Finance Theory

Robert C. Merton
 Table of Contents

I. Introduction ............................................................................................................. 1

II. On the Arithmetic of Compound Interest: The Time Value of Money................ 8

III. On the Theory of Accumulation and Intertemporal Consumption Choice by

Households in an Environment of Certainty ............................................... 34

IV. On the Role of Business Firms, Financial Instruments and Markets in an

Environment of Certainty............................................................................. 57

V. The "Default-Free" Bond Market and Financial Intermediation in Borrowing

and Lending .................................................................................................. 76

VI. The Value of the Firm Under Certainty ................................................................. 115

VII. The Firm's Investment Decision Under Certainty: Capital Budgeting and
Ranking of New Investment Projects........................................................... 134

VIII. Forward Contracts, Futures Contracts and Options............................................... 151

IX. The Financing Decision by Firms: Impact of Capital Structure Choice on

Value............................................................................................................. 165

X. The Investor's Decision Under Uncertainty: Portfolio Selection ......................... 185

XI. Implications of Portfolio Theory for the Operation of the Capital Markets: The
Capital Asset Pricing Model ........................................................................ 225

XII. Risk-Spreading via Financial Intermediation: Life Insurance .............................. 241

XIII. Optimal Use of Security Analysis and Investment Management .......................... 249

XIV. Theory of Value and Capital Budgeting Under Uncertainty................................. 270

XV. Introduction to Mergers and Acquisitions: Firm Diversification ......................... 287

XVI. The Financing Decision by Firms: Impact of Dividend Policy on Value ............ 296

XVII. Security Pricing and Security Analysis in an Efficient Market............................. 312

Copyright © 1982 by Robert C. Merton. These Notes are not to be reproduced without the author’s
written permission. All rights reserved.
I. INTRODUCTION

Output
Product Markets
Consumption
Manufacturing
Labor Markets or Business
Firms

Investment
Capital Markets Capital
Households
• Stock
Savings • Bond
• Money
• Futures

Savings
Financial Intermediaries
(Borrowings)

Domain of Finance

This course is an introduction to the theory of optimal financial management of households,

business firms, and financial intermediaries. For the term "optimal" to have meaning, a criterion for
measuring performance must be established. For households, it is assumed that each consumer has
a criterion or "utility" function representing his preferences among alternatives, and this set of
preferences is taken as "given" (i.e., as exogenous to the theory). This traditional approach to
households and their tastes does not extend to economic organizations and institutions. That is,
they are regarded as existing primarily because of the functions they serve instead of functioning
primarily because they exist. Economic organizations and institutions, unlike households and their
tastes, are endogenous to the theory. Hence, in the theory of the firm, it is not a fruitful approach to
treat the firm as an "individual" with exogenous preferences. Rather, it is assumed that firms are
created as means to the ends of consumer-investor welfare, and therefore, the criterion function for
judging optimal management of the firm will be endogenous.
In a modern large-scale economy, it is neither practical nor necessary for management to
"poll" the owners of the firm to make decisions. Instead certain data gathered from the capital
markets can be used as "indirect" signals for the determination of the optimal investment and
financing decisions. What the labor and product markets are to the marketing, production and

1
 Robert C. Merton

product-pricing managers, the capital markets are to the financial manager. Hence, a good financial
manager must understand how capital markets work.
Since the capital markets are central, it is quite natural to begin the study of Finance with
the theory of capital markets. To derive the functions of financial markets and institutions, we
investigate the behavior of individual households. Using portfolio selection theory, the households'
demand functions for assets and financial securities are derived to develop the demand side of
capital markets. Taking as given the supply of available assets (i.e., the investment and financing
decisions of business firms), the demands of households are aggregated and equated to aggregate
supplies to determine the equilibrium structure of returns of assets traded in the capital market.
Inspection of the structure of these demand functions leads in a natural way to an introductory
theory for the existence and optimal management of financial intermediaries.
In the second part of the course, the supply side of the capital markets is developed by
studying the optimal management of business firms (given the demand functions of households).
The two elements which make Finance a nontrivial subject are time and uncertainty.
Capital investments often require substantial commitments of resources to earn uncertain cash
flows which may not be generated before some distant future date. It is the financial manager's
responsibility to determine under what conditions such investments should be taken and to ensure
that sufficient funds will be available to take the investments. Because future flows and rates of
return are not known with certainty, to make good decisions, the financial manager must have a
thorough understanding of the tradeoff between risk and return.
While the basic mode of approach has universal application, it should be understood that
the assumed environment is the (reasonably) large corporation in a large-scale economy with well-
developed capital markets and institutions similar to those in the United States. Although the
emphasis is on the private sector, most of the analysis can be applied directly to public sector
financing and investment decisions. However, certain assumptions made in developing the theory
(which are quite reasonable in the assumed environment) will require modification before being
applied to small businesses with limited access to the capital markets or to foreign countries with
significantly different institutional and social structures.

2
Finance Theory

Summary of Different Parts of Finance

Households (Personal Finance)

Taken as Given: 1. A criterion function for choice among alternative consumption
programs
2. Initial endowments

To be Determined: 1. Optimal consumption-saving decision

2. Optimal allocation of savings (portfolio selection)

Manufacturing or Business Firms (Corporate Finance)

Taken as Given: 1. Owners of the firm are households [either directly or through
financial intermediaries]
2. Proper management is to operate the firm in the best interests of the
owners or shareholders
3. The technology or "blueprints" of available projects (including cost
and revenue forecasts) are known either as point values (certainty) or
as probability distributions.

To be Determined: 1. An operation criterion for measuring good management

2. Investment decision in physical assets (capital budgeting)
a. Which assets to invest in
b. How much to invest in total
3. The long-term financing decision
a. Dividend policy
b. Capital structure decisions and the cost of capital
4. The short-term financing decision
a. Management of working capital and cash

3
 Robert C. Merton

5. Mergers and Acquisitions: Firm diversification

6. Taxation and its impact on 2-5 (above)

Financial Intermediaries (Financial Institutions)

Taken as Given: 1. Owners of the intermediary are households [either directly or
through other financial intermediaries]
2. Proper management is to operate the intermediary in the best
interests of the owners or shareholders
To be Determined: 1. Why they exist and what services they provide
2. How the management of financial intermediaries differs from the
management of business firms
3. Efficient management and measurement of performance
4. The role of market makers

Capital Markets and Financial Instruments (Capital Market Finance)

To be Determined: 1. Why they exist and what services they provide
2. The characteristics of an "efficient" capital market
3. How an efficient capital market permits decentralization of decision
making
4. The role of capital markets as a source of information (or "signals")
for efficient decision making by households and managers of
business firms and financial intermediaries
5. The empirical testing of finance theories using capital market
data

4
Finance Theory

Basic Methodology and Approach of the Course

1. How should the system work?

2. Does it work that way?
3. If not, is there an opportunity for improvement (and hence, a profit opportunity)?
4. If you and the market "disagree," then who is right?

Frequently-Used Concepts

Equilibrium: To understand each element of the system, one must frequently analyze the whole
system. To do so, we look at the aggregated resultant of the actions of each unit. If each unit is
choosing the "best" plan possible and the aggregation of the actions implied by these plans are such
that the market clears (i.e., supply equals demand for every item), then these "best" plans can be
realized, and the market is said to be in equilibrium. In general, it will be assumed that the markets
are in or tending toward equilibrium.

Competition: The basic paradigm adopted is that markets operate such that the very best at their
"job" will earn a "fair" return and those that are not will earn a less-than-fair return. This is in
contrast to the view that anyone can earn a "fair" return and the "smart" people will earn a "super"
return. In certain situations, it will be assumed that the capital markets satisfy the technical
conditions of pure competition.

"Perfect" or "Frictionless" Markets: At times, we will use the abstract concept of a perfect market.
That is, there are no transactions costs or other frictions; that there are no institutional restrictions
against market transactions of any sort; there are no divisibility problems with respect to the scale of
transactions; that equal information is available to all market participants. In some cases, actual
markets will be sufficiently "close" to this abstraction to use the resulting analysis directly. In other
cases, it provides a "benchmark" for the study of imperfections.

5
 Robert C. Merton

Summary 53-Year Return Experience: Stocks and Bonds (1926–1978)

Source: “Stocks, Bonds, Bills, and Inflation: Historical Returns (1926–1978),”

R.G. Ibbotson and R.A. Sinquefield, Financial Analysts Foundation (1979).

Average Annual Standard Growth of $1000

Type Return Deviation (Average Compound Return)

Common Stocks (S&P 500) 11.2% 22.2% $89,592

(8.9%)

Long-Term Corporate Bonds 4.1% 5.6% $ 7,807

(4.0%)

Long-Term Government Bonds 3.4% 5.7% $ 5,342

(3.2%)

U.S. Treasury Bills 2.5% 2.2% $ 3,728

(2.5%)

“Inflation-Adjusted” (Consumers Price Index) (“Real”) Returns

Average Annual Standard Growth of $1000

Type Return Deviation (Average Compound Return)

Common Stocks (S&P 500) 8.7% 22.3% $23,399

(6.1%)

Long-Term Corporate Bonds 1.6% NA $ 2,018

(1.3%)

Long-Term Government Bonds 0.9% NA $ 1,377

(0.6%)

U.S. Treasury Bills 0.0% 4.6% $ 965

(0.0%)

6
Finance Theory

7
II. ON THE ARITHMETIC OF COMPOUND INTEREST: THE TIME VALUE OF
MONEY

From our everyday experiences, we all recognize that we would not be indifferent to a
choice between a dollar to be paid to us at some future date (e.g., three years from now) or a
dollar paid to us today. Indeed, all of us would prefer to receive the dollar today. The
assumption implicit in this common-sense choice is that having the use of money for a period of
time, like having the use of an apartment or a car, has value. The earlier receipt of a dollar is
more valuable than a later receipt, and the difference in value between the two is called the time
value of money. This positive time value of money makes the choice among various
intertemporal economic plans dependent not only on the magnitudes of receipts and expenditures
associated with each of the plans but also upon the timing of these inflows and outflows.
Virtually every area in Finance involves the solution of such intertemporal choice problems, and
hence a fundamental understanding of the time value of money is an essential prerequisite to the
study of Finance. It is, therefore, natural to begin with those basic definitions and analytical tools
required to develop this fundamental understanding. The formal analysis, sometimes called the
arithmetic of compound interest, is not difficult, and indeed many of the formulas to be derived
may be quite familiar. However, the assumptions upon which the formulas are based may not be
so familiar. Because these formulas are so fundamental and because their valid application
depends upon the underlying assumptions being satisfied, it is appropriate to derive them in a
careful and axiomatic fashion. Then, armed with these analytical tools, we can proceed in
subsequent sections with the systematic development of finance theory. Although the emphasis
of this section is on developing the formulas, many of the specific problems used to illustrate
their application are of independent substantive importance.
A positive time value of money implies that rents are paid for the use of money. For goods
and services, the most common form of quoting rents is to give a money rental rate which is the
dollar rent per unit time per unit item rented. A typical example would be the rental rate on an
apartment which might be quoted as "$200 per month (per apartment)." However, a rental rate
can be denominated in terms of any commodity or service. For example, the wheat rental rate

8
Finance Theory

would have the form of so many bushels of wheat rent per unit item rented. So the wheat rental
rate on an apartment might be quoted as "125 bushels of wheat per month (per apartment)."
In the special case when the unit of payment is the same as the item rented, the rental rate
is called the own rental rate, and is quoted as a pure percentage per unit time. So, for example, if
the wheat rental rate on wheat were ".01 bushels of wheat per month per bushel of wheat rented,"
then the rental rate would simply be stated as "1 percent per month." In general, the own rental
rate on an item is called that item's interest rate, and therefore, an interest rate always has the
form of a pure percentage per unit time.
Because it is so common to quote rental rates in terms of money, the money rental rate
(being an own rental rate) is called the money interest rate, or simply the interest rate, and the
rents received for the use of money are called interest payments. Moreover, as is well known, to
rent money from an entity is to borrow, and to rent money to an entity is to lend. If one borrows
money, he is a debtor, and if he lends money, he is a creditor.
Throughout this section, we maintain four basic assumptions:

(A.II.1) Certainty: There is no uncertainty about either the magnitude or timing of any
payments. In particular, all financial obligations are paid in the amounts and at
the time promised.

(A.II.2) No Satiation: Individuals always strictly prefer more money to less.

(A.II.3) No Transactions Costs: The interest rate at which an individual can lend in a
given period is equal to the interest rate at which he can borrow in that same
period. I.e., the borrowing and lending rates are equal.

(A.II.4) Price-Taker: The interest rate in a given period is the same for a particular
individual independent of the amount he borrows or lends. I.e., the choices made
by the individual do not affect the interest rate paid or charged.

In addition, we will frequently make the further assumption that the rate of interest in each
period is the same, and when such an assumption is made, that common per period rate will be

9
 Robert C. Merton

denoted by r. Although no specific institutional structure for borrowing or lending is presumed,

the reader may find it helpful to think of the described financial transactions as being between an
individual and a bank. Indeed, for expositional convenience, we will call loans made by
individuals, "deposits."

Compound Interest Formulas

Compound Value

Let V n denote the amount of money an individual would have at the end of n periods if he
initially deposits V o dollars and allows all interest payments earned to be left on deposit (i.e.,
reinvested). V n is called the compound value of V o dollars invested for n periods. Suppose
the interest rate is the same each period. At the end of the first period, the individual would have
the initial amount V o plus the interest earned, rV o , or V 1 = Vo + rV o = (1+ r)V o . If he

redeposits V1 dollars for the second period at rate r, then

2
V 2 = (1 + r) V 1 = (1 + r)[(1 + r) V o ] = (1 + r ) V o . Similarly, at the end of period (t - 1), he will
t
have V t -1 and redeposited, he will have V t = (1 + r) V t -1 = (1 + r ) V o at the end of period t.
Therefore, the compound value is given by
n
(II.1) V n = (1 + r ) V o ,
n
and (1 + r ) is called the compound value of a dollar invested at rate r for n periods.

