
indicator of the degree of participation in the ADR plans.
Several factors should be considered in determining the promi-
nence and applicability of each participation measure. First of all
what is the role that financial variables play in determining whether
or not investors participate in ADR plans? That is, does a particu-
lar type of investor tend to affiliate himself with a firm with
characteristics that suit his individual preferences? Such prefer-
ences could be in terms of a risk-return criterion or perhaps in
terms of a preference for income or capital gains which might be in-
herent in the growth and dividend payout ratio of the firm. Since
it was impossible to get information on individual shareholders and
study their characteristics as independent variables influencing the
participation ratio, it was assumed that the characteristics of the
shareholder are inherent in the characteristics of the firm. The
decision was thus made to use characteristics of the firm as a sur-
rogate for shareholder characteristics. With this assumption also
is an acceptance of a tendency for shareholders to invest in ventures
that satisfy other preferences in addition to earnings. In other
words, the assumed link between investor preferences is a result of
accepting the "clientele theory" as well as the notion of
"rationality." This linkage between investor preferences or
characteristics and firm characteristics is clouded by the prominence
of the institution investor in the market. It is presumed that the
institutional investor's position in a stock may affect the degree of
participation profoundly. While it is possible to infer that the
institutional investor may be particularly attracted to certain risk-
return considerations, the resolution of the conflict between income
and capital gains may be even more dominating. Endowments, for
instance, have more or less traditionally limited expenditures to
cash spinoffs from dividends and have shown a reluctance to liquidate
stock holdings. Presumably, institutional investors being less af-
fected by transfer charges and more sophisticated with regard to the
concept of diversification may not rely on a dividend reinvestment
mechanism to the extent that the small investor does. The same is
true of the large investor but his preferences and characteristics,
if logical, should be more discernible in the characteristics of the
firm than the institutional investor. Since the institutional
investor's behavior is different with respect to ADR plans than the
individual, the position that an institutional investor has in a
particular stock may then serve to obscure characteristics of share-
holders if that position is quite large. Participation percentages
are thus quite likely to be affected by the position of the
institutional investor as well as the preference of shareholders for
income or capital gains. For this reason the percent of shareholders
participating may be a better measure than the percentage of shares
outstanding. If the percent of shareholders is used as a dependent
variable, the weight given to the institutional investor is minimized
and its decision is not as prominent as it would be if the percentage
of shares outstanding were used as the dependent variable.
Sufficient data for computation of variables were available for
34 industrial firms and 15 public utilities. The primary reason for
the sharp reduction from the sample of firms employed in the discrim-
inant model was due to the fact that the dependent variable came from
the questionnaire. Consequently, only those firms in the discriminant
model sample which were classified as users and answered the question-
naire made up the regression model sample.
CHARACTERISTIC PROFILES OF ADR USERS
One result of the rapid growth in the number of existing ADR
plans is that financial executives in firms that do not offer an ADR
option have been troubled by the decision of whether or not an ADR
plan would be beneficial to the firm. Two questions that they most
commonly asked have been, "What type of firm has historically
offered ADR plans?" and "Have firms in some industries utilized ADR
plans to a greater extent than firms in other industries?". The
purpose of this chapter is to answer questions of this nature by
developing profiles of users and nonusers.
The research technique employed to differentiate characteris-
tics of users and nonusers were multiple stepwise discriminant
analysis and a test of the difference between two sample means. The
multiple discriminant analysis (MDA) model employed eleven indepen-
dent variables and one dependent variable. The dependent variable
was one if the firm was a user and zero if the firm was a nonuser
of ADR plans. The independent variables used were classified into
the seven following categories:
XI Three-year weighted average yield (change in market
price plus dividend income divided by initial market
price) for the years 1970-1972.
X Average price earnings ratio for 1972 (high + low share
2 price divided by EPS).
X3 Growth rate in EPS, the arithmetic mean, based on yearly
X4 Growth rate in EPS, the geometric mean, based on yearly
X5 Current ratio for 1972.
X6 Debt over net worth for 1972.
X7 Total dollar dividends paid in 1972.
X8 Total number of shareholders as of the end of 1972.
X9 Total assets as of the end of 1972.
X10- Average dividend payout ratio for 10 years, 1963-1972,
after throwing out the high and the low.
X11- Industry utilization factor based on S&P industry clas-
sification and equal to the number of firms in an
industry using ADR's divided by the number of firms in
The Collection of Information
The data for the above variables were obtained from the two
surveys and from the Compustat Tapes. Identification of the firms
as ADR users or nonusers was a joint effort of the bank and firm
surveys. The a priori classification of firms for the MDA model
required the identification of all firms in the model as users or
nonusers. In the test of the difference between two sample means
the users became one sample and the nonusers the other. In both
statistical procedures the industrials and the public utilities
were handled separately. The empirical results of these statisti-
cal techniques for the industrials and the utilities will be con-
sidered in turn.
The Generation of Samples
Multiple discriminant analysis and a test of the statistical
difference between two sample means were performed separately for
the 205 industrial firms and the 41 public utilities.
In the MDA model the decision was made to split the industrial
sample into two parts to eliminate the possibility of search bias.
A function-generating sample of 103 firms and a holdout sample of
102 firms were then constructed by using a random number generator
to separate the industrial firms into two groups. The function-
generating sample contained 36 users and 67 nonusers. The holdout
sample contained 36 users and 66 nonusers. Stepwise multiple dis-
criminant analysis was then employed to fit a function to the
generating sample. In turn, that function was used to classify
the holdout sample.
The utilities were handled a little differently in that the
total sample was too small to split; the function-generating
sample and the classification sample were the same. Frank, Massey,
and Morrison1 noted that search bias may result unless the sample
is split. Consequently, randomness may be responsible for the
classifying power of some variables rather than any functional re-
lationship. It was felt that splitting the sample would result in
1Ronald E. Frank, William F. Massey, and Donald G. Morrison,
"Bias in Multiple Discriminant Analysis," Journal of Marketing
Research, II (August, 1965), pp. 252-253.
such a small function-generating and holdout sample, particularly
for the nonuser, that the possible error due to search bias would
be more than offset by the fact that sample sizes were so small.
Additionally, it should be pointed out that the total sample of 41
utilities comprised a large portion of the 55 utilities that the
sample sought to describe. The utility samples were identical to
those in the test of the difference in two means.
Test of the Difference Between Two Means
A test of the statistical difference between the means of the
two samples provided some information on the differences in profiles
between the sample of users and nonusers. A Z score was determined
for each of the eleven variables presented in Chapter III. The Z
scores for the industrials are presented in Table II. A t test was
performed on the utilities because of the small sample size. Both
the t value and the Z value were determined by the following equa-
tion where the subscript denotes the sample (i.e., 1 is for the
users and 2 is for the nonusers):
(4-1) = Z value or t value.
2 2 1/2
(S1 /N1 + S2 /N2)
The t value or the Z value is then defined as the difference between
the two sample means divided by the square root of the sum of the
respective variances over the sample size of the group. In the case
of the industrial firms where the sample sizes are 72 and 133, a Z
score cutoff can be used. to determine, within the confidence level,
if the difference is of statistical significance. Because of the
smaller sample sizes used for the utilities, a t score cutoff is used
with the figures in Table II for utilities. If the t value or Z value
is greater than the t cutoff or the Z cutoff, the hypothesis that
there is no statistical difference is rejected and the mean of the
group of ADR users is greater than the mean of nonusers for that vari-
able. If the t value or the Z value is less than a minus t cutoff or
a minus Z cutoff, the hypothesis that there is no statistical differ-
ence is again rejected. However, this time the mean of the user group
is statistically smaller than for the nonuser group for that particu-
lar variable. Using a two-tailed test for statistical significance
at a 95 percent confidence level, the cutoff score for the normal dis-
tribution was determined to be 1.96. The only statistically signifi-
cant variables at the .05 level are (from Table II) the average price
earning ratio, growth in EPS (arithmetic mean), debt over net worth,
the average dividend payout ratio, and the industry utilization
Determination of the degrees of freedom needed to characterize
the student distribution for a small sample size with unequal
variances2 is as follows:
2 2 2
(S12/n1 + S22/n2)2
(4-2) v = 2 .
n1 1 n 1
None of the t values for the variables of the utilities were
significant at the .05 level. The largest t value was 1.4756 for
2Roger L. Burford, Statistics: A Computer Approach (Columbus,
Ohio: Charles E. Merrill Publishing Company, 1968), p. 245.
Z VALUES OF INDUSTRIAL FIRMS CALCULATED FOR SELECT VARIABLES
Variable Z Value
1 Weighted Average Yield -0.4792
2 Average Price Earnings Ratio -3.3037
3 Growth Rate in EPS (Arithmetic Mean) -2.1708
4 Growth Rate in EPS (Geometric Mean) 0.9558
5 Current Ratio -1.0686
6 Debt Over Net Worth -1.9827
7 Total Dollar Dividends 1.4699
8 Number of Shareholders 1.3870
9 Total Assets 1.3265
10 Average Dividend Payout Ratio 2.5893
11 Industry Utilization Factor 8.2785
the dividend payout ratio. If a normal distribution had been assumed
for the utilities, a t value of 1.4756 would not be significant. The
cutoff values for a given level of significance are always higher for
the t distribution than for the normal distribution unless the degrees
of freedom are very large and then the t cutoff would approach the Z
cutoff. Consequently, the t value for the average dividend payout
ratio would not be significant even if a normal distribution could be
assumed; likewise, it would not be statistically significant for a t
distribution. The test is thus inconclusive at least for the
Multiple Discriminant Analysis
Stepwise multiple discriminant analysis was used to distinguish
between users and nonusers for both industrial firms and utilities.
Tables III and V summarize the steps by denoting which variables
entered, the F value to enter, and the number of degrees of freedom
remaining in the equation for industrials and utilities. The F values
of the first two variables to enter the stepwise regression models for
the industrials are both statistically significant at the .05 level.
The critical F value for the .05 level with 1 degree of freedom in the
numerator and 100 degrees of freedom in the denominator is 3.94.
The discriminant function developed by the stepwise procedure
was then employed to classify companies as being either users or non-
users of ADR plans. The results are provided in Table IV. As Table
IV reveals, the function correctly classified 24 of the 36 ADR users
or 66.7 percent and 49 of the 66 nonusers or 74.2 percent. A chance
classification3 could be expected to identify 35.3 percent of the
users and 64.7 percent of the nonusers.
3A chance classification is a rather complicated issue when ap-
plied to a confusion matrix such as the one in Table IV. Obviously
the best grouping in terms of percentages would be to classify all
firms as nonusers. In the industrial sample the resulting classifi-
cation would thus correctly identify all 66 nonusers but would mis-
classify all 36 users. The percentage correctly classified would
thus be 64.7 percent. However, the resulting classification would
not be consistent with the purpose of this chapter in that none of
the users would be correctly identified. An alternate chance ap-
proach would be to classify half of the 102 firms as users and the
other half as nonusers. On the average this approach would correct-
ly classify 18 of the users and 33 of the nonusers or 50 percent.
The problem with this approach lies in the fact that equal weights
are given to two groups of unequal size. The particular chance clas-
sification used assumes that firms are classified in accordance with
group size. Therefore, 66 firms would be classified as nonusers and
36 as users. On the average 42.7 firms would be correctly classified
SUMMARY STATISTICS OF STEPWISE MULTIPLE
DISCRIMINANT ANALYSIS FOR INDUSTRIALS
Step Entering Variable F-value Dgrees of
to Enter Freedom
1 Industry Utilization Factor 36.7656 1/101
2 Average Price Earnings Ratio 9.6351 1/100
3 Total Dollar Dividends 2.3560 1/ 99
CLASSIFICATION MATRIX FOR INDUSTRIAL
User 24 12
Nonuser 17 49
When the MDA model was applied to the sample of public utilities,
none of the F values to enter were statistically significant at the
.05 level for the utility variables in Table V. The critical value
for the F distribution at that level with 1 degree of freedom in the
as nonusers and 12.7 firms correctly classified as users. The result
is that an average of 55.4 of the 102 firms are correctly classified.