Problem II.1. "Doubling Your Money": Given that the interest rate is the same each period, how
many periods will it take before the individual doubles his initial deposit? This is the same as
asking how many periods does it take before the compound value equals twice the initial deposit
*
(i.e., V n = 2 V o ). Substituting into (II.1), we have that the number of periods required, n , is
given by

10
Finance Theory

*
(II.2) n = log(2)/ log(1 + r) = .69315/ log(1 + r)

where "log" denotes the natural logarithm (i.e., to the base e). Two "rules of thumb" used to
*
approximate n in (II.2) are:

n ≈ 72/100r
*
(II.3) (" Rule of 72" )

and

n ≈ 0.35 + 69/100r
*
(II.4) (" Rule of 69" )

Of the two, the Rule of 69 is the more precise although the Rule of 72 has the virtue of requiring
*
only one number to remember. Both rules provide reasonable approximations to n . For
*
example, if r equals 6 percent per annum, to one decimal place, the Rule of 72 gives n = 12.0
*
years while the Rule of 69 and the exact solution gives n = 11.9 years. Moreover, in this day of
hand calculators, any more accurate estimates should simply be computed using (II.2). For
further discussion of these rules, see Gould and Weil (1974).

Present Value of a Future Payment

The present value of a payment of $x, n periods from now, PV n (x),

is defined as the smallest number of dollars one would have to deposit today so that with it and
cumulated interest, a payment of $x could be made at the end of period n. It is therefore, equal
to the number of dollars deposited today such that its compound value at the end of period n is
$x. If one can earn at the same rate of interest r per period on all funds (including cumulated
interest) for each of the n periods, then the present value can be computed by setting V n = x in
n n
(II.1), and solving for V o = V n /(1 + r ) = x/(1 + r ) . I.e.,

n
(II.5) PV n (x) = x/(1 + r ) ,

11
 Robert C. Merton
n
and 1/(1 + r ) is the present value of a dollar to be paid n periods from now.
If one were offered a payment of $x, n periods from now, what is the most that he would
pay for this claim on a future payment today? The answer is PV n (x). To see this, suppose that
the cost of the future claim were P > PV n (x). Further, suppose that instead of buying the future
claim, he deposited $P today and reinvested all interest payments for n periods. At the end of
n n
n periods, he would have $P(1 + r ) which by hypothesis is larger than PV n (x)(1 + r ) = $x.
I.e., he would have more money at the end of n periods by simply depositing the money rather
than by purchasing the future claim for P. Therefore, he would be better off not to purchase the
future claim.
If one owned a future claim on a payment of $x, n periods from now, what is the least
amount that he would sell this claim for today? Again, the answer is PV n (x). Suppose that the
price offered for the future claim today were P < PV n (x). If he sells, then he will have $P
today. Suppose that, instead of selling the future claim, he borrows $ PV n (x) today for one
period. At the end of the first period, he will owe PV n (x) plus interest, rPV n (x), for a total of
(1 + r) PV n (x). If he pays off this loan and interest by borrowing $(1 + r) PV n (x) for another

period (i.e., he "refinances" the loan), then at the end of this (the second) period, he will owe
(1 + r) PV n (x) plus interest, r(1 + r) PV n (x) for a total of (1 + r )2 PV n (x). If he continues to

refinance the loans in the same fashion of n periods, then at the end of period n, he will owe
(1 + r )n PV n (x) or $x which he can exactly pay off with the $x payment from the claim he

owns. The net of these transactions is that he will have received $ PV n (x) initially which by
hypothesis is larger than $P. I.e., he would have more money initially by borrowing the money
"against" the future claim rather than by selling the future claim for $P, and therefore he would
be better off not to sell the future claim.
In summary, if the price of the future claim, P, exceeds its present value, PVn(x), then
the individual would prefer to sell the claim rather than hold it (or if he did not own it, he would
not buy it). If the price of the future claim, P, is less than its present value, PV n (x), then the

12
Finance Theory

individual would prefer to hold it rather than sell it (or if he did not own it, he would buy it).
Therefore, at P = PV n (x), the individual would have no preference between buying, holding, or
selling the future claim. Hence, the present value of a future payment is such that the individual
would be indifferent between having that number of dollars today or having a claim on the future
payment.

Present Value of Multiple Future Payments

The present value of a stream of payments with a schedule of $ xt paid at the end of
period t for t = 1,2,..., N is defined as the smallest number of dollars one would have to deposit
today so that with it and cumulated interest, a payment of $ xt could be made at the end of
period t for each period t, t = 1,2,..., N. We denote this present value by PV( x1 , x 2 ,..., x N ).
To derive the formula for its present value, we proceed as follows: Suppose that we establish
today N separate bank accounts where in "Account #t," we deposit PV t ( x t ) dollars,
t = 1,2,..., N. If we let the interest payments accumulate in Account #t until the end of period t,
then the amount of money in the account at that time will equal the compound value of
PV t ( x t ). By the definition of the present value of a single future payment, we will have just
enough money to make a payment of $ xt at the end of period t by liquidating Account #t. If
we follow this procedure for each of the N separate accounts, then we would be able to make
exactly the schedule of payments required. Hence, the present value of the stream of payments
with this schedule is equal to the total amount of deposits required for these N accounts. I.e.,

PV( x1 , x 2 ,..., x N ) = PV 1( x1 ) + PV 2 ( x 2 ) + ... + PV N ( x N )

N
(II.6)
= ∑ PV t ( x t ) .
t =1

So, the present value of a stream of payments is just equal to the sum of the present values of
each of the payments. Hence, if one can earn at the same rate of interest r per period on all

13
 Robert C. Merton

funds (including cumulated interest) for each of the N periods, then from (II.5) and (II.6), we
have that
N
PV( x1 , x 2 ,..., x N ) = ∑ xt /(1 + r ) .
t
(II.7)
t =1

As this derivation demonstrates, a claim on a stream of future payments is formally

equivalent to a set of claims with one claim for each of the future payments. As was shown, an
individual would be indifferent between having $ PV t ( x t ) today or a payment of $ xt at the
end of period t. It, therefore, follows that he would be indifferent between having
$PV( x1 , x 2 ,..., x N ) today or a claim on the stream of future payments with the schedule of $ xt

paid at the end of period t for t = 1,2,..., N .

As may already be apparent, the present value concept is an important tool for the
solution of intertemporal choice problems. For example, suppose that one has a choice between
two claims: the first, call it "claim Y," provides a stream of payments of $ y t at the end of

period t for t = 1,2,..., N, and the second, call it "claim X," provides a stream of payments of
$ xt at the end of period t for t = 1,2,..., N . Which claim would one choose? We have
already seen that one would be indifferent between having a claim on stream of future payments
or having its present value in dollars today. So one would be indifferent between having claim Y
or $PV( y1 , y 2 ,..., y N ) today, and similarly, one would be indifferent between having claim X

or $PV( x1 , x 2 ,..., x N ) today. Hence to make a choice between having $PV( y 1 , y 2 ,..., y N )

today or $PV( x1 , x 2 ,..., x N ) today is formally equivalent to making a choice between claim Y
or claim X. But, as long as one prefers more to less, the former choice is trivial to make:
Namely, one would always prefer the larger of $PV( y 1 , y 2 ,..., y N ) or $PV( x1 , x 2 ,..., x N ) today.

Thus, one would prefer claim Y to claim X if PV( y 1 , y 2 ,..., y N ) > PV( x1 , x 2 ,..., x N ), and

would prefer claim X to claim Y if PV( y 1 , y 2 ,..., y N ) < PV( x1 , x 2 ,..., x N ) . Moreover, if the

two present values are equal, then one would be indifferent between the two claims.

14
Finance Theory

In the formal notation, both claim X and claim Y had the same number of payments:
namely N. However, nowhere was it assumed that some of the xt or y t could not be zero.
Thus, the timing of the payments need not be the same. Moreover, nowhere was it assumed that
some of the xt or y t could not be negative. Since the xt or y t represent cash payments to

the owner of the claim (i.e., a receipt) a negative magnitude for these variables is interpreted as a
cash payment from the owner of the claim (i.e., an expenditure). Indeed, it is entirely possible for
the present value of a stream of payments to be negative which simply means one would be
willing to make an expenditure and pay someone to take the claim. Hence, the present value tool
provides a systematic method for comparing claims whose schedules of payments can differ
substantially both with respect to magnitude and timing. While our illustration applied it to
choosing between two claims, it can obviously be extended to the problem of choosing from
among several claims. Its use in this intertemporal choice problem can be formalized as follows:

Present Value Rule:

If one must choose among several claims, then proceed by: first, computing the present
values of all the claims. Second, rank or order all the claims in terms of their present values from
the highest to the lowest. Third, if one must choose only one claim, then take the first claim (i.e.,
the one with the highest present value). More generally, if one must choose k claims out of a
larger group, then take the first k claims in the ordering (i.e., those claims with the k largest
present values in the group). This procedure for choosing among several claims is called the
Present Value Rule.
Note that if the rate of interest in every period were zero, then the present value of a
N

stream of payments is just equal to the sum of all the payments (i.e., PV( x 1 , x 2 ,..., x N ) = ∑ x t ).
t=1

In this case, the Present Value Rule would simply say "choose that claim which pays one the
most money in total (without regard to when the payments are received)." However, because of

15
 Robert C. Merton

the time value of money, the interest rate will not be zero, and no such simple rule will apply.
That one cannot rank or choose between alternative claims without taking into account the
specific interest rate available is demonstrated by the following problem:

Problem II.2. Choosing Between Claims: Suppose that one has a choice between "claim X"
which pays $100 at the end of each year for ten years or "claim Y" which provides for a single
payment of $900 at the end of the third year. Given that the interest rate will be the same each
year for the next ten years, which one should be chosen? The Present Value Rule says "Choose
the one with the larger present value." However, as the following table demonstrates, the claim
chosen depends upon the interest rate.

Interest Rate, r Present Value of Claim X Present Value of Claim Y

0% $1000 $900
2% 898 848
5% 772 777
8% 671 714
10% 614 676
12% 565 641

While the present values of both claims decline as one moves in the direction of higher interest
rates, the rate of decline in the present value of Claim Y is smaller than the rate of decline for
Claim X. Hence, for interest rates below 5 percent, one should choose Claim X and for rates
above 5 percent, one should choose Claim Y.
The result obtained here that one claim is chosen over the other for some interest rates
and the reverse choice is made for other interest rates often occurs in choice problems and is
called the switching phenomenon. It is called this because an individual would "switch" his
choice if he were faced with a sufficiently different interest rate. Hence, without knowing the
interest rate, the choice between two claims will, in general, be ambiguous. So, in general,
unqualified questions like "which claim is better?" will not be well posed without reference to
the specific environment in which the choice must be made. Note, however, that for a specified

16
Finance Theory

interest rate, the present value of each claim is uniquely determined, and therefore the choice
between them at that interest rate level is always unambiguous.
In Problem II.2, it was stressed that, in general, the solution to the problem of choosing
among alternative claims will depend upon the interest rate at which the individual can borrow or
lend. However, it is equally important to stress that the solution depends only upon that interest
rate. Specifically, given that rate of interest, the solution is not altered by the existence of other
claims that an individual owns (i.e., his endowment). Moreover, the solution does not depend
upon whether he plans to use the payments received for current consumption or to save them for
consumption in the future. That is, the solution does not depend upon the individual's
preferences or tastes for future consumption. While this demonstrated independence of the
solution to either the individual's tastes or endowments has far-ranging implications for the
theory of Finance, further discussion is postponed to Section III where the general intertemporal
choice problem for the individual is systematically examined.

Continuous Compounding

It is not uncommon to see an interest rate quoted as "R% per year, compounded n times
a year." For example, a bank might quote its rate on deposits as "7% per year, compounded
quarterly (i.e., every three months or four times a year)" or "7% per year, compounded monthly
(i.e., every month or twelve times a year)." Provided that funds are left on deposit until the end
of a compounding date, such quotations can be interpreted to mean that n times a year, the
account is credited with cumulated interest earned at the rate, (R/n), per period of (1/n) years.
The "true" annual rate of interest, call it in, when there are n such compoundings per year can
be derived using the compound value formula (II.1). From that formula, one dollar will grow to
$(1 + R/n )n in one year, and therefore,

(II.8) 1 + i n = (1 + R/n )n .

17
 Robert C. Merton

By inspection of (II.8), for a given value of R, more frequent compoundings (i.e., larger n)
result in a larger "true" annual interest rate, in . The limiting case of n → ∞ is called
continuous compounding, and the limit of (II.8) is

(II.9) 1 + i∞ = eR

R
where "e" is a constant equal to 2.7183..., and e is called the exponential factor. The
difference between the true or effective annual rate i∞ and the stated rate R will be larger, the
larger is R although for typical interest rates, this difference will not be large. For example, at a
stated rate of R = 5%, i∞ = 5.13%. However, the cumulative difference in compound value for
higher interest rates and over several years can be significant as is illustrated in the following
table:

Compound Value of $100 at the End of N Years

At 10% At 10% per Year,

N per Year Compounded Continuously

1 $ 110.00 $ 110.52
2 121.00 122.14
5 161.05 164.87
10 259.37 271.83
15 417.72 448.17
20 672.75 738.91
30 1,744.93 2,008.55

One can, of course, invert the original question and ask "What continuously-compounded
rate, r c , will produce a "true" annual interest rate, r?" From (II.9), we have that

(II.10) er c ≡ 1 + r ,

or by taking (natural) logarithms of both sides of (II.10), we can rewrite (II.10) as

18
Finance Theory

(II.11) r c ≡ log(1 + r) .

In the analysis of interest rate problems, it is frequently more convenient to work with the
continuously-compounded rate, r c , rather than the actual rate, r. For example, in Problem II.1,
*
we derived a formula for the number of periods required to double our money, n . Substituting
from (II.11) into (II.2), we have that

*
II.12) n = log(2)/ r c = .69315/ r c .