The chance model should be expected to correctly classify 54.3 per-
cent. The discriminant model above actually classified 73 of the 102
firms correctly or 71.6 percent. For additional information on the
appropriate chance criterion to use see Donald G. Morrison, "On the
Interpretation of Discriminant Analysis," Journal of Marketing
Research, VI (May, 1969), pp. 157-158.
numerator and 39 degrees of freedom in the denominator is 4.11. Since
the F value of the dividend payout ratio is not greater than 4.11,
that variable is not statistically significant at the .05 level. There-
fore, none of the other variables are statistically significant.
SUMMARY STATISTICS OF STEPWISE MULTIPLE
DISCRIMINANT ANALYSIS FOR UTILITIES
Average Dividend Payout
Average Price Earnings Ratio
Total Number of Shareholders
The utility discriminant function correctly classified 20 of the
26 users or 76.9 percent and 9 of the 15 nonusers or 60 percent. A
chance classification4 could be expected to correctly identify 63.4
percent of the users and 37.6 percent of the nonusers.
CLASSIFICATION MATRIX FOR UTILITY
User 20 6
Nonuser 6 9
4A chance classification could be expected to correctly identify
16.5 users and 5.5 nonusers. The percentage correctly classified by
chance would then be 53.7 percent. The discriminant model actually
correctly classified 29 of the 41 firms correctly or 70.7 percent.
Problems and Limitations
Application of the multiple discriminant model to the data col-
lected was subjected to the following problems: (1) the technique
employed to identify ADR users and nonusers may not have been complete-
ly accurate, (2) the proper classification techniques may not have
been linear and (3) statistical problems such as specification bias,
multicollinearity and spurious correlation may have been present.
In fact, positive identification of the users was not made with
certainty. The procedure followed was to identify as many users as
possible from the survey of the S&P 500 and the banks. Firms that were
not subsequently identified as users were then concluded to be nonusers.
Thus, there exists the possibility that some firms having established
ADR plans may erroneously be classified as nonusers. While the possi-
bility of this error does exist, the probability is not very large.
Actually, slight errors of misclassification should not have a great ef-
fect upon the ability of the discriminant function to correctly identify
users and nonusers.
The linear relationship assumption was of greater concern. Linear
decision rules are appropriate in discriminant analysis when the group
dispersion matrices covariancee matrices) are equal. While this study
utilized linear classification rules, no determination was made as to
whether or not the covariance matrices of users and nonusers were equal.
The alternative of using a quadratic function was not viewed as being
attractive even if the covariance matrices were unequal.
It has been substantiated that it may not be necessary to employ
quadratic function. One argument that has been advanced is that the
respective covariance matrices only have to approximate each other for
a linear relationship to suffice. Documentation to this effect from
the Eisenbeis and Avery book on discriminant analysis is as follows:
There is some evidence that nonmultivariate normal data may be
used in a discriminant analysis without significantly biasing
the results. For the two group case, Gilbert has compared the
performance of linear classification rules (estimating only
dichotomous variables) with two rules based upon a logistic
model and a rule formed on the assumption of mutual indepen-
dence of the variables. The results imply that the loss using
a linear rule "for classification as opposed to any other pro-
cedure is too small to be of much importance." Of course, as
the number of variables increases, the Central Limit Theorem
implies that the distribution of the discriminant scores for
each group approaches a normal distribution. Hence, classifi-
cation can be performed in the reduced discriminant space.5
Another argument is that if the discriminant function is able to iden-
tify the holdout sample well, a linear function is not inapplicable.
It is conceivable that a quadratic function may perform better, but not
Specification bias and multicollinearity were the two statistical
problem areas felt to be inherent in the use of MDA and in a test of
the difference between two sample means for ADR users and nonusers.
Specification bias is present when the dependent variable is a function
of an omitted independent variable. It is also necessary that the
omitted variable be correlated with an independent variable that is in
the regression or discriminant equation. In particular, stepwise pro-
cedures are subjected to this problem because variables that are cor-
related with an independent variable in the equation are suppressed in
the F test. A univariate t test is subjected to the same criticism.
Multicollinearity is a problem for which stepwise discriminant
analysis adjusts to a great extent. Variables enter the stepwise pro-
Robert A. Eisenbeis and Robert B. Avery, Discriminant Analysis
and Classification Procedures (Lexington, Massachusetts: D.C. Heath
and Company, 1972), p. 37.
cedure in terms of the magnitude of the predictive power that each adds.
Thus, a variable that is highly correlated with one that is already in
solution will be suppressed in that its F ratio will not be very high.
Variables with the highest F ratio enter the stepwise procedure first.
Consequently, it was not surprising that the coefficients of correla-
tions between variables in the final discriminant equations were small.
The highest correlation coefficient between any of the variables in the
final discriminant equation for the industrials was .07374. The value
of the coefficient in the discriminant equation for utilities indicat-
ing the greatest correlation was -.18432. Since the t value calculated
in the test for a statistical difference between two sample means is
univariate, multicollinearity is not a problem.
A third potentially serious statistical problem follows from the
manner in which the industry utilization factor was calculated. The
numerator of the independent variable is the number of firms in an in-
dustry with established ADR plans. Its denominator is the number of
firms in an industry. The dependent variable in the MDA model is one
if a firm is a user and zero if the firm is a nonuser. Since the a
priori identification of whether a firm is a user or nonuser was an in-
put into the calculation of the utilization factor and determined the
independent variable as well, spurious correlation must exist.
The question then arises as to whether or not the predictive
power of the utilization factor is a result of the spurious correlation.
If not, is there some kind of industry effect that has a bearing on
whether or not a firm may be inclined to establish an ADR plan?
In defense of using the variable, it should be pointed out that
the a priori classification of each firm is only a component of the
utilization factor. The actual proportion of the variable that a par-
ticular firm accounts for is equal to the reciprocal of the number of
firms in an industry. Consequently, as the size of the industry
increases, the degree of spurious correlation should decrease. The av-
erage number of firms per industry in the discriminant model for the
industrials was 5.4 The modal class contained 6 firms. Of the 63 in-
dustry classifications used in the model, 52 had 4 or more firms. The
smallest number of firms in any classification was 2; the largest was
Also, the utilization factor will be the same for users and non-
users within an industry. Therefore, the utilization factor does not
reveal whether a firm is a user or nonuser. What it does reveal is
the percentage of firms in an industry which have established ADR plans.