If, in addition, one approximates the stream of payments from a claim, { x t } , by a

continuous stream of payments, {x(t)}, then the discrete-time formula for the present value of a
stream of payments, (II.7), can be approximated by the integral formula,
N
(II.13) PV(x1 , x2 ,...,xn ) ≈ ∫ x(t)e - r ctdt,
0

and in some cases, the integral expression in (II.13) provides an easier way to compute formula
for the present value than its discrete-time counterpart in (II.7).

Annuity Formulas

A claim which provides for a stream of payments of equal fixed amounts at the end of
each period for a specified number of periods is called an annuity. Suppose that one owned an
annuity claim which pays $y at the end of each year for N years. How much money would one
have at the end of year N if payments are immediately deposited in an account which earns r%

19
 Robert C. Merton

per year (on both cumulated interest and the initial deposit) in each year? Using the compound
value formula, (II.1), we have that:

N -1
year 1's payment will grow to y(1 + r )
N -2
year 2's payment will grow to y(1 + r )
N -3
year 3's payment will grow to y(1 + r )
. . . . . .
. . . . . .
. . . . . .
year (N-1)'s payment will grow to y(1 + r)
year N's payment will grow to y .

Hence, the total amount accumulated, S N , will be the sum of all N terms. I.e., SN =
N N -1

∑ y(1 + r ) = y ∑ (1 + r ) . To further simplify the formula, we make a brief digression to

N -t t

t =1 t =0

develop a mathematical formula. The sum of a geometric progression,

N -1
1 + x + x2 + ... + x N -1 = ∑ xt , is given by the formula
t=0

N -1

∑ x = (x
t N
(II.14) - 1)/(x - 1).
t=0

From (II.14), we also have that

N

(II.14a) ∑ x = x( x
t =1
t N
- 1)/(x - 1).

Applying (II.14) with x = 1 + r to the expression for S N , we can rewrite it as

N
(II.15) S N = y[(1 + r ) - 1]/r.

20
Finance Theory
N
SN is called the compound value of an annuity, and [(1 + r ) - 1]/r is called the annuity
compound value factor.
Maintaining the assumption that the interest rate is the same each year, what is the present
value of an annuity (denoted by AN )? From (II.7), we have that
N N

A N = ∑ y/(1 + r ) = y ∑ 1/(1 + r ) .
t t
From (II.14a), we can rewrite the expression for the
t =1 t =1

present value as
N
(II.16) A N = y[1 - 1/(1 + r ) ]/r

N
and [1 - 1/(1 + r) ] /r is called the annuity present value factor.

Formula (II.16) could have been derived by a different (but equivalent) method. From
(II.15), we know that a N-year annuity paying $y per year is equivalent to a claim which
provides a single payment of $ S N paid at the end of year N. From (II.5), we have that
N
PV N ( S N ) = S N /(1 + r ) . But, the present values of two equivalent streams are the same, and

therefore A N = S N /(1 + r ) . The reader may verify that this is the case by inspection of (II.16).
N

Note that if one has a N-period annuity at time (t=) zero, then this same claim will
become a (N-1) period annuity at time t = 1, and at time t, it will be an (N–t) period annuity.
Hence, the change in the present value of an N-period annuity over one period is equal to
AN -1 - AN , and from (II.16), can be written as

N
(II.17) AN -1 - A N = - y/(1 + r ) .

Inspection of (II.17) shows that the present value of an annuity declines each period until at time
t = N (called its expiration date), its present value is zero. Note further that the rate of decline is
larger the closer the annuity is to its expiration date. However, in the special limiting case of a

21
 Robert C. Merton

perpetual annuity or perpetuity where N = ∞, the present value remains unchanged through
time, and is given by

(II.18) A∞ = y/r.

Problem II.3. Mortgage Payment Calculations: Probably the annuity claim with which
households are most familiar is the mortgage which is a specific form of loan used to finance the
purchase of a house. The terms of a standard or conventional mortgage call for the borrower to
repay the loan with interest by making a series of periodic payments of equal size for a specified
length of time. In effect, the house buyer "issues" to the lender (usually a bank) an annuity claim
in exchange for cash today. Typically, the length of time, the periodicity of the payments, and
the interest rate are quoted by the bank. Given this information, one can then determine the size
of the periodic payments as a function of the amount of money to be borrowed. Suppose the
bank quotes its mortgage terms as follows: the length of the mortgage's life or term is 25 years;
the periodicity of the payments is once a year; and the interest rate charged is 8 percent per year.
If the amount of money to be borrowed is $30,000, then what will be the annual payments
required? To solve this problem, we use formula (II.16). The amount of money received in
return for the annuity, $30,000, equals the present value of the annuity, AN . The number of
payments, N, equals 25, and the annual interest rate, r, equals .08. Thus, the required annual
payments, y, are given by the formula

(II.19) y = rA N /[1 - 1/(1 + r )N ].

The annuity present value factor for r = .08 and N = 25 equals 10.675. Therefore, y =
$30,000/10.675 or approximately $2810 per year.
Although the size of the payments remains the same over the life of the mortgage, the
amount of money actually borrowed (called the principal of the loan) does not. In addition to

22
Finance Theory

covering interest payments, a portion of each year's payment is used to reduce the principal. In
the example above, during the first year of the mortgage, the amount of money borrowed is
$30,000, and therefore, the interest part of the payment is .08 × $30,000 or $2,400. However,
because the total payment made is $2,810, the balance after interest, $410, is used to reduce the
principal. Hence, for the second year in the life of the mortgage, the amount actually borrowed is
not $30,000, but $29,590. The following table illustrates how the level of payments are
distributed between interest payments and principal reduction over the life of the mortgage.

25-Year 8% Mortgage: Distribution of Payments

Interest Payments Principal Reduction Amount of Loan

Year Total Payment Amount % of Total Amount % of Total Outstanding

1 $2,810 $2,400 85.4% $ 410 14.6% $29,590

2 2,810 2,367 84.2 443 15.8 29,147
5 2,810 2,252 80.1 558 19.9 27,589
10 2,810 1,990 70.8 820 29.2 24,052
15 2,810 1,605 57.1 1,205 42.9 18,855
20 2,810 1,039 37.0 1,771 63.0 11,220
25 2,810 208 7.4 2,602 92.6 0

Note that early in the life of the mortgage, almost all of the total payment goes for interest
payments. However, by the seventeenth year, the distribution of the payment is approximately
half interest payment and half principal reduction, and as the mortgage approaches its expiration
date, virtually all the payment goes for the reduction of principal.
The general case for the distribution of the payments between interest and principal
reduction can be solved by using formulas (II.16) and (II.17). Because the amount of the
mortgage outstanding always equals its present value, the principal at time t, AN -t , is given by
N -t
A N - t = y[1 - 1/(1 + r ) ]/r. We can rewrite this expression in terms of the initial size of the
mortgage, AN , as

23
 Robert C. Merton

N t N
(II.20) A N - t = A N [(1 + r ) - (1 + r ) ]/[(1+ r ) - 1].

Moreover, the change in principal between t and t + 1 is equal to AN -t -1 - AN -t which from

(II.17) can be written as
N -t
(II.21) AN -t -1 - AN -t = - y/(1 + r ) ,

and the percentage of the total payment used to reduce principal between t and t + 1 can be
written as

(II.22) [ AN -t - AN -t -1 ]/y = 1/(1 + r )N -t .

Problem II.4. Saving for Retirement: A bank recently advertised that if one would deposit $100
a month for twelve years, then at that time, the bank would pay the depositor $100 a month
forever. This is an example of a regular saving plan designed to produce a perpetual stream of
income later, and frequently arises in analyses of retirement plans. For example, how many years
in advance of retirement should one begin to save $X a year so that at retirement, one would
receive $C a year forever?
If it is assumed that the annual rate of interest is the same in each year and if one starts
T
saving T years prior to retirement, then from formula (II.15), a total of $X[(1 + r ) - 1]/r will
have been accumulated by the retirement date. From formula (II.18), it will take $C/r at that
time to purchase a perpetual annuity of $C per year. Hence, the required number of years of
saving is derived by equating the accumulated sum to the cost of the annuity. By taking the
logarithms of both sides and rearranging terms, we have that

(II.23) T = log [1 + C/X]/ log [1 + r],

or alternatively, using (II.11), we can rewrite (II.23) in terms of the equivalent continuously-
compounded interest rate as

24
Finance Theory

(II.24) T = log [1 + C/X]/ r c .

Note that for a fixed ratio of C/X, the length of time required is inversely proportional to the
(continuously-compounded) interest rate. So, if that rate is doubled, then the required saving
period is halved. In the special case where C = X, (II.24) reduces to

(II.25) T = 0.69315/ r c

where 0.69315 ≈ log(2). Comparing (II.25) with (II.2), the number of years of required saving is
exactly equal to the number of years it takes to "double your money," and therefore a "quick"
solution for T can be obtained by using either the Rule of 72 or the Rule of 69. Applying (II.25)
to the bank advertisement, we can derive the monthly interest rate implied by the bank to be 0.48
percent per month or 5.93 percent per year.

Problem II.5. The Choice Between a Lump-Sum Payment or an Annuity at Retirement: Having
participated in a pension plan, it is not uncommon for the individual to be offered the choice at
retirement between a single, lump-sum payment or a lifetime annuity. Suppose one is offered a
choice between a single payment of $x or an annuity of $y per year for the rest of his life.
Given that the interest rate at which he can invest for the rest of his life is r, which should he
choose? Provided that y > rx, the proper choice depends upon the number of years that the
individual will live. Clearly, if he expects to live long enough, then he should choose the
annuity. Otherwise, he should take the lump-sum payment. We can determine the "switch point"
*
in terms of life expectancy by solving for the number of years, N , such that the present value
of the annuity is just equal to the lump-sum payment x. Substituting x for AN in (II.16) and
rearranging terms, we have that

*
(II.25) N = log[y/(y - rx)]/ log[1 + r].

25
 Robert C. Merton

*
Hence, if he expects to live longer than N years, then he should choose the annuity.

Problem II.6. Tax-Deferred Saving for Retirement: Under certain provisions of the tax code,
individuals are permitted to establish tax-deferred savings plans for retirement (e.g., Individual
Retirement Accounts or Keogh Plans). Contributions to these plans are deductible from current
income for tax purposes and interest on these contributions is not taxed when earned. These
plans are called "tax-deferred" rather that "tax-free" because any amounts withdrawn from the
plan are taxed at that time. Suppose that an individual faces a proportional tax rate of τ which is
the same each period and that the interest rate r is the same each period. Further suppose that
he contributes $y each year to the plan until he retires N years from now at which time he
begins a withdrawal program on an annuity basis for n years. Assuming that his first
contribution to the plan takes place one year from now, what is the economic benefit of the tax-
deferred saving plan over an ordinary saving plan?
Using formula (II.15), his total before-tax amount accumulated at retirement,
N
S N , is $y[(1+ r ) - 1]/r. From formula (II.16), he can generate a withdrawal plan of

$q = rS N /[1- 1/(1+ r )n ] per year for n years from this accumulated sum. However, he must

pay taxes of $τq each year on the withdrawals. Hence, the tax-deferred plan will produce an
after-tax stream of payments for n years beginning at retirement of

(II.26) $ q1 = (1 - τ )y[(1 + r )N - 1]/[1 - 1/(1 + r )n ].

If, instead, he had chosen an ordinary saving plan, he would have had to pay $τy additional
taxes each year during the accumulation period because contributions to an ordinary saving plan
are not deductible. So, without changing his expenditures on other items during the
accumulation period, he could only contribute $(1 - τ )y each year. Moreover, the interest
earned in an ordinary saving plan is taxable at the time it is earned. Therefore, instead of earning

26
Finance Theory

at the rate r each year on invested money, he only receives rate (1 - τ )r after tax. Again using
formula (II.15), his total amount accumulated at retirement from the ordinary saving plan, S 2 ,

is $(1 - τ )y[(1 + (1 - τ )r ) - 1]/(1 - τ )r.

N
Because he has paid the taxes on contributions and
interest along the way, the $ S 2 accumulated is not subject to further tax. However, any interest
earned on invested money during the subsequent withdrawal period is taxed at rate τ. Thus,
from formula (II.16), he can generate an after-tax withdrawal plan of
$ q 2 = (1 - τ ) rS 2 /[1 - 1/(1 + (1 - τ )r )n ] per year for n years which can be rewritten as

(II.27) $ q 2 = (1 - τ )y[(1 + (1 - τ )r )N - 1]/[1 - 1/(1 + (1 - τ )r )n ].

Clearly, the tax-deferred plan provides a positive benefit because q1 > q2. Inspection of
(II.26) and (II.27) shows that this differential can be expressed in terms of a higher effective
interest rate on accumulations in the tax-deferred plan. Specifically, the tax-deferred plan is
formally equivalent to having an ordinary saving plan where the interest earned is not taxed.

Problem II.7. The Choice Between Buying or Renting a Consumer Durable: For most large
consumer durables (e.g., a house or car), the individual can either choose to buy the good or rent
it. Suppose an individual faces the decision of whether to buy a house for $I or rent it where the
annual rental charge is $X per year. If he buys the house, then he must spend $M for
maintenance and $PT for property taxes each year. These are both included in the rent.
Suppose that the individual faces a proportional tax rate of τ which is the same each period and
that the interest rate r is the same each period. His problem is to choose the method of
obtaining housing services with the lowest (present value of) cost.
The present value of cost equals the discounted value of the after-tax outflows discounted
at the after-tax rate of interest, (1 - τ )r. Because property taxes can be deducted from income

27
 Robert C. Merton

for federal income tax purposes, the after-tax outflow for property taxes each year is (1 - τ )PT.
Hence, the cost of owning the house, PCO, can be written as
∞
PCO = I + ∑ [M + (1 - τ )PT]/(1 + (1 - τ )r )
t

(II.28) t =1

= I + PT/r + M/(1 - τ )r

where we have assumed that the (properly-maintained) house continues in perpetuity and applied
the annuity formula. Similarly, the cost of renting the house, PCR, can be written as
∞
PCR = ∑ X/(1 + (1 - τ )r )t
(II.29) t=1

= X/(1 - τ )r.