For example, in an industry of 5 firms with 2 users, the utilization
factor is .40 or 40 percent for all firms in that industry.
The principal argument against the use of the utilization factor
contends that to use the variable is akin to employing knowledge of
whether a firm is a user or nonuser as an independent variable. Thus,
a priori classification of firms is necessary in order to discriminate
between users and nonusers.
This study acknowledges that spurious correlation does exist and
may be responsible for some of the predictive power of the model. How-
ever, the industry effect may still be of merit. As previously stated,
discriminant analysis has a dual application. It can be used as a pre-
dictive tool or it can be used to determine characteristics that
distinguish dichotomous groups such as users and nonusers.
In the purely predictive sense, spurious correlation does not
constitute a shortcoming. In fact, the a priori classification of users
and nonusers is a prerequisite for employing discriminant analysis. Any
problem that arises is attributable to the use of MDA to develop charac-
teristic profiles of users and nonusers.
Since spurious correlation may have invalidated any conclusions
that the discriminant model reached on the utilization factor, a chi-
square (X2) test was also employed to ascertain whether there was any
evidence of an industry effect. The hypothesis tested was that the
number of firms in an industry utilizing ADR plans is independent of
the industry classification. The categories for the test of independ-
ence were the number of firms in an industry employing ADR plans.Since
109 out of the 425 industrial firms have established ADR plans, the
probability of an industrial firm having ADR plans is 25.65 percent.
Expected values were then calculated using a binomial distribution with
the probability equal to .25. The number of firms in an industrial
classification became the sample size.
The binomial distribution was used to determine the expected
number of firms for different industrial classifications that would
have employed ADR plans if the utilization of these plans were independ-
ent of any industrial effect. The expected number of firms for each
category was determined by summing the above vAlues. Chi-square was
calculated using the following equation:
N (0 Ei)2
(4-3) X2 = E E
where i denotes the category, N is the number of categories, 0 represents
observed number of firms in an industrial classification and E is the
expected number of firms. The calculated value of X2 was determined
to be 9.98. The X2 value from the table with 3 degrees of freedom at
the .05 level was 7.815. The hypothesis is thus rejected. Evidently,
there appears to be evidence that some industry effect may exist.
The discriminant model was also run for industrials without the
utilization factor. The only variable with an F to enter significant
at the .05 level was the average price earnings ratio. Other variables
in the discriminant equation were (in order of appearance) total dollar
dividends, debt to net worth, average dividend payout ratio and growth
in EPS (geometric mean). The equation, thus generated, correctly clas-
sified 64 percent of the users and 46 percent of the nonusers in the
holdout sample. It was previously determined that a chance classifica-
tion could be expected to identify 35.3 percent of the users and 64.7
percent of the nonusers. Consequently, the model could not be said to
differ much from a chance model in terms of predictive power. The loss
of the utilization factor evidently had a pronounced effect upon the
predictive power of the MDA model.
Both the test of the difference between two sample means and the
stepwise discriminant model failed to provide any conclusive results
for the utility sample. Part of the problem appeared to be that the
financial characteristics within the utility industry are quite similar.
This should not be unexpected for a regulated industry producing
products which are practically identical. Even though none of the
three variables in solution have statistically significant F values, it
is particularly interesting to note that the first variable to enter
the stepwise procedure was the dividend payout ratio. A size variable
(total number of shareholders outstanding) was the final variable to
enter. Although there were slight differences in the variables included
in the two empirical models, the results for the industrial firms and
the public utilities were comparable.
Five independent variables were significant in the tests for
statistical differences between ADR users and nonusers in the sample
for industrial firms. The significant variables were: the average
price earnings ratio, growth in EPS (arithmetic mean), debt over net
worth, the average divident payout ratio and the industry utilization
factor. Only two variables had statistically significant F values to
enter in the MDA model. They were the industry utilization factor and
the average price earnings ratio.
It was interesting to note that the industry utilization factor
had, by far, the largest F to enter and Z value. It was apparent that
the industry norm was responsible for much of the predictive power of
the MDA model. However, caution should be urged in determining the
role that the industry effect plays. While the X2 test would appear
to indicate that there is some industry effect, it may not be as pro-
nounced as the discriminant model for industrial firms may lead one to
The second variable to enter the equation, the average price
earnings ratio, had the second largest F to enter and Z value. While
it is true that the variable with the largest Z value should be the
first to enter a stepwise discriminant model and should have the
largest F to enter, a similar analogy does not necessarily hold for
the variable with the second largest Z value. This is likely to be
true if there is strong correlation between the variables that have
the two largest Z values. Once the first variable is in solution in the
stepwise model, the discriminatory power of the second variable is
reduced and its F value to enter may drop below the F value to enter of
variables with smaller Z values which are not as strongly correlated
with the first variable.
There exists a strong a priori reason to believe that correlation
exists in varying degrees between the industry utilization factor and
many of the other independent variables in the model. The reason this
correlation is likely to exist is that an industrial classification at-
tempts to group firms with similar products. Since financial charac-
teristics often prove similar within industrial groupings, less varia-
bility in these characteristics takes place within an industrial group-
ing than between the different industrial classifications. Consequently,
it appears that the MDA model may suffer from specification bias if the
utilization factor is credited with explaining differences in the two
groups that are actually a function of other financial variables with
which the utilization factor is correlated. On the other hand, the
utilization factor seems to afford more explanatory power than the
other independent variables. Thus, the variable may offer some tangible
source of explanation of true differences between users and nonusers.
Perhaps that difference is little more than an industrial leader-
ship effect where the largest firms in some industries implement ADR
plans and the smaller firms follow in turn. This effect of trying to
"keep up with the Joneses" may exist in the industry sector as well as
in consumer behavior. Some evidence to support such an effect is evi-
dent. The Z value for all the size variables are positive. This
means that the average of total assets, number of shareholders, and
dollar dividends are higher for firms with ADR plans than for those
without. In the classical test of hypothesis sense, statistical
significance cannot be proven in that none of the Z values are greater
than 1.96. However, it should be remembered that this does not dis-
prove the possibility either. Also, the fact that all three are posi-
tive and that the smallest Z value of the three is 1.3265 should rein-
force the premise that large firms are more likely to establish ADR
plans than small firms.
However, the result of the test of the difference between two
sample means was felt to be superior in developing profiles because of
the a priori belief that the utilization factor and the use of step-
wise discriminant analysis resulted in the omission of variables that
characterize ADR users and nonusers. It was apparent from this statis-
tical test that industrial firms which are ADR users have lower average
price earnings ratios, lower growth rates, lower ratios of debt to net
worth and higher average dividend payout ratios. Perhaps firms that
establish ADR plans tend to be low risk and low growth firms.
As a result of the ADR movement, financial executives are show-
ing concern for answers to questions such as: (1) How do ADR plans
work? (2) Who administers them? (3) What are the costs to partici-
pants? (4) Are there any legal complications? (5) Do the firms and
shareholders realize any benefits? (6) What kind of participation
rate should a firm expect? (7) Why have some firms established ADR
plan options while others have not? (8) Is there any inherent infor-
mation in the ADR experience which should affect dividend policy?
The purpose of this chapter is to answer these and other ques-
tions that are a result of the rapid growth in the ADR movement. In
collecting information to answer questions of this nature, banks and
firms were surveyed on the subject. Information from the surveys
will be presented and analyzed after the mechanics of automatic divi-
dend reinvestment plans have been discussed.
The Nature of Automatic Dividend Reinvestment Plans1
Operating Characteristics of ADR Plans
Although the automatic reinvestment of dividends has for some
time been an option to shareholders of mutual funds, savings and loan
associations, mutual savings banks, and some insurance companies,
nonfinancial corporations have only recently begun to provide this
'Much of the discussion on the nature of ADR plans parallels
Richard H. Pettway and R. Phil Malone, "Automatic Dividend Reinvest-
ment Plans of Nonfinancial Corporations," Financial Management (in
option to stockholders. While the concept is quite comparable, the
actual plans available for shareholders in nonfinancial corporations
are still distinctly different from those of financial institutions.
For instance, open-end mutual funds provide for the issuance of new
shares to participants while the majority of corporate ADR plans
deal in outstanding issues and charge shareholders a fee to partici-
pate. Some mutual funds do charge a load fee which is likely to be
higher than the administrative and brokerage charges to participate
in ADR plans.
The similarities between corporation ADR plans and the dividend
reinvestment programs of mutual funds are multifold. Each utilizes
the concept of pooling ownership of certificates in an account with-
out necessitating the issuance of the security in the shareholder's
name. Ownership information is provided by the issuance of periodic
statements showing the number of shares held in trust, the receipt of
cash dividends, and the subsequent purchase of shares. Information
of this nature is necessary since dividends are taxable even though
shareholders may not actually receive a cash dividend.
The automatic dividend reinvestment plans offered by nonfinan-
cial firms are basically of two types: the bank-trust department
administered plans and the firm-administered plans. As of November 1,
1973, 265 nonfinancial corporations had provided their shareholders
with an ADR option. Over 98 percent of these plans are bank
administered. Under bank-administered plans the corporations pay the
cash dividends of stockholders, who elect to participate, directly to
the trust department of the bank. The trust department maintains ac-
2Only three of the nonfinancial plans are firm managed.
Only three of the nonfinancial plans are firm managed.
counts for each participating shareholder. The managing trust depart-
ment purchases the firm's common shares in the open market and dis-
tributes these shares to the shareholders' accounts on a prorata basis
of the actual amount of the dividends received in each account less
proratedbrokerage costs and less some administrative costs as a per-
centage of the dollar dividend reinvested. The marginal rate of the
predetermined administrative fee commonly decreases with the size of
dividends reinvested. This fee is frequently subjected to some
maximum. The brokerage fees are quite likely to be significantly less
than they would have been for a comparable purchase on the market by
the individual shareholder. The total charge is likely to be less for
the small investor who may purchase in odd lots and is consequently
subjected to the odd-lot differential.
The bank, on the other hand, aggregates the purchases of all the
participating shareholders and in some instances may purchase large
blocks off the exchange. Many banks, which are currently administer-
ing ADR plans for firms, have purchased or developed computer programs
to handle the administrative details which carry share division among
individual investors to three decimal places.
In most cases a stockholder who elects to participate in an ADR
plan may make contributions in addition to the total dollar dividends.
Frequently, limits (such as $500 or $1,000 in additional investments
per dividend period) are placed on non-dividend reinvestments. Most
plans also allow stockholders to liquidate their investment at any
time. Typically the shareholder can withdraw and elect to take shares
of stock or often have the bank convert the shares into cash at little
or no charge. Thus, stockholders are actually given a remarkable
degree of flexibility and make only a short-run commitment to reinvest
Recently a unique type of bank-managed plan has evolved. Unlike
the traditional plan it allows shareholders to purchase newly issued
rather than outstanding shares of the company's common stock. The
first company to offer such a plan was Long Island Lighting Company in
November 1972. Currently three additional firms have a similar plan.
Allowing for reinvestment into a new issue is a recent and interesting
departure from the normal type of ADR plans and more will be said
about this departure in later chapters.
Although specific details of the purchase of outstanding stock
by the administrator (bank or firm) vary from plan to plan, generally
the critical price used in the process of reinvesting dividends into
existing shares is the average of all costs actually incurred during
the purchase period. The administrator is allowed to purchase in any
market; all purchases normally are made within 30 days of the divi-
dend payment date. When the plan purchases new shares, the adminis-
trator either uses the closing price or an average of high and low
prices of the shares traded on the NYSE on the dividend payment date
as the exchange price of new stock for the shareholders' dividends.
There are only three firm-managed programs. In July 1972,
General Motors was the first company to establish such a program. Cur-
rently the General Motors' plan deals only in outstanding issues of
stock. Two other large firms, American Telephone and Telegraph and
International Paper Company, have plans that deal with new issues of
The experiences of American Telephone and Telegraph are inter-
testing and worth summarizing. American Telephone and Telegraph Company
was one of the first firms to establish an ADR plan (in 1969 with the
First National City Bank). In June of 1973 the firm allowed its share-
holders to participate in a firm-administered plan which superseded the
old ADR bank-administered plan. The new ADR plan allowed the
reinvestment of participating shareholders' dividends to be used to
purchase only new common stock of the firm. Under the old plan the
firm experienced a participation of about 7 1/2 percent of the number
of shares outstanding with a volume of about $70 million of market
purchases in 1972. Under the new plan AT&T expects about 12 percent
participation and about $160 million of new equity sales annually.3
Classifying a plan as firm-administered does not mean that a
bank is not acting as transfer agent but rather that the firm itself
has retained the responsibility of the record keeping and administra-
tion of the ADR plan. Firms which manage their own plan do employ
banks to act as transfer agents.