Hence, if PCR > PCO, then it is better to own rather than rent. Of course, the relationship
between PCR and PCO depends upon the rent charged. In a competitive market, the rent
charged should be such that the landlord earns a return competitive with alternative investments.
Hence, X should be such that the present value of the after-tax cash flows to the landlord equals
the cost of his investment I. The pretax net cash flow to the landlord each year is (X-M-PT). In
computing his tax liability, the landlord can deduct depreciation, D, a non-cash item. Hence,

his taxes are (X-M-PT-D)τ where τ is his proportional tax rate. Therefore, his after

tax cash flow is (X-M-PT)(1 -τ ) + τ D . Discounting these after-tax cash flows at his after-

tax interest rate, (1 - τ )r, we have that X must satisfy

I = [(X-M-PT)(I - τ )+τ D]/(I - τ )r or

(II.30) X = rI + M + PT - τ D/(1 - τ ).

From (II.28), (II.29), and (II.30), we have that the cost saving of owning over renting can be
written as

28
Finance Theory

(II.31) PCR - PCO = τ [I + PT/r]/(1 - τ ) - τ D/[(1 - τ )(1 - τ )r].

The advantage to ownership is that one is not taxed on the rent paid to oneself. The disadvantage
is that one cannot take a tax deduction for the (non-cash) depreciation item. So if the
depreciation rate on the property is high or the individual is in a low tax bracket, then renting is
less costly. On the other hand, if property taxes are high and the individual is in a high tax
bracket, then owning is probably less costly.

"Pure" Discount Loan

A pure discount loan calls for the borrower to repay the loan with interest by making a
single lump-sum payment to the lender at a specified future date called the maturity or expiration
date. Hence, unlike an annuity-type loan, there are no interim payments made to the lender. This
form of loan is most common for short maturity loans, and the best known examples are U.S.
Treasury Bills and corporate commercial paper. If it is assumed that the interest rate is the same
each period, then the present value of a discount loan (denoted by D N ) which has a promised
payment of $M to be paid N periods from now can be written as
N
(II.32) D N = M/(1 + r ) .

If one has a N-period discount loan at time (t=) zero, then this same loan will become a (N – 1)
period discount loan at time t = 1, and at time t, it will be a (N - t) period discount loan.
Hence, the change in the present value of a N-period discount loan over one period is equal to
D N -1 - D N , and from (II.32), can be written as

N
D N -1 - D N = rM/(1+ r )
(II.33)
= rD N .

29
 Robert C. Merton

Inspection of (II.33) shows that unlike an annuity, the present value of a discount loan increases
each period until at t = N, its present value is M. Hence, the amount of money actually
borrowed increases over the life of the loan. The rate of increase each period is the same and
equal to the interest rate r.

"Interest-Only" Loans

Another common form for a loan is an "interest-only" loan which calls for the borrower to
make a series of periodic payments equal in amount to the interest payments for a specified
length of time and, in addition, at the end of that length of time, to make a single payment equal
to the initial amount borrowed (i.e., the principal). The periodic payments are called coupon
payments, and the single, lump-sum (or "balloon") payment at the end is called the return of
principal or simply the principal payment. This form of loan is most common for long maturity
loans, and the best known examples are U.S. Treasury Notes and corporate bonds.

The structure of "interest-only" loans is a mixture of the annuity and pure discount forms
of loans. With the exception of the principal payment, the payment patterns are like those of an
annuity because the size of the coupon payments are all the same. Like a discount loan, there is a
lump-sum payment at the maturity date. However, unlike both the annuity and discount loans,
the amount of the loan outstanding or the principal remains the same throughout the term of the
loan. If it is assumed that the interest rate is the same each period, then the present value of an
interest-only loan (denoted by I N ) which has a coupon payment of $C per period and a
balloon payment of $M can be written as
N

I N = ∑ C/(1 + r ) + M/(1 + r )
t N

(II.34) t=1

= C[1 - 1/(1 + r )N ]/r + M/(1 + r )N .

30
Finance Theory

If the initial amount borrowed is $M and the coupon is set equal to the interest on the amount
borrowed (i.e., C = rM), then substituting into (II.34), we have that

(II.35) IN=M

independent of N. Hence, the present value of the loan remains the same over the life of the
loan.

Compound and Present Values When the Interest Rate Changes Over Time

To this point, all the formulas were derived using the assumption that the interest rate at
which the individual can borrow or lend is the same in each period. We now consider the general
case where the interest can vary, and we denote by r t the one-period rate of interest which will
obtain for the period beginning at time (t – 1) and ending at time t. If, as before, V n denotes
the compound value of Vo dollars invested for n periods, then
V 1 = (1 + r 1 ) V o ; V 2 = (1 + r 2 ) V 1 = (1 + r 2 )(1 + r 1 ) V o ; and
V t = (1 + r t )V t -1 = (1 + r t )(1 + r t -1 )(1 + r t - 2 )...(1 + r 1 )V o . Hence, the analogous formula to (II.1)
for the compound value is

⎡ n ⎤
(II.36) V n ⎢Π
= ( 1+ r )
t ⎥V o
⎣ t=1 ⎦

where "Π" is a shorthand notation for the "product of." I.e.,

n

∏ (1+ rt ) ≡ ( 1+ r1 )( 1+ r2 ) ... ( 1+ rn−1 )(1+ rn ) . For notational simplicity, we define

t=1

the number R n as that rate such that compounding at that (equal) rate each period for n periods

31
 Robert C. Merton

will give the same compound value as compounding at the actual (and different) one-period
rates. That is,

n
(1+ R n ) ≡ Π (1+ r t ),
n
(II.37)
t=1

and therefore, 1 + R n is the geometric average of the {1 + r t }, t = 1,2,..., n. Hence, we can

rewrite (II.36) as
n
(II.38) V n = (1 + R n ) V o .

From (II.38) and the definition of present value, the present value of a payment of $x, n
periods from now, can be written as
n
(II.39) PV n (x) = x/(1 + R n ) ,

and the present value of a stream of payments with a schedule of $ xt paid at the end of period
t, t = 1,2,...,N, can be written as

N
PV( x1 , x 2 ,..., x N ) = ∑ PV t ( xt )
t =1
(II.40) N
= ∑ x /(1+R t t
t
).
t =1

Using the formalism of R n , the compound and present value formulas when interest rates vary
look essentially the same as in the constant interest rate case. However, care should be exercised
to ensure that one does not confuse the " R n" with the " r n " . The former depends upon the
entire path of interest rates from time t = 1 to time t = n while the latter is simply the one-
period rate that obtains between t = n – 1 and t = n. For example, from (II.37), we have that

32
Finance Theory

> >
(II.41) R n = R n -1 if and only if r n = R n -1 .
< <

Hence, R n = r n if and only if R n = R n-1 . Moreover, r n > R n-1 does not imply that r n > r n -1 .
Further discussion of the relationship between the { R t } and { r t } is postponed until Section V
where they will be placed in substantive context.
This completes the formal preparation on the time value of money, and, as promised, we
now turn to the systematic development of finance theory.

33
III. ON THE THEORY OF ACCUMULATION AND INTERTEMPORAL
CONSUMPTION CHOICE BY HOUSEHOLDS IN AN ENVIRONMENT OF
CERTAINTY

Begin the study of Finance with the analysis of an economy where all future outcomes are
known with certainty, but households receive income (their endowments) and consume at
different points in time. In particular, it is shown how the consumption-saving decision is made
and why the introduction of a capital market and financial securities can improve consumer
welfare.
As was discussed in the Introduction, the major decisions of the financial manager are to
choose which (physical) investments to make and to choose the appropriate means for financing
them. It is assumed that the "correct" policies chosen will be those that maximize some criterion
function (or performance index) specified by the firm. We prepare for the study of corporate
finance by deducing here and in Section IV a rational criterion function for the firm and the
management rules which optimize this criterion function in the simplified world of perfect
markets and certainty. Despite the simplicity of the model relative to the "real" world, the results
derived from this model form a basis for the rationalization of the more complex decision rules
developed later. Hence, while the manifest functions of the analysis are to show how
intertemporal allocations are made and to show what role capital markets play in these
allocations, an important latent function of the analysis is to provide a foundation for corporate
financial theory.
We begin the analysis by solving the two-period problem and then extend it in a natural
fashion to the general case of many periods.

Consumer Behavior: The Two-Period Case

The four assumptions of Section II (A.II.1) - (A.II.4), are maintained throughout the
analysis. It is further assumed that each consumer has a well-behaved utility function expressing
his preferences between current consumption, C0 , and next period's consumption, C1.
Because the emphasis is on the intertemporal allocation of consumption, it is assumed that there

34
Finance Theory

is a single consumption good in each period. The consumer's utility function is denoted by
U[C0,C1]. Because both period's consumptions are considered goods (in contrast to "bads"), it is
assumed that U1[C0,C1] ≡ ∂U[C0,C1]/ ∂C0 > 0 and U2[C0,C1] ≡ ∂U[C0,C1]/∂C1 > 0. By
assuming the strict inequality, we rule out the possibility of satiation. I.e., consumers will always
strictly prefer more to less of either C0 or C1. We also assume sufficient regularity and
concavity of U to ensure existence of unique interior maximums.
An indifference curve is the set of all combinations of current and next-period
consumption, (C0,C1), such that the consumer is indifferent among these alternative
combinations i.e., they are curves of equal utility or iso-utility curves. Formally, it is the
functional relationship between C0 and C1 such that U[C0,C1] = U , where U is a constant.

Figure 1 illustrates the general shape of the indifference curves, and as they are drawn,
1 2 3
U >U >U . Analytically, by the Implicit Function Theorem or heuristically, by using

differentials, we have that d U = 0 = U 1 d C 0 + U 2 d C 1 , or that

⎛ dC 1 ⎞
(III.1) ⎜⎜ ⎟⎟ = - U 1 [ C 0 ,C 1 ]/ U 2 [ C 0 ,C 1 ] < 0,
⎝ dC 0 ⎠U =U
_
where (dC1/dC0) is the slope of the indifference curve defined by U[C0,C1] = U at the point
(C0,C1). As shown in Figure 1, this slope is always strictly negative.

Case 1. The Simplest Capital Market: Pure Exchange

For this case, we assume that there are no means of physical production. I.e., there is no
way of using the current period's goods to produce additional goods next period. However,
suppose there does exist a market for trading current period's goods in return for a claim on
goods next period. So, an individual can go to the market and exchange current period goods for
"pieces of paper" which, in turn, can be exchanged next period for goods. Alternatively, he can

35
 Robert C. Merton

receive current period goods by issuing "pieces of paper" which he must redeem for goods next
period. In effect, in the former case, he is lending and in the latter, he is borrowing.
If, by convention, the price per unit of current period goods is set equal to one (i.e., a unit
of current period goods is numeraire), then the (current) price per unit of next period goods, P,
is the rate of exchange for claims on next period goods in terms of current period goods. So, P
units of current period goods can buy a claim on one unit of next period goods. In an
intertemporal context, this price is also written as P ≡ 1/(1+ r) where r is the rate of interest.
Hence, one unit of current goods can be exchanged for (1 + r) units of goods delivered next
period.

Figure III.1
Indifference Curves

36
 A consumer's endowment of exogenous income is denoted by (y0,y1) where y0 is the
number of units of current goods he owns and y1 is the number of units of goods that he will
receive next period. The consumer's current wealth, W0, is equal to the value of his endowment
i.e., W0 = y0 + Py1. The consumer's feasible consumption set is the set of all combinations
(C0,C1) which he can afford to buy. Thus, if (C0,C1) are in the consumer's feasible
consumption set, then the cost of that consumption program, C0 + PC1, can be no larger than his
wealth W0. Moreover, as long as a consumer prefers more consumption to less, he would never
choose a program which costs less than his wealth. Hence, if it is assumed that the consumer
will choose the most preferred feasible consumption program, then he will act so as to maximize
U[C0,C1] subject to his budget constraint that W0 = C0 + PC1.
Substituting for C0 in U from the budget constraint, we can write the consumer choice
problem as

(III.2) Max U[ W
C1
0 - PC 1 , C 1 ]

which leads to the first-order condition for an interior maximum

dU
(III.3) = 0 = - U 1 [ W 0 - PC*1 ,C*1 ]P + U 2 [ W 0 - PC*1 ,C*1 ] ,
dC 1

* * * *
where (C0 ,C1 ) is the optimal consumption program. Noting that C0 = W0 - PC1 , we can
rewrite (III.3) as

(III.4) ( dC 1 / dC 0 )U = U * = - 1/P = - (1 + r) ,

* * *
where U ≡ U [C0 ,C1 ] is the maximum feasible value of utility. Hence, the optimum occurs at
the point where an indifference curve is tangent to the budget constraint as shown in Figure 2.
Note that in arriving at the optimality condition (III.3), we have used assumption (A.II.4)
that the consumer acts as a pure competitor or price-taker. So, in solving for his most preferred
consumption program, the consumer treats the price (or interest rate) as a given number which
does not change in response to the different consumption choices that he might make.

37
 Robert C. Merton

Figure III.2

In the absence of an exchange market and without physical storage of goods through time,
the optimal consumption program for the consumer will simply be to consume current income.

Co = yo and C1 = y1 .
*
I.e., Hence, if the solution to (III.3) yields C0 ≠ y0 (and
*
therefore, C1 ≠ y1 ), then the consumer will be better off as a result of the creation of an

exchange market. Moreover, he can be no worse off because he always has the option not to use
the market and choose C0 = y0 and C1 = y1 which is called the autarky point.
Even if physical storage of goods is feasible, then in the absence of an exchange market,
the feasible consumption choices are constrained to have C0 ≤ y0. That is, physical storage

38
Finance Theory

allows one to "move" goods "forward" in time for consumption, but it does not allow one to
"move" goods "backward" in time.
So, for example, suppose that one had an income stream of (y0 =) ten bushels of wheat
this period and (y1 =) fifty bushels of wheat next period. In the absence of an exchange market,
there is no way that he can consume more than ten bushels of wheat this period even if costless
storage of wheat were available. However, in the presence of an exchange market, in addition to
the ten bushels he has, he could consume up to 50/(1+r) bushels of wheat in the initial period
where r is the market interest rate. Even if his endowment had been y0 = 50 and y1 = 10, then
he would still be better off to save wheat for next period through the exchange market rather than
by storage provided that the interest rate is positive.

Problem III.1: Choosing an Optimal Consumption Allocation: Suppose that one has a
preference function given by U[C0,C1] = log(C0) + log(C1)/(1+δ) and an endowment of y0 = y1
= y. If r is the market rate of interest, then what is the optimal allocation
* * * * * * * *
(C0 , C1 )? From (III.3), we have that U1[C0 ,C1 ]/U2[C0 ,C1 ] = (1+δ)C1 /C0 = 1+r, or that

C1 = (1+r)C0 /(1+δ ). W0 = y + Py = (2+r)y/(1+r). From

* *
the budget constraint,

C 0 = W0 - PC1 = ⎡
⎣( 2+r ) y − C1 ⎤⎦ (1 + r ) . Substituting
* * *
into the budget constraint for

*
C1 from the optimality condition, we have that

C 0 = { (1 + δ )(2 + r)/[(1 + r)(2 + δ )] } y

*
(III.5a)

and

C 1 = (2 + r)y/(2 + δ ) .
*
(III.5b)

39
 Robert C. Merton

Time Preference
A consumer is said to have a positive time preference if for every (a,b) such that b > a,
U[b,a] > U[a,b]. He has no time preference if U[b,a] = U[a,b], and a negative time preference
if U[b,a] < U[a,b].
In the example of preferences used in Problem III.1, δ can be interpreted as the
consumer's rate of time preference. If δ > 0, then he has positive time preference. If δ = 0,
then he has no time preference, and if δ < 0, then he has negative time preference. Note that in
*
that example, if the interest rate exceeds his rate of time preference (r > δ), then C0 < y , and

he will save some of his current period's income to consume next period. If r < δ, then
*
C0 > y and he borrows against next period's income to consume more than his current income.
*
If r = δ, then C0 = y, and he does not trade, but consumes exactly his income in each period.
Suppose that the consumer in this example were the only person in the economy (i.e., a
"Robinson Crusoe" economy). Because he can only trade with himself, the autarky solution is
the only feasible solution. However, we can compute the "equilibrium" rate of interest consistent
with autarky and that rate clearly must be r = δ. Hence, by this example, we have illustrated one
of the possible explanations for a positive rate of interest: namely, consumers' impatience to
consume or a positive time preference.

Case 2. A No-Exchange Market Economy: Pure Production

As in the first case, we assume that the consumer has an endowment of exogenous income
(y0,y1), but in addition, he has the opportunity to use some of his current income to produce
next-period goods. One may wish to think of the "good" as seed which can either be eaten
(consumed) or planted (invested). However, because there is no exchange market, physical
production is the only means he has to increase his next period's consumption beyond next
period's income. Moreover, because there is no exchange market, the only way that he can

40
Finance Theory

produce is by forgoing some current consumption i.e., if X0 denotes the amount he invests in
production, then

(III.6) X 0 = y 0 - C 0 > 0.

The technology available to him is described by a production function f, such that X0

units of current goods invested will produce X1 = f(X0) units of the good next period. It is
assumed that f(0) = 0 and df/dX0 ≡ f′ (X0) > 0. It is further assumed that the production

technology exhibits non-increasing returns to scale (i.e., d f / dX 0 ≤ 0 ) . Figure 3 illustrates

2 2

the production function for decreasing returns to scale, and for 0 ≤ X0 ≤ y0, describes his
max
Production Possibility Frontier. The maximum output that he can produce is X1 = f(y0)

which corresponds to X0 = y0 and C0 = 0. Hence, f(y0) ≥ X1 ≥ 0. His next period's

consumption can be written as

C 1 = y1 + X 1
(III.7)
= y1 + f( y 0 - C 0 )

41
 Robert C. Merton

Figure III.3
Production Function

which for y0 - C0 ≥ 0, describes his feasible consumption set of Consumption Possibility

Frontier. Because there is no exchange market and therefore, no prices, the consumer does not
have a budget constraint of the type in Case 1. However, his consumption choices are
constrained by (III.7) which is called a technological budget constraint. Hence, if, as in Case 1,
it is assumed that the consumer will choose the most-preferred feasible consumption program,
then he will act so as to maximize U[C0,C1] subject to his technological budget constraint.
Substituting for C1 in U from (III.7), we can write the consumer choice problem as

Max U[ C
C0
0 , y 1 + f( y 0 - C 0 )]

which leads to the first-order condition for an interior maximum

42
Finance Theory

(III.8) 0 = U 1 [ C*0 , y 1 + f( y 0 - C*0 )] - f ′( y 0 - C*0 )U 2 [ C*0 , y 1 + f( y 0 - C*0 )].

Assuming that the optimum is interior, we can rewrite (III.8) as

U 1 [ C 0 ,C 1 ]/ U 2 [ C 0 ,C 1 ] = f ′( X 0 )
* * * * *
(III.9)

* * *
where C1 = y1 + f(X 0 ) and the optimal amount to plant, X 0 is given by
* *
X 0 = y 0 - C0 . As was done in Case 1, we have from (III.1) that (III.9) can be rewritten in
* *
terms of the slope of an indifference curve through (C0 ,C1 ) as

(III.10) ( dC 1 / dC 0 )U =U * = - f ′( X *0 )

* * *
where U = U[C0 ,C1 ]. Figure 4 plots the Consumption Possibility Frontier along with a
* * *
graphical solution of the optimal consumption-production program (C0 ,C1 ,X0 ). Because

there is no exchange market, he cannot "borrow" against next period's income, y1, to consume

more in the current period. (i.e., C1 ≥ y1 and C0 ≤ y0). Hence, as shown in Figure 4, the

Consumption Possibility Frontier has a vertical portion for C1 ≤ y1.

43
 Robert C. Merton

Figure III.4

Although there is no market rate of interest in Case 2, we can define an "implied" or

"technological" rate of interest, r , by 1 + r ≡ f (′ X 0 ). By comparing (III.10) with (III.4), we

*

see that r serves as a surrogate for the market rate r, and hence illustrates a second reason for
a positive rate of interest: namely, the productivity of (physical) investment.

Case 3. Production Within an Exchange Market Economy

We maintain the same assumptions about the consumer's endowment of exogenous income
and a production technology as in Case 2. However, we now allow for an exchange market as in
Case 1 where the current market price of next period's goods is P = 1/(1+r). In this environment
his current wealth, W0, can be written as

44
Finance Theory

(III.11) W 0 = y0 + Py1 + Pf( X 0 ) - X 0

where the first two terms on the right-hand side represent the current value of his endowment of
exogenous income and the last two terms represent the net current value of operating his
production technology with an input intensity of X0. That is, if he buys inputs today with a
current value of X0, then he will receive an output next period of f(X0) which has a current
value of Pf(X0). The difference between the two is the net increment to his current wealth from
operating the technology at that intensity. Note that unlike in Case 1, the consumer's current
wealth is affected by one of his decisions: namely, the amount of physical production he
undertakes, X0.
As in Cases 1 and 2, the consumer chooses an investment-consumption program,
(X0,C0,C1), so as to maximize U[C0,C1] subject to the budget constraint that W0 = C0 + PC1.
Because there now exists an exchange market, (III.6) in Case 2 is no longer a constraint i.e., the
consumer can borrow against future income to either consume or invest in physical production in
the current period. Substituting for C0 from the budget constraint, we can write the consumer
choice problem as

Max U[ y
{ X 0 ,C 1 }
0 + Py1 + Pf( X 0 ) - X 0 - PC 1 , C 1 ]

which leads to the set of first-order conditions for an interior maximum

(III.12a) ∂U/ ∂ X 1 = 0 = U 1[ C *0 ,C*1 ](Pf '( X *0 ) - 1)

and

(III.12b) ∂U/∂ C 1 = 0 = - PU 1 [ C*0 , C*1 ] + U 2 [ C*0 , C*1 ] ,

where
* * *
(X0 ,C0 ,C1 ) denotes the quantities chosen for the optimal investment-consumption program, and

45
 Robert C. Merton

* * * *
C0 = y 0 + Py1 + Pf(X 0 ) - X 0 - PC1 . Because the consumer is assumed never to be
* *
satiated, U1[C0 ,C1 ] > 0, and we can rewrite (III.12a) as

f (′ X *0 ) = 1/P
(III.13a)
= 1+ r .

By inspection of (III.13a), we see that, unlike in (III.9) of the Robinson Crusoe Case 2, the
*
optimal amount to invest in physical production, X 0 , does not depend either upon the
consumer's preferences, U, or his endowment, (y0,y1). Hence, two consumers with quite
different preferences between current and future consumption and with quite different
endowments, but who face the same market rate of interest and have the same production
technologies, will choose the same level of physical investment in their technologies,
*
X 0 . Such a result about physical production is called an efficiency condition because it is
independent of either preferences or endowments, and hence independent of who owns the
production technology.
One interpretation of the optimality condition (III.13a) can be derived as follows: as
previously noted, the current wealth of the consumer is affected by the choice of production
**
intensity. I.e., W0 can be written as W0(X0). If X0 denotes that amount of physical

investment which maximizes the current wealth of the consumer, then from (III.11),
**
X0 is the solution to the problem:

Max [ y
{ X0}
0 + P y 1 + Pf( X 0 ) - X 0 ]

which leads to the first-order condition for an interior maximum

**
(III.14) dW0 /dX 0 = 0 = Pf '( X 0 ) - 1

46
Finance Theory

** *
which is identical to (III.13a). i.e., X 0 = X 0 . Hence, optimality condition (III.13a) can be

interpreted as saying "Choose physical investment so as to maximize one's current wealth." This
is called the Value Maximization Rule and it has significant implications for the theory of
Finance. However, discussion of these implications is postponed until Section IV.
Consider now the second optimality condition (III.12b). From (III.1), it can be rewritten in
* *
terms of the slope of an indifference curve through the point (C0 , C1 ) as

(III.13b) ( dC 1 / dC 0 )U =U * = - 1/P = - (1 + r) .

Comparing (III.13b) with (III.4), we find that it is identical to the optimality condition in the Pure
* * *
Exchange Case 1 if we use as current wealth, W0 ≡ y 0 + Py1 + Pf(X 0 ) - X 0 . Hence, one

can describe the solution of the optimal investment-consumption program for the consumer as
taking place in two steps. First, choose physical investment so as to maximize current wealth.
Second, as in the case of pure exchange, use the exchange market to borrow or lend (against this
maximized wealth) so as to achieve the most-preferred, feasible consumption allocation. Figure
5 provides a graphical solution of the problem, and is, in essence, a composite of Figures 3 and 4.
As inspection of Figure 5 clearly demonstrates, the consumer is better off in the presence of an
exchange market than he was in the Robinson Crusoe framework of Case 2.
Hence, the existence of an exchange or capital market will not only affect the patterns of
consumption chosen but also will alter the allocation of physical investment among the various
technologies, and in so doing affect the total output for the economy.

47
 Robert C. Merton

Figure III.5
Note: By trading, he reaches a higher indifference curve.

The Multi-Period Consumption and Allocation Decision: The T-Period Case

We now extend the previous analysis to a consumer who lives for T-periods with a utility
function for lifetime consumption described by U[C0,C1,...,CT-1,CT] where Ct is his
consumption in period t, t = 0,...,T. Let yt denote the exogenous income he will receive in
period t, t = 0,...,T. There exists an exchange market which is open each period and allows for
trading the current period's consumption good and claims on consumption goods in the future.
th
Specifically, at each point in time, there are (T+1) different claims traded in units where the τ

48
Finance Theory

such claim gives its owner the right to one unit of the consumption good payable τ periods from
the date at which it is issued, τ = 0,...,T. In effect, these claims are pure discount loans as
defined in Section II. Let Pt(τ) denote the price at date t of a discount loan which pays one unit
of the consumption good τ periods from date t (i.e., at date t + τ). If, by convention, the
current period's (or "spot") price of the consumption good is taken as numeraire', then Pt(0) = 1,
for all t.
In the absence of any production capabilities, the consumer's current wealth, W0, at date t
= 0 can be written as

W 0 = ∑ P0 ( τ
T

(III.15) ) yτ .
τ =0

As in the two-period analysis, the consumer's feasible consumption set is the set of all
consumption programs that he can afford to buy. Hence, for a consumption program to be
T

feasible, it must satisfy ∑ P ( τ )C ≤ W

0 τ 0 which defines the feasible consumption set.
τ =0

Provided that satiation is ruled out, the T-period consumer allocation problem is formulated as

maximize U[C0,C1,...,CT] subject to the budget constraint that W 0 = ∑ P0 ( τ ) Cτ

T
. Noting
τ =0

that P0(0) = 1, we can substitute for C0 from the budget constraint, and rewrite the problem as

U[ W 0 - ∑ P0 ( τ )C τ ,C 1 ,C 2 ,...,C T ]
T

Max
{ C 1 ,C 2 ,...,C T } τ =1

which leads to T first-order conditions

(III.16) 0 = -U 1 [ C*0 ,C*1 ,...,C*T ] P0 ( τ ) + U τ +1 [ C*0 ,C*1 ,...,C*T ],τ = 1,2,...,T,

th
where Uτ ≡ ∂U[C0,C1,...,CT]/∂Cτ-1 denotes the partial derivative of U with respect to its τ
* * *
argument and (C 0,C 1,...,C T) is the optimal consumption program with

49
 Robert C. Merton

∑ P ( τ )C .
* *
C =W 0 -
0 0 τ In words, (III.16) says that at the optimum, the ratio of the marginal
τ =1

utility of consumption should just equal the ratio of the marginal cost of consumption in period τ
to the marginal utility of current consumption in period τ, P0(τ), to the marginal cost of current
consumption, P0(0) = 1. From (III.16), we have that

* * * * * *
U t+1[ C 0 , C 1 ,...,C T ]/ U s+1[ C 0 ,C 1 ,..., C T ]
(III.17)
= P 0(t)/ P 0(s), s, t = 0,1,...,T .