Perhaps one of the most interesting differences between the four
types of plans is the cost factor. Table VII was prepared to facili-
tate cost comparisons between the various plans. Not all 265 firms
are included as cost information was not obtainable either because of
reluctance on the part of some banks to provide this information on a
firm-by-firm basis or due to the failure of firms to reply to inquiry.
However, sufficiently detailed information was obtained on the cost of
235 different plans.
3Pettway and Malone (in press).
COST TO PARTICIPANTS
FOR VARIOUS TYPES OF ADR PLANS*
Type of Plan Administrative Fee Brokerage Number
Charges of Firms
Bank Administered 2% up to $2.00 maximum prorated 3
dealing only in 2 1/2% up to $2.50 maximum prorated 1
outstanding issues 4% up to $2.50 maximum prorated 25
5% up to $2.50 maximum prorated 188
5% up to $3.00 maximum prorated 11
Bank Administered no charge no charge 4
dealing in new
Firm Administered 2% up to $2.00 maximum prorated 1
dealing in out-
Firm Administered no charge no charge 2
dealing in new
*Richard H. Pettway and R. Phil Malone, "Automatic Dividend
Reinvestment Plans of Nonfinancial Corporations," Financial Manage-
ment (in press).
Most of the firms on which information was available charge an
administrative fee of 5 percent and prorated brokerage cost. How-
ever, none of the 4 plans dealing in new issues assessed charges
against participating shareholders and the one firm-managed plan
charged participating shareholders considerably less than the norm
for bank-managed plans.
While there are currently no legal requirements specifically
relating to ADR plans, certain existing rules and regulations are being
applied by the Securities and Exchange Commission (S.E.C.). The
Investment Company Act of 1940 and the Securities Act of 1933, in par-
ticular, have been applied if an ADR plan is managed by an "affiliate"
or if the plan provides for the issuance of new shares of common stock.
If there is no "affiliate" relationship between the firm and the
administrative agent and if the plan deals only in outstanding shares
purchased on the market, requirements under these two acts are general-
ly inapplicable. The majority of ADR plans, being bank administered
and dealing only in outstanding issues, have been exempt from these
If the S.E.C. determines that an "affiliate" relationship exists
such as with firm-administered plans, then the rules in Release Number
4790 are applied. Release Number 4790 was designed to apply to
employee stock purchase plans and determines whether registration under
the Securities Act of 1933 is necessary. Departures from the standards
set by that release may require registration. In the event that a
plan deals in new issues of common stock, such as the AT&T plan, then
all pertinent requirements of the Securities Act of 1933 are applicable.
The development of ADR plans has been closely followed by the
Securities and Exchange Commission and current interpretations may
change at any time. Further rules and regulations may even be forth-
Advantages and Disadvantages
Financial executives indicated that firms allowing stockholders
to participate in ADR plans often realize several advantages. Plans
which purchase only existing shares of common may generate some share-
holder goodwill among participants. Secondly, small shareholders are
further encouraged to own a larger portion of the company's outstand-
ing stock in that the cost of purchasing additional shares is quite
often less for them than it would have been otherwise. Also, many
corporations have found that they save considerably in terms of the
operating expense normally incurred with the cash dividend payment.
Operating expense reductions result from elimination of the expenses
for that fraction of stockholders who are participating in the rein-
vestment plan. Executives claim that these cost savings offset the
charges which are paid by the company to the administering bank in
the form of incidental expenses for mailing solicitations to partici-
pate, account statements, etc. A final possible advantage to the
firm is that the plans may generate greater shareholder demand for the
stock than would have been present otherwise. Thus, additional buying
pressure may be created for the stock.
If the plan deals in new rather than outstanding issues then
perhaps some additional advantages may be recognized. First, the ad-
dition to equity accounts through the sale of new shares could reduce
the debt ratio and improve the firm's access to capital markets.
Second, it could provide a regular source of new equity as long as
shareholders elected to reinvest dividends. Finally, the firm would
be able to sell new equity at higher net prices than with conventional
flotation due to the almost negligible issue cost if a reinvestment
plan is used.
Possible advantages to participating shareholders were also sug-
gested: (1) purchasing costs for small investors are considerably
less than they would be if the investor had to pay the brokerage
charge that he would incur on an individual purchase, (2) investors
need not wait and accumulate enough cash to purchase a round lot in
order to avoid the odd-lot differential, (3) the purchase of frac-
tional shares is even possible and (4) finally, ADR plans encourage
forced saving, as the shareholder never receives his dividend check
and may not be as likely to consume the proceeds.
From the standpoint of the firm it would also seem that several
possible disadvantages exist. Efforts to terminate an existing plan
that has not been successful might be looked upon unfavorably by
those shareholders who are participants. Also, institutional investors
as well as shareholders who are not interested in this option may view
any additional expenses by the firm that benefit a segment of owner-
ship more than another as unethical.
Even the participating shareholder may be placing himself in a
disadvantageous position. The decision to participate in an ADR, while
only a short-run commitment, still reduces the investor's liquidity in
that he has elected to forego the receipt of cash. The investor's
flexibility is thus somewhat decreased, at least for the short-run.
Also, shareholders who simply reinvest shares may not tend to diversi-
fy themselves as well since the decision to automatically reinvest
dividends limits the resources that can be invested elsewhere.
Two populations were surveyed for information in gathering data
on the ADR movement. Much of the information from the banks was in-
corporated into the background material on the ADR movement earlier
in this chapter. The second population surveyed was the firms which
comprise the S&P 500. In fact, the questionnaire survey of the firms
was the central information collecting effort of this study. Actually,
two questionnaires were employed. The first (see Exhibit I, Chapter
III) sought to provide information on the rationale and the degree of
success, as measured by the reinvestment volume and the participation
percentage, realized by firms with established ADR plans. The second
questionnaire (see Exhibit II, Chapter III) was intended for firms
that had not yet implemented an ADR plan. It sought to provide infor-
mation as to the degree of consideration that had been given to the
possibility of establishing ADR plans. Inherent in such information
should be some indication of the potential for growth in the ADR
movement as well as the rationale for not establishing ADR plans.
The first questionnaire was designed to be answered by firms
with established ADR plans. Of the 284 firms responding to the sur-
vey out of a possible 500, 944 had established or expected to estab-
lish an ADR plan within two months. Thus, of the firms returning the
survey, 33.1 percent had established or soon would establish an ADR
plan. The remaining 190 firms did not have an established ADR plan.
Question one was designed to provide some historical evidence
on the momentum of the ADR movement. The answers to this question
are summarized in Table VIII. A frequency tally by years reveals
that the year 1972 and the first ten months of 1973 were particu-
larly prominent in the ADR movement for the 90 firms that answered
the first question.
4One firm expected to have its plan functional by December,
1973. The other 93 plans were functional.
YEARS IN WHICH FIRMS RESPONDING
TO THE QUESTIONNAIRE ESTABLISHED ADR PLANS*
Year Number of Firms Percent of Total
1968 1 1.1
1969 2 2.2
1970 1 1.1
1971 7 7.8
1972 36 40.0
1973** 43 47.8
Information from questionnaire survey of S&P 500.
** Includes only the ten-month period of January through
The second question sought to ascertain the prevalence of bank-
managed ADR plans as opposed to those in which the administrative pro-
cedure is assumed by the firms themselves. Only three of the firms
indicated that they administered their own plans. Two firms failed to
respond and the remaining eighty-nine employed a commercial bank as
The next question tried to ascertain how successful the firm's
ADR plan was. Tables IX and X represent an attempt to summarize the
information given by the firms in answering question three. Despite
the fact that different class intervals were used, .99 for Table IX
and .49 for Table X, the data show that participation by number of
shareholders is greater than by the number of shares outstanding.
ESTIMATED PARTICIPATION BY
PERCENTAGE OF SHAREHOLDERS FOR 1973 AND 1974
Class Limits Number of Percent Number of Percent
Firms of Total Firms of Total
.00 .99 0 0.0 0 0.0
1.00 1.99 2 2.3 0 0.0
2.00 2.99 3 3.5 1 1.3
3.00 3.99 13 14.9 9 11.4
4.00 4.99 14 16.1 4 5.1
5.00 5.99 12 13.8 9 11.4
6.00 6.99 8 9.2 14 17.7
7.00 7.99 13 14.9 10 12.6
8.00 8.99 11 12.6 11 13.9
9.00 9.99 2 2.3 5 6.3
10.00 10.99 6 6.9 9 11.4
11.00 or above* 3 3.5 7 8.9
87 100.0 79 100.0
*In 1973 only three firms had greater than 11 percent of their
shareholders participating in an ADR plan. One firm had 11.8 percent
and the other two 12 percent. The 1974 estimates result in 7 firms in
that category. The respective estimates are for 11 percent, 11 percent,
11.8 percent, 12 percent, 12 percent, 13.5 percent and 15.0 percent.
THE ESTIMATED PERCENTAGE OF SHARES
OUTSTANDING PARTICIPATING IN ADR PLANS FOR 1973 AND 1974
Class Limits Number of Percent Number of Percent
Firms of Total Firms of Total
.00 .49 10 12.4 4 5.6
.50 .99 23 28.4 14 19.7
1.00 1.49 18 22.2 15 21.1
1.50 1.99 10 12.4 10 14.1
2.00 2.49 4 4.9 8 11.3
2.50 2.99 4 4.9 5 7.1
3.00 3.49 6 7.4 5 7.1
3.50 3.99 0 0.0 2 2.8
4.00 4.49 1 1.2 2 2.8
4.50 4.99 0 0.0 1 1.4
5.00 5.49 3 3.7 2 2.8
5.50 or over* 2 2.5 3 4.2
81 100.0 71 100.0
*In 1973 only two firms had estimated that greater than 5 1/2
percent of the outstanding shares were participating in the plan. One
firm estimated that the participation rate was 6 percent and the other
6 1/2 percent. Three firms had estimates in this category for 1974.
The estimates were 6 1/2 percent, 7 percent, and 8 percent.
Also, in spite of the fact that the number of responses decreases
respectively from 87 to 79 and from 81 to 71 in the two tables, the
expectations are that participation will be higher in 1974 than it
was in 1973, regardless of the measure of participation used.
The fourth question sought to determine the volume of funds
reinvested annually by shareholders through the ADR mechanism. Table
XI is a frequency tally of the responses to that question. The range
of the annual dollar volume is from $20,000 to $160,000,000. Because
of excessive positive skewness the mean was considered to be a rela-
tively poor measure of central tendency and was not calculated. The
modal class is the first class with limits of $0 and less than
$250,000. The median is $383,000. Of the 81 firms answering ques-
tion four, 29 had an annual dollar reinvestment below $250,000. While
only 9 of the responding firms had an annual dollar volume under
$100,000, 22 firms estimated their annual reinvestment as being over
$1,000,000. Approximately 68 percent of the firms had an annual
dollar volume less than $750,000. An annual dollar volume of less
than $1,250,000 was the estimate of 80.1 percent of the responding
Question five made an effort to reveal the most prominent
reasons for the implementation of ADR plans by firms. The responses
to this question are summarized in Table XII. If the respondent
failed to designate the importance of his answers, as indicated in
the questionnaire, all of his selections were treated as first
choices. Two reasons appear to account for the establishment of
most ADR plans. By far the most important reason appears to be that
the plans provide a service for those shareholders who want to save
THE ESTIMATED ANNUAL DOLLAR VOLUME
OF FUNDS INVESTED THROUGH ESTABLISHED ADR PLANS
Class Limits Number of Firms Percent of Total
Less than 250*
5,250 or more**
*The smallest estimated annual dollar volume was $20,000. Nine
of the responding firms or 11.25 percent had an estimated annual
dollar volume less than $100,000.
**The two largest plans in terms of estimated annual dollar vol-
ume were General Motors and American Telephone and Telegraph with
respective estimates of $20,000,000 and $160,000,000. Twenty-two firms
had estimates above $1,000,000. Twenty-seven and one-half percent of
the 81 firms answering that question had estimated annual investments
of over a million dollars.