As with Case 3 of the two-period analysis, we now expand the analysis of the T-period
case to allow for production. Generalizing the production function description of the technology
from the two-period case, let ft(X0t,X1t,...,Xt-1,t) denote the production function for output in
period t, (t=1,2,...,T) where Xjt is the amount of input required to be invested in period j,
(j=0,1,2,...t–1), in order to produce output ft in period t. In an analogous fashion to (III.11) in
the two-period case, we can write the current wealth of the consumer as
T T
W 0 = ∑ P0 ( τ ) yτ + ∑ P0 ( τ ) f τ ( X 0τ , X 1τ ,..., X τ -1,τ )
τ =0 τ =1
(III.18)
- ∑ P0 ( τ
T -1
)Xτ
τ =0

T
where X τ ≡ ∑ X τ j is the total amount of inputs required in period τ to allow production
j=τ +1

plan {ft}, τ = 0,...,T–1. Define the net increment to the consumer's current wealth of production
plan {ft}, V0, by

V 0 ≡ ∑ P0 ( τ ) f τ ( X 0τ , X 1τ ,..., X τ -1,τ ) - ∑ P0 ( τ
T T -1

(III.19) ) Xτ .
τ =1 τ =0

50
Finance Theory

The combined investment-consumption choice problem is formulated as choose the production

and consumption program so as to maximize U[C0,C1,...,CT] subject to the budget constraint
T T

that W 0 = ∑ P0 ( τ ) yτ + V 0 = ∑ P0 ( τ ) Cτ . Substituting for C0 from the budget constraint, the

τ =0 τ =0

problem can be rewritten as choose (C1,...,CT) and (Xjt, j = 0,...,t-1 and t=1,...,T-1) so as to

⎡T ⎤
Max U ⎢∑ P0 ( τ
T
) yτ + V - ∑ P ( τ ) Cτ , C
0 0 1 ,..., C T ⎥
⎣τ =0 τ =1 ⎦
which leads to T(T+1)/2 first-order conditions for the production choices

∂U/ ∂ X jt = 0 = U 1[ C *0 , C *1 ,..., C *T ]
(III.20a)
∂V0 / ∂ X jt , j = 0,1,...,t - 1 and t = 0,...,T - 1

and T first-order conditions for the consumption choices

∂U/∂ C τ = 0 = - U 1 [ C*0 , C*1 ,..., C*T ] P0 ( τ )

(III.20b)
+ U τ +1 [ C*0 , C*1 ,..., C*T ] , τ = 1,...,T .

Noting that U1 > 0, the first-order conditions (III.20a) can be rewritten as ∂V0/∂Xjt = 0, j =
0,1,...,t-1 and t = 0,1,...,T-1, and in that form, are simply the generalization of condition
(III.13a) in the two-period case. Indeed, the interpretation given to (III.13a) in the two-period
case of choosing a physical production program so as to maximize the consumer's current wealth
carries over exactly to the T-period case. From (III.18) and (III.19), the set of {Xjt} which
maximizes W0 are the ones that maximize V0. But, the set of first-order conditions that
maximize V0 are simply ∂V0/∂Xjt = 0. Hence, (III.20a) simply says choose physical production
so as to maximize current wealth, and therefore the Value Maximization Rule applies in the
general T-period case.
Inspection of (III.20b) shows that it is identical to the first-order conditions for the pure-
exchange case (III.16) where the level of current wealth used is the maximized value,

51
 Robert C. Merton

*
W 0. Hence, as was shown in the two-period case, the solution of the T-period optimal
investment-consumption program for the consumer can be described as taking place in two steps:
namely, first, choose physical investments so as to maximize current wealth. Second, use the
exchange market to borrow or lend so as to achieve the most preferred feasible consumption
allocation.

On the Connection Between the T-Period and Two-Period Analyses

While the T-period consumer choice is a more realistic description of the world than the
two-period formulation, the analysis is more complex and is burdened by a barrage of notation.
Moreover, it does not readily lend itself to the relatively intuitive graphical display of the
solution. We have already shown that the fundamental behavioral characteristics (such as the
Value Maximization Rule) deduced in the two-period case carry over to the general T-period
case. We now show that, in essence, the general T-period problem can always be structured so as
to "look like" a two-period problem. Not only does this connection between the two problems
make the analysis of the T-period problem more tractable, but it also provides the appropriate
framework for studying the intertemporal consumption-investment choice problem in an
uncertain environment.
In the previous analysis, we solved the entire lifetime consumption choice problem by
having the consumer choose at date t = 0, (C0,C1,...,CT) so as to maximize U[C0,C1,...,CT]

subject to his budget constraint W 0 = ∑ P0 ( τ ) Cτ

T
. Suppose we move ahead one period to
τ =0

date t = 1. Suppose further that the consumer consumed C 0 units at date t = 0. The consumer

choice problem at date t = 1 can be formulated as choose (C1,C2,...,CT) so as to maximize

T
U[C0 ,C1 ,...,CT ] subject to his budget constraint W 1 = ∑ P1 ( τ - 1) Cτ where W1 is his
τ =1

52
Finance Theory

wealth at date t = 1. Note: C0 is not a choice variable at t = 1 because whatever was

consumed at time t = 0 is now past history.
We can solve the optimal choice problem at t=1 in the same way that the problem was
solved at t = 0, and in analogous fashion to (III.16), we arrive at the (T-1) first-order conditions
that

0 = - U 2 [ C 0 ,W 1 - ∑ P 1 ( τ - 1) Cτ ,C ,...,C ] P ( τ - 1)
T
* * *
2 T 1
(III.21) τ =2

+ U τ [ C ,C ,...,C ] , τ = 2,...,T,
+1 0 1
*
T

T
C 1 = W 1 - ∑ P1 ( τ - 1) Cτ .
* *
where From (III.21), it is clear that the optimal solution
τ =2

* * *
(C1 , C 2 ,...,CT ) will depend upon the amount of wealth W1, the prices

{P1 (1),...,P1 (T-1)}, C0 , and the form of the utility function U.

Define the function J by

(III.22) J[ C 0 ,W 1; P 1(1),..., P 1(T - 1)] ≡ U[ C 0 , C *1 ,...,C *T ] .

* * *
J is the "level" of utility associated with a consumption program of ( C 0 ,C 1 ,C 2 ,...,C T ) , and is
the maximal level of utility (corresponding to the most preferred feasible program) conditional

on consuming C0 units at date t = 0 and having wealth W1 at time t = 1. Because the prices

{P1(τ)} are not affected by the choices made by the consumer, they can be treated as
parameters. Hence, a shortened form for J is simply to write it as " J[ C 0 ,W 1 ]."

Return now to the original problem of selecting an optimal consumption program at time
t = 0. Of course, at t = 0, the consumer is free to choose any (feasible) level for C0. A
necessary condition for a consumption program to be optimal is that whatever level of
consumption is chosen for C0, the choices made for C1,C2,...,CT must be the best one can do

53
 Robert C. Merton

conditional on having chosen level C0 . That, of course, is exactly what

* * *
C1 ,C 2 ,...,CT represent in the t = 1 problem just solved where they represent the best the

consumer can do conditional on having chosen to consume C0 at time t = 0. Further, we have

that wealth at time t = 1, W1, can be expressed in terms of wealth at time t = 0 as

(III.23) W 1 = [ W 0 - C 0 ]/ P0 (1) .

I.e., whatever part of current wealth that is not currently consumed will grow in one period by the
one-period interest rate. Hence, having solved the conditional (on t = 0 consumption)
optimization problem as of t = 1, we can reformulate the consumer choice problem at t = 0 as:
Choose current consumption, C0, so as to maximize J[C0,W1] subject to the budget constraint
W0 = C0 + P0(1)W1. Expressed in this way, except for some notational differences, this problem
is in essence the same as the two-period choice problem solved in Case 1 of this section where
the utility function "J" replaces "U[Co,C1]" and "W1" replaces next period consumption "C1"
i.e., the T-period consumption problem can be reformulated as a two-period problem.
Although in the formulation the utility function, J, has utility depending upon (next
period's) wealth, the consumer still only gets direct utility from consumption. In effect, wealth
W1 acts as a "surrogate" for future consumption so that the utility "tradeoff" between C0 and
W1 is really a tradeoff between current and future consumption. J is sometimes called the
indirect or derived utility function, and provided that the direct utility function, U[C0,C1,...,CT],
is a well behaved, (quasi) concave function, J will be a well behaved, (quasi) concave function
in (C0,W1).
To solve the problem, we substitute for W1 using the budget constraint to get:

Max J [ C
C0
0 ,( W 0 - C 0 )/ P0 (1)]

which leads to the first-order condition

(III.24) 0 = J 1 [ C*0 ,W *1 ] - J 2 [ C*0 ,W *1 ]/ P0 (1)

54
Finance Theory

where subscripts denote partial derivatives of J with respect to the appropriate arguments and
* *
W1 ≡ (W0 - C0 )/P0 (1). As in (III.4), we can rewrite (III.24) in terms of the slope of an
* *
indifference curve through the point (C 0,W 1) as

(III.25) ( dW *1 /d C 0 )J = J * = - 1/ P0 (1),

* * *
where J ≡ J[C 0,W 1]. Figure 6 provides the graphical solution which is

Figure III.6

analogous to the one displayed in Figure 2 for the two-period problem. Although the derivation
presented here is more descriptive than rigorous, the analysis can be made rigorous by using the
mathematical technique of dynamic programming. While the explicit development of this
technique is more appropriately the subject of an advanced treatment of Finance, the interested

55
 Robert C. Merton

reader can find its development in the context of this problem in Fama [American Economic
Review, March 1970].
In summary, we have solved the general intertemporal consumption-investment problem in
a certainty environment. In so doing, we have shown that the creation of financial securities and
an exchange market will make the consumer better off. In particular, we showed that an
exchange market was the only means by which an individual consumer can convert future
income or output into current consumption. While this manifest function of the exchange market
more than justifies its existence (indeed, if such markets did not exist, we would have to invent
them), it has an important latent function as the means for permitting an efficient organization of
the economy's production. This important latent function is the topic of the next section.

56
IV. ON THE ROLE OF BUSINESS FIRMS, FINANCIAL INSTRUMENTS, AND
MARKETS IN AN ENVIRONMENT OF CERTAINTY

Every modern economy has as part of its institutional structure financial instruments,
capital markets, and business firms. In the previous section, we saw that the creation of a capital
market made households better off even in the simplistic environment of certainty. In this
section, we expand upon that analysis to explain the role of business firms.
In Section III, it was shown that in the presence of a capital market the optimal production
rule was to choose investment so as to maximize one's current wealth. In that model, the
consumer-owner of the technology made the production decisions, and therefore all technologies
were presumed to be owner-managed. However, for most modern economies, a majority of
production is carried out by business firms whose managers are not the (sole or even majority)
owners of the firm. This empirical fact raises several questions. In particular, why is this the
structure that we observe? What changes in the analysis of Section III are induced by this
separation of ownership from management? How can an efficient allocation of resources be
achieved with this (at least partial) centralization of production decisions?
To answer these questions, we begin with the following stylized description of the
formation of a business firm. First, the individual of Section III (the "founder") forms a
corporation and contributes his technology described by the production function f to the firm.
In return, he receives ownership of the whole firm, and therefore the right to one hundred percent
of the output of the firm. Second, he hires a manager (or technocrat) whose job it is to run the
firm. Specifically, the manager must choose the amount of physical production to undertake, and
then raise the additional resources necessary to carry out the production plans. The former is
called the investment decision and the latter is called the financing decision.
In this structure, the consumer has "turned over" the production and financing decisions
to the manager but still retains complete (albeit indirect) ownership of the technology through his
ownership of the firm's stock. Clearly, since the manager is hired by the owner, the manager's
job is to make decisions which are in the best interests of the owner. What is not so clear is how
he can achieve this goal. Of course, the manager could review each decision with the owner

57
 Robert C. Merton

including the production choices, cost of obtaining capital, etc., and ask him which combination
he prefers. However, in that case, the owner would have to have the same knowledge and spend
essentially the same amount of time as he would as an owner-manager, and therefore there would
be little point in hiring a manager to "run the business". Moreover, while this procedure might be
feasible when there is a single owner of the firm, it becomes increasingly more difficult as the
number of owners becomes large. Indeed, for a large corporation in the United States, the
number of shareholders or "owners" can range from several thousand to over a million. Hence, a
feasible or operational rule for managing the firm should not require the manager to "poll" the
owner(s) about his decisions. Furthermore, to be effective, the "right rule" should not require the
manager to know the tastes or endowments of the owner(s) because such data are virtually
impossible to obtain, and even if the data were available as of one point in time, they would
change over time. Indeed, since shares of stock change hands every day, the owners of the
corporation change every day. Thus, to be feasible, the right rule should be independent of who
the owner or owners are.
If a feasible rule for the manager to follow were found which would lead him to make the
same investment and financing decisions that each of the individual owners would have made
had they made the decisions themselves, then such a rule would clearly be the "right rule."
Because it was shown in (III.13a) and (III.20a) of Section III that an individual owner would
choose the investment plan which maximizes his current wealth, it follows that the right rule for
the manager is to choose investment so as to maximize current stockholders' (owners') wealth.
Moreover, inspection of (III.13a) and (III.20a) will show that the optimal investment decision
depends only upon the structure of the production technology and market interest rates (i.e., bond
prices). Specifically, it does not depend upon the tastes or endowments of the owners, and so it
can be made without any specific information about the owners. Therefore the manager can
follow the "right rule" without polling the owners with respect to his decisions.
There are a variety of ways to restate the operational criterion by which the manager
should make the investment decision for the firm. One such restatement is: "Choose investment
so as to maximize profits." To see this, consider the two-period case of Section III where the

58
Finance Theory

production technology available to the firm is described by f(X0) with X0 denoting the input
provided at time zero. If, as defined in Section III, P = 1/(1+r) is the price today of a unit of
output delivered next period, then by selling the (future) output of the firm today, the current
revenues of the firm are Pf(X0) and the current profit, ∏ , equals current revenues minus costs
or ∏ = Pf(X0) - X0. If the manager chooses X0 so as to maximize ∏ , then the chosen amount
* *
of investment X 0 will satisfy Pf ‘(X 0) = 1 which is exactly condition (III.13a).
This restatement of the operational criterion is also valid in the general case of T-periods
and uncertain cash flows to the firm provided that "profit" is defined in a very technical fashion.
However, it can be misleading if one applies the common (accounting or flow) usage for the
word "profit": namely, "profit in period t" is equal to period t gross cash flow minus period t
costs. So, for example, if the production process requires many periods, then which period's
profit is to be maximized? Or if either future revenues or costs are uncertain, then what is the
meaning of "maximize profits" when profits are described by a random variable?
A second restatement of the operational criterion is: "Using market interest rates, choose
investment so as to maximize the present value of the firm's net cash flows." This is the Present
Value Rule deduced for choosing among claims in Section II where the discount rates used are
the market interest rates because these represent the "cost of money" to the firm. In the two-
period case of Section III, the net cash flow in period 0 is – X0 and in period 1 is f(X0). Hence,
from (II.40), the Present Value Rule says "choose X0 so as to maximize – X0 + f(X0)/(1+r)."