0 r-l 0 C
mc CO OO CC
C-') H H
r-1 r-H C
ON r,- -o
CU n 0) 'U
aH 0 0
. -4 1
. ( 4- 14.J
0 H r
, ) a-)
and do not want to use current dividends for consumption. The second
most important reason given was that ADR plans encourage small share-
holders to own a larger portion of the company and increase ownership
dispersion. Other reasons cited by respondents are that the plans pro-
vide an inexpensive way for small investors to purchase additional
stock and, from the viewpoint of the firm, ADR plans may improve stock-
Responses to the sixth question revealed that only 5 of the
existing ADR plans allowed for the purchase of new shares of common
stock. Two firms failed to answer the question and the remaining 87
had an ADR plan which dealt only in already outstanding issues of com-
mon stock. The 87 firms which did not utilize new common issues were
then asked whether or not they expected to issue new shares through
ADR plans in the future. Of the 82 responding firms only 3 expected
to utilize new common issue in the near future while 26 firms had re-
jected the possibility for some time. The remaining 53 firms were
either uncertain or had never really seriously considered the possibi-
Finally, Questionnaire I sought to determine if financial execu-
tives would draw upon their ADR experience and change the dividend pay-
out ratio if a "significant" number of shareholders participated in
their automatic dividend reinvestment plan. The seventh question was
responded to by 86 executives. Only 6 felt that the firm's dividend
payout ratio policy might change if a "significant" number of the
firm's shareholders participated in their ADR plan.
Problems arose in the interpretation of answers to this question.
First respondents seemed to have difficulty with the word "significant."
r- L o\ T
r-l C4 Cn CO
t( o U '0
rl U 0
r-14 (J 0
:3 41 4-1
4-1 ( CO
o o a
*H o 0
rl 0 '
a- 0) 4-J
4- ., '-P
(0 ) CO
> 0 c
0 0 U
0 ( 4-
a a o
Secondly, the historical participation by some firms has been so low
that many executives could not conceptually accept the possibility
that participation might be this high. In many cases the firm is un-
likely to see a very high participation rate. In that respect the
question asked them to assume something contrary to experience. Un-
fortunately, no apparent way around these two problems was evident
in the study. There is no way to substitute a participation percent-
age for the word "significant" because different percentages might
well draw different responses from the various executives. Also no
apparent remedy was known which would enable some of the executives
to relate to the possibility that their participation percentage might
be quite high in the future.
The second questionnaire was intended for those firms that had
not yet implemented an automatic dividend reinvestment plan for their
shareholders. Of the firms responding to the survey 190 firms identi-
fied themselves as being in this position. That is, 66.9 percent of
the responding firms did not have an established ADR plan.
The first question sought to gauge these firms as to their pres-
ent position with respect to establishing an ADR plan. The results
are summarized in Table XIV. Only 19 firms had never considered the
possibility while 95 firms had rejected the possibility for the pres-
ent time. The remaining 76 firms which were currently exploring the
possibility were further subdivided according to assessments of their
present position. Consequently, 37 firms had not yet made a decision
while 31 firms felt that it was quite likely that they may establish
an ADR plan in the near future. Only 8 firms felt that it was
C 4 o o o o
P 0 0
14-4 n m
4MH H l ;
(n H 0 w C -) 4o
0 a, 3 0)4J 4
W ) 4) r-l
OO I I d I4
0 to U) w (
Sm ao r
I a ) o ) a)
0E a 0 M
0 d -r 0 C *
H0 3 0 0r 0) 4-
1 4 Or 4 c4
o a a) a a
Se4 m Cj 1Y)
0 0 3
H 0 0 *, 1 0 -i *
S U U 0 -
P M 0) o
3 Q' 4-I 4-I
likely that they would not establish such a plan.
Question number two tried to determine the number of firms that
were exploring plans that involved new equity issues. For this
reason, only those firms that were currently exploring the possibility
of implementing an ADR plan were of interest. The 8 firms that did
not expect to establish an ADR plan in the near future were also
excluded because their response was almost equivalent to a rejection.
Only 7 firms were considering plans that involved new equity issues.
Plans involving outstanding common stock were considered by 55 firms.
Some firms indicated that they were considering both types of plans.
The intent of the next question was to ascertain the most promi-
nent reasons that firms have for rejecting ADR plans. Consequently,
only firms that had rejected the option or felt it likely that they
would reject were tallied in collecting information from this question.
A summary of that tabulation is available in Table XV. Because re-
spondents were asked to order their responses, frequencies in the
table are ranked according to the preferences of the respondents. In
the event that respondents failed to rank choices, all selections were
treated in the tabulations as if each was a first choice.
The response data in Table XV reveal that the predominant
reasons for not establishing an ADR plan were that the dollar volume
of dividend payments for the firm were too low to make such a plan
practical, and that the firm was convinced that most of its share-
holders wanted current income (ie., most stockholders would not be
interested in participating in an ADR plan). Numerous other reasons
were also given by the respondents. The most common was that the firm
paid either little or no cash dividend. The second most common reason
0 ,H 0
r-i L) c' 1- 0
-I f) C14 -It -.I
\ C m
cd ..0 u -U a)
43 I *H
4J- 4. 14-
S ClC) (
* 0 $4 >
4 44 0
U ) U) a) ()
*t ,.- H
4-1 CC > Cd
co "0 .
4J 0 ) a)
a) o0 a)
S-.rl 4J 4J
Cd 3d <
0- 4J O0 4J
o) V1 4 41 a)
0.U 0 p a
*- d 41 0
o 40 44-l
4,l0) C o a
3 ,-H 0 -
0. (o .
4 4 J 0 4
o E o 0
04 0 0 Q) WO
0a) 4UJ 0
4 0 o0 0
0 4-J'3 ()
to Hdo )4
13 0 0 & S
0a a) j W -4J
0 C r
oa 9U it
( C -H U 4
0 4-J E0
P 3 -ri 00)
a) 0 aC) 9
0 w M El 0
aV) 3) -
$ u *- o a) 4
4.1 0 9
o3 0) 0 t) a
0 4 *)Crl
E4 n W