The maximizing amount of investment,

*
( ) *
X 0 , will satisfy f ' X 0 = 1 + r which is exactly

(III.13a).
In the general T-period case of Section III, P0(t) denotes the current market price of $1
payable t periods in the future. The discount rate for period t cash flows, Rt, is defined in
(II.37). Because the Present Value Rule is to be applied using market interest rates, we have
t
from the present value formula (II.39) with x = 1 that P0(t) = 1/(1+Rt) . If, as in Section III, ft
denotes production output in period t and Xt denotes the total amount of inputs required in

59
 Robert C. Merton

period t, then (ft - Xt) is the net cash flow in period t, t = 0,1,...,T. Therefore, from (II.40), the
Present Value Rules says "choose production inputs, (Xjt, j = 0,...,t-1 and t = 1,...,T-1) so as to
maximize the present value of the net cash flows, PV0" which can be written as

∑ [ f τ ( X 0τ , X 1τ ,..., X τ -1,τ ) - X τ ]/(1+ Rτ )

τ
(IV.1) PV 0 ≡ .
τ =0

Noting that f 0 ≡ 0 and X T ≡ 0, we can rewrite (IV.1) as

T T -1
PV 0 = ∑ f τ ( X 0τ , X 1τ ,..., X τ -1,τ )/( 1 + Rτ ) - ∑ X τ /( 1+ Rτ ) .
τ τ
(IV.2)
τ =1 τ =0
τ
However, P 0 ( τ ) = 1/(1 + Rτ ) . Therefore, from (III.19) and (IV.2), PV0 = V0. Hence, the set
of choices for Xjt which maximize PV0 will satisfy
∂ PV 0 / ∂ X jt ≡ ∂V 0 / ∂ X jt = 0, j = 0,1,...,t - 1 and t = 1,...,T-1 which is exactly condition
(III.20a). Thus, unlike the "Profit Maximization" restatement, the "Present Value Rule"
restatement causes no ambiguities when the production process involves many periods.
However, like the "Profit Maximization" restatement, the Present Value Rule is not well defined
if the future net cash flows are uncertain i.e., what does it mean to maximize the discounted sum
of T random variables?
A third restatement of the operational criterion is: "choose investment so as to maximize
the current market value of the firm." In the two-period case, we determine the current market
value of the firm as a function of the investment decision X0 as follows. Suppose the firm
chooses to operate its technology at the intensity X0, and makes known to the public what its
plans are. At this time which is prior to the actual raising of the necessary funds to implement
the production plan, a market price for the firm is established. Call this market value V_. Note:
since, at this point, the original owner or "founder" still owns one hundred percent of the firm,
V_(X0) is the market value of his ownership (contingent on the firm being operated at intensity
X0). Moreover, although the firm has neither implemented its production plan nor even raised

60
Finance Theory

the necessary funds to purchase the inputs, the founder could actually sell either all or part of his
holdings for λV_ where λ is the fraction of his holdings that he chooses to sell. To determine
V_, we first establish what value the firm will have after it has raised the necessary additional
capital and entered into production. This value, call it V+(X0), is determined by noting that it
must be priced to yield a return competitive with other securities available to investors. In the
certainty environment, this competitive rate of return will be the interest rate r. Since the end of
period value of the firm will be f(X0), V+ must satisfy (1+r)V+(X0) = f(X0), and therefore
V+(X0) = f(X0)/(1+r).
The firm can raise the additional capital by either issuing debt or more equity. In either
case, it must raise $X0 to realize the production plan. If it is done by a debt issue, then the firm
issues a claim promising to pay a fixed amount, b, at the end of the period. If investors are to
provide $X0 to the firm today, then b must be chosen so that they will earn the competitive
interest rate r on their investment i.e., b = (1+r)X0, and the current market price of the debt will
be $X0. By definition, the market value of the firm is equal to the market value of its liabilities
which in this case are debt and equity. Hence, the market value of equity will equal V+(X0) - X0.
But, under this financing arrangement, the founder retains ownership of all the equity, and

therefore V_ (X 0 ) = V+ ( X 0 ) - X 0 or V_(X 0 ) = f(X 0 )/(1+r) - X 0 . Hence, if the

manager chooses X0 so as to maximize V_(X0), then that X0 will satisfy (III.13a) and will,
therefore, coincide with the decision which would have been made by the owner had he made it.
If the necessary capital is raised by issuing additional equity, then the original owner(s)
must give up some percentage of the equity. As with debt, the additional equity must be priced
to yield a competitive return. If γ is the percentage ownership given to the new shareholders,
then the value of their holdings as of next period will be γ f(X0). Therefore, to raise $X0 today,
γ must be chosen so that γ f(X0) = (1+r)X0 or γ = (1+r)X0/f(X0). Under this financing method,
the original owner's holdings will be worth (1–γ)V+(X0) = f(X0)/(1+r) – X0 = V_(X0) which is the
same as for the debt financed case. So, for either form of financing the right rule is to maximize
V_(X0).

61
 Robert C. Merton

To complete the analysis, suppose that in fact the firm has other assets in addition to the
production technology represented by f. Suppose these other assets are simply cash in the
amount of $C. By an analysis similar to the ones just used, one can show that V_(X0) =
f(X0)/(1+r) – X0 + C. Hence, even in the case where C ≥ X0 so that no external financing is
required to implement the production plan, the value-maximization rule leads to the right
* *
decision: Namely, choose X0 = X 0 such as to satisfy f ‘(X 0) = 1+r.
Using similar arguments in the general T-period case, one can show that the current
market value of the firm, V_, is equal to V0 as defined in (III.14). Hence, if the manager
chooses the production inputs {Xjt} so as to maximize V_, then the resulting choices will
maximize V0 which is exactly condition (III.20a). Therefore, the "Maximize Current Market
Value" Rule leads to the correct decisions when the production process involves many periods.
Although we have not as yet analyzed the case where future cash flows are uncertain, it is
clear that in that case the current market value of the firm is still well defined. (E.g., the future
cash flows of the IBM corporation are uncertain, but there is a current price for its stock which is
not uncertain). Hence, unlike the other two restatements, the "Maximize Current Market Value"
Rule causes no ambiguities if future cash flows of the firm are uncertain. Moreover, as will be
shown later in these Notes, provided that the capital markets are competitive, this Rule leads to
the "right" decision even in an uncertain environment.
In summary, the objective or criterion function for the firm is its current market value,
and good management is to make decisions so as to maximize the firm's criterion function.
Provided that managers operate in this fashion, an efficient allocation of the economy's
productive resources can be achieved with the ownership and management functions separated.
Note that the existence of a well-functioning capital market is essential to the feasibility
of this efficient separation. Of course, the manifest function of the capital market in terms of the
firm's actual transactions is to provide a means for the firm to raise the necessary resources to
carry out its production plans. However, an equally important, but latent function is to provide
information which is necessary for the manager so that he can make the "correct" decisions about
operating the firm. Specifically, while it is reasonable to assume that a good manager will have

62
Finance Theory

as much information about his firm's production technology, {ft}, as anyone, such "internal (to
the firm)" information is not sufficient to make decisions. Indeed, in the absence of a capital
market, we saw in (III.9) that, in addition, the manager would require "external (to the firm)"
information: Namely, the tastes and endowments of the owner. While, in the presence of a
capital market, the manager no longer requires this specific set of external information, he still
requires external information in the form of interest rates or prices. The existence of a capital
market allows the manager to substitute one set of external information which is relatively easy
to obtain for another set which is virtually impossible to obtain. In essence, prices in the capital
market "capture" all the essential information about tastes, endowments, and other investment
opportunities that the manager requires to make the correct decisions.
In reaching these results about the appropriate criterion function for the firm and the role
that capital markets play in the allocation of the economy's productive resources, we have made a
number of abstractions from reality: Namely, we assume perfect certainty about all current and
future events, and a "frictionless" world with no transactions costs, no indivisibilities, all
information available to everyone at no cost; and no explicit labor costs including management's
compensation. Moreover, we assumed that both individuals and firms behave competitively with
respect to their transactions in the capital markets. While, under these hypothesized conditions,
the owner would be just indifferent between the owner-manager structure or the separated
structure of a non-owner manager who makes decisions so as to maximize market value, the
introduction of the slightest "frictions" will generally lead to a definite preference for the
separated structure.
For example, a standard division of labor argument would lead to a definite preference for
the separated structure if either the cost of paying the professional manager is less than that
which the owner could earn in some other occupation or for the same cost, a professional
manager could be found who has a superior understanding of the firm's technology. Indeed, in an
owner-manager structure, the owner must have both the talents of a manager and the financial
resources necessary to carry out production. In the separated structure, no such coincidence is

63
 Robert C. Merton

required. Further, there is the "learning curve" or "going concern" effect which favors the
separated structure. Suppose the owner wants to sell all or part of his technology either now or at
a later date. In an owner-manager structure, the new owner will incur additional costs while he
becomes familiar with the operations of the firm. If there are economies of scale (a form of
"synergism"), then the separated structure is again favored because more than one person's
technology can be managed within a single entity at lower costs than within separate entities. As
will be shown later in the Notes, the introduction of uncertainty will cause individuals to want to
diversify their investments across many technologies, and diversification is difficult to achieve
within an owner-manager structure. Finally, provided that the manager has the most accurate
information about the firm's technology available (i.e., he is technically competent) and provided
that he uses this information to maximize the market value of the firm (i.e., he is benevolent),
then the owners of the firm need to know nothing about either the technology of the firm or the
intensity at which it is being operated. Hence, the separated structure allows for savings in the
costs of information gathering.
Thus, in an economy with production activities and a well functioning capital market, one
would expect to find that, in general, the owners of business firms will not be the managers and
that the ownership of such firms is dispersed among many individuals. Further, one would
expect to observe that, over time, the changes in the composition of the ownership would be far
more volatile than the changes in the composition of the management. However, if the
management follows the value-maximization rule, then it will be acting in the best interests of
the owners at each point in time.
Of course, one might be skeptical about the realism of such "mutual-admiration society"
behavior. It is certainly possible for the current management of a firm to be either incompetent
or malevolent, or both. Of course, the owners could "fire" the management by voting them out.
However, since a major benefit of the separated structure is that the owners can remain relatively
uninformed about the operations of the firm, it is not apparent how these owners could know
whether their firm is being mismanaged or not. The feasibility of voting rights being a solution
to the problem is further aggravated if ownership of the firm is widely dispersed. If that is the

64
Finance Theory

situation, then the holdings of any single owner are likely to be so small that he would not incur
the expense to become informed and to convey this information to other owners. Thus, voting
rights alone can do little to solve this dilemma. However, there is another mechanism called the
takeover which, at least in part, can.
Suppose some entity has identified a significantly mismanaged firm (i.e., one whose
management has chosen an investment plan which leads to a market value that is significantly
less than the maximum value that could be achieved). Specifically, the firm has production
technology f(X0) and the management has announced that their investment plan is to operate at

intensity
+ + * *
( ) *
X 0 , and X 0 ≠ X 0 where X 0 satisfies f ' X 0 = 1 + r. Moreover, by

supposition, V _ X 0 ( )
+
is significantly less than ( )*
V _ X0 . In response to the current
+
management's announced plan, the market value of the firm will be V_(X 0). Suppose that
the entity buys all the shares of this firm at the current market value. Having done so, it fires the
management and installs a new management that will choose to operate the firm at intensity
*
X 0 . Having announced the change in the firm's investment plans, the entity now sells the
*
shares of the firm at the new market price, V0 (X 0 ), based upon the new investment plan.

Hence, by taking over the firm and changing its investment plans, the entity earns an immediate
* +
profit of V_(X 0 ) - V_(X 0 ). Note: the entity did not have to add any tangible resources to

the firm to achieve this profit. Hence, the only expenses incurred are the cost of identifying a
mismanaged firm and the cost of acquiring the firm's shares.
While the cost of identifying a mismanaged firm will vary, it can be quite low if the entity
happens to be a supplier, customer, or competitor of the firm because much of the information
required may have been gathered for other purposes already. For this reason, the takeover
mechanism can work even if resources are not spent for the explicit reason of identifying
mismanaged firms. However, if significant mismanagement of firms were widespread, then it

65
 Robert C. Merton

would pay to spend resources in search for such firms in much the same way that resources are
spent on research for new physical investment projects. Therefore, the threat of a takeover and
the subsequent removal of management provides a strong incentive for current management
(acting in its own self interest) to act in the interests of the firm's current stockholders by
maximizing market value. Indeed, even in the absence of any explicit instructions from the
shareholders or knowledge of the theory for "good management," one might expect managers to
move in the direction of value maximization as simply a matter of self-preservation. Moreover,
it should be noted that the analysis depends in no way on whether the source of the
mismanagement is incompetence or malevolence (i.e., whether the current management are
"fools" or "knaves"), and therefore the takeover mechanism serves equally well to correct either
one. Of course, the effectiveness of the takeover mechanism will depend upon how much of a
threat it poses for current management. For example, in an attempt to prevent the formation of
monopolies in various product markets, the Justice Department will take legal action under the
anti-trust laws to prevent mergers or acquisitions which might reduce competition. Because it is
more likely that a supplier, customer, or competitor will be the entity to identify a mismanaged
firm, this public policy will tend to reduce the threat of takeover. For much the same reason, the
managements of larger firms are probably less vulnerable to a takeover bid. As an aside, this
example illustrates how public policy objectives can be in conflict with one another where no
simple resolution of the conflict is available.
In summary, the gains in efficient resource allocation and reduced costs from the
combined institutional structure of a well functioning capital market and owner-separated-from-
manager business firms does not rest upon the delicate and naive assumption of a mutual-
admiration society with no conflicts between the interests of owners and managers. Indeed, in
the absence of any external "checks," the management of a firm with dispersed ownership
certainly has the opportunity to enrich themselves at the expense of the owners. However,
because a larger market value for the firm reduces the chances for a takeover and makes the
owners better off, the external check of the takeover mechanism forces management to act as if
its interests were coincident with the owners.

66
Finance Theory

In concluding this analysis of the business firm, it is worthwhile to reiterate remarks made
in the Introduction. Firms are economic organizations designed to serve people by performing
specific functions. While in the corporate form, the firm legally has a "corpus," it does not have
a "soul," and therefore has no independent right to existence. Thus, the reader should examine
with care those theories that treat the firm "as if" it were an individual and then deduce the
"proper" rules for good management based upon an exogenously specified utility function for the
firm. Similarly, the reader should be skeptical of theories that treat the firm as if it were "an
island unto itself" and, as such, have management decisions based only upon data which are
"internal" to the firm. The decision as to whether or not a specific project is to be undertaken
should not be based solely upon the engineering and economic specifications of the project.
Such decisions must take into account the economic environment in which the firm is operating,
and to do so require external information. In the analysis presented here, market interest rates
provide the appropriate connection with the outside environment. The switching phenomenon
(illustrated in Problem II.2 of Section II) clearly demonstrates that the right decision will not be
invariant to these external rates.
Stocks, bonds, and other financial instruments which are an essential part of the
proceeding analyses are all examples of financial assets. In its purest form, a financial asset,
unlike a physical asset, has no value for itself but derives all its value from what it gives its
owner a claim on. For example, a stock certificate has virtually no value as a physical asset (i.e.,
as a physical asset, its value is that of a used piece of paper), but it may have great value as a
financial asset because it represents a percentage ownership or claim on the firm's physical assets
and their associated earnings flows. In our stylized description of the formation of a business
firm, the founder gave a physical asset (the technology) to the firm in exchange for a financial
asset (shares of stock in the firm) giving him a claim on the output of the firm. When the firm
raised the necessary additional capital for production, investors may have given the firm physical
assets (the inputs required for production) in return for a financial asset (either debt or equity in

67
 Robert C. Merton

the firm). Or, more likely, the investors exchanged one financial asset (money) for another (debt
or equity), and then the firm gave this money to another firm in return for raw materials.
The principal function of a financial asset is to serve as a store of value. Indeed, the
capital markets could not exist without financial assets. While it is not necessary that a financial
asset have no intrinsic worth (i.e., stock certificates could take the form of engravings on gold
bars), there are two good reasons why it is preferable that it not have any significant value as a
physical asset. First, because to have a positive value as a physical asset is not required for a
financial asset to serve its function, to use something which has a significant physical value as a
financial asset is to waste scarce economic resources. Second, if a financial asset also has value
as a physical asset, then its value will be determined as the maximum of its value as either a
physical asset or a financial asset. If its value as a physical asset should exceed its value as a
financial asset, then it will cease to serve the function of a financial asset. For example, coins
made from metals (e.g., copper, silver or gold) have frequently had the value of their (melted-
down) metal content exceed their stated monetary value in which case they have ceased to be
used as money.
Another function served by financial assets is to allow divisibility of ownership of
physical assets which are not generally divisible. For example, to physically divide a race horse
would be to destroy virtually all its value. However, by issuing a financial asset which provides
for a fractional ownership of the race horse (i.e., a right to a certain percentage of all purses, stud
fees, and sales) accomplishes divisibility without affecting the underlying physical asset.
The types of financial assets that are traded in markets are easily identified and are of a
standard form. Hence, they are reasonably liquid in that they can be sold within a short period of
time at something near to the current market price. In general, the existence of financial assets
lowers the requirements for information needed by both parties in order to have trade. Unlike
many physical assets, financial assets are relatively easy to transport from one physical location
to another.
While these services alone would be sufficient to explain the existence of financial assets,
the most important reason for their creation is that without financial assets, it is necessary that

68
Finance Theory

saving must equal investment for each economic unit. For households, saving equals income
minus consumption. In general, for all units, savings equals current income minus current
expenditures. In every case without financial assets, the saving of each unit would have to equal
its investment in physical assets. Indeed, in the "Robinson Crusoe" economy with no financial
assets in Section III, this constraint was specifically stated in (III.6), and it was the relaxation of
this constraint through the introduction of financial assets which allowed the individual to choose
a better allocation. While, even with financial assets, it is still necessary that for the economy as
a whole aggregate saving must equal aggregate investment, it is no longer necessary that saving
must equal investment for each unit.
As a form of summary, we illustrate the benefits of financial assets and a capital market
for both the pure exchange and production cases with a three-period, two-person economy
example. In all cases, we assume that the preferences and "wage income" endowment of person
#I are given by:
I I I I I I I
(IV.3) U ( C 0 , C 1 , C 2 ) = log( C 0 ) + log( C 1 ) + log( C 2 )
and
I I I
y0 = 300 ; y 1 = 1500 ; and y 2 = 2700 .

Similarly, for person #II, we assume that the preference and "wage income" endowment are given
by
II II II II II II II
(IV.4) U ( C 0 , C 1 , C 2 ) = log( C 0 ) + log( C 1 ) + log( C 2 )
and
II II II
y0 = 2700; y1 = 1500; and y 2 = 300.

Note that in this example, both people have identical preferences and similar magnitudes of
income except the time patterns of their receipt are reversed.

Problem IV.1: No Production and No Exchange Market:

69
 Robert C. Merton

In this case, both individuals have no choice but to consume their current income each period
because "no production" implies "no storage." Hence,
I I I I I I I
C0 = y0 = 300; C1 = y1 = 1500; C 2 = y 2 = 2700, and therefore, U = 9.08.
II II II II II II II
Similarly, C 0 = y 0 = 2700; C1 = y1 = 1500; C 2 = y 2 = 300; and U = 9.08.

Problem IV.2: No Production with an Exchange Market:

Suppose now that there is an exchange market with market interest rates (r1,r2). The current
j j j j
wealth of person j is given by W0 = y 0 + y1 /(1+r1 )+ y 2 /(1+r1 )(1+r2 ), j = I,II. As

discussed in Section III, person j will choose a consumption program as follows:

Max { log( C 0j ) + log( C 1j ) + log( C 2j )}

j j j j
subject to the constraint that W0 = C0 + C1/(1+r1 ) + C 2 /(1+r1 )(1+r2 ). The optimal
solution is given by
j* j
C 0 = W 0 /3
j* j* j
(IV.5) C 1 = (1 + r 1 ) C 0 = (1 + r 1 )W 0 /3
j* j* j
C 2 = (1 + r 2 ) C 1 = (1 + r 1 )(1 + r 2 )W 0 /3
for j = I,II.
To determine the market interest rates (r1,r2), we impose the equilibrium condition that
aggregate (planned) saving must equal aggregate (planned) physical investment for the economy
in each period. Because there are no means of production (including storage), aggregate
investment, and hence aggregate saving, must equal zero i.e.,
I* II* I II
Ct + Ct = y t + y t , t = 0,1,2. The equilibrium set of interest rates that allow these
market clearing conditions to be satisfied is (r1 = 0, r2 = 0). Hence,

70
Finance Theory

I II j*
W0 = W0 = 4500, and from (IV.5), we have that C t = 1500 for j = I, II and t
I II
= 0,1,2 and that U = U = 9.53. Hence, both people are better off with an exchange
market than they were in Problem IV.1. Note that in both problems, aggregate saving,
I II j j j*
St + St , equals zero in each period where St ≡ y t - C t , (j = I, II and t = 0,1,2).
I II j
However, in Problem IV.1, St = St = 0 whereas in this problem, St ≠ 0 i.e., saving need

not equal investment for each unit.

Note that even if we had relaxed the no-production condition in Problem IV.1 to allow for
costless storage, the resulting solution would not have been the same as with an exchange
market. Person #II could achieve the optimal (1500, 1500, 1500) consumption plan using
storage. However, even with storage, Person #I would still choose the same (300, 1500, 2700)
allocation chosen in the absence of storage. This underscores the point that storage only allows
one to "transport" goods forward in time and not backwards whereas by having financial assets
and an exchange market, one can change his allocation in either direction.

Problem IV.3: Production and No Exchange Market

Suppose that Person #I has, in addition to his wage income endowment, a production technology
which transforms one unit of input this period into two units of output next period
I
(i.e ., f t+1 = 2X t , t = 0,1). Suppose further that Person #II has the (storage) production

technology which transforms one unit of input this period into one unit of output next period
II
(i.e., f t+1 = X t , t = 0,1). Even though Person #I's technology provides for a 100 percent rate

of return per period, his optimal choice is to not use his technology to produce any goods. The
I
reason is that his current period's income (y 0 = 300) is so small by comparison with his later

period's income that he prefers to consume all his current income rather than produce. Therefore,

71
 Robert C. Merton

he derives no benefit from his technology, and has the same consumption allocation as in the no-
production case of Problem IV.1. On the other hand, Person #II does use his production
technology to achieve the optimal allocation (1500, 1500, 1500). Because Person #II is
producing goods with a technology which is inferior to the one owned by Person #I who is not
producing at all, there is an obvious "loss" to the economy. Under this allocation, there is a total
of 3000 available to the economy in each period. However, if the 1200 that Person #II carries
over from the current period by storage had been employed in Person #I's technology, then there
would have been an extra 1200 available to the economy in the second period with a
corresponding compound increase for the third period. As we now show, this inefficient
allocation is corrected by the introduction of a competitive exchange market.

Problem IV.4: Production with an Exchange Market

Suppose we now combine the production technologies of Problem IV.3 with the exchange
market in Problem IV.2. For a competitive exchange market and the given technologies the
equilibrium interest rates (r1,r2) must each be greater than or equal to 100 percent. Otherwise,
Person #I would register an indefinitely large demand for current period goods. Indeed, by
requiring that markets clear, the equilibrium rates will just equal 100 percent. At r1 = r2 = 1, the
present value of the (superior) technology will be zero. Hence, the wealth of each person will be
equal to the present (or "capitalized") value of his wage income. Thus,
I II
W0 = 300 + 1500/2 + 2700/4 = 1725 and W0 = 2700 + 1500/2 + 300/4 = 3525. From
(IV.5), the optimal consumption program chosen by Person
I* I* I* I
#I is C0 = 575; C1 = 1150; and C2 = 2300 with U = 9.182. Similarly for
II* II II* II
Person #II, C0 = 1175; C 2 = 2350; and C2 = 4700 with U = 10.113. As a
result of the introduction of an exchange market, both people are better off than they were in
Problem IV.3.

72
Finance Theory

In summary, the need for financial assets arises form the discrepancy between (desired)
saving and investment of individual economic units. If investment exceeds saving for a given
economic unit, it can finance this "saving deficit" by either issuing a financial asset (a liability to
the issuer) or by selling an already existing financial asset. Purchase and sale transactions of
already existing financial assets take place in a secondary market. Hence, with such a market,
the outstanding stock of financial assets need not change even if some units have a saving deficit.
Primary and secondary markets are an efficient means of channeling required investment funds
to the most productive units. While the analyses and examples have been centered around
private sector investment for a closed economy with no government, the same principles apply to
public investment. Thus, the same analyses could be applied to less developed countries where
the government is the main investment unit. While the analyses in Problems IV.1 - IV.4 were
structured along the lines of two individuals, the same analyses and resulting benefits would
apply if the two people were reinterpreted as two countries where international capital flows
replaced individual savings-investment deficits. The following flow and balance sheet
statements provide a detailed description of savings and investment flows for the case examined
in Problem IV.4.

73
 Robert C. Merton

Flow Statement t = 0

Person #I Person #II Aggregate (I + II)

Wage Income 300 2,700 3,000

Production Income 0 0 0
Operating Income 300 2,700 3,000
Interest Income (Expense) 0 0 0
Net Income 300 2,700 3,000
–Consumption –575 –1,175 –1,750
Savings (275) 1,525 1,250
–Investment –1,250 0 –1,250
Savings Surplus (Deficit) (1,525) (1,525) 0

Balance Sheet t = 0+

Person Person Person Person

#I #II Aggregate #I #II Aggregate
Assets Liabilities

Capital 1,250 0 1,250 1,525 0 1,525 Debt

Bonds 0 1,525 1,525
Capitalized
Wage Income 1,425 825 2,250 1,150 2,350 3,500 Net Worth
2,675 2,350 5,025 2,675 2,350 5,025

Flow Statement t = 1
Person #I Person #II Aggregate (I + II)

Wage Income 1,500 1,500 3,000

Production Income 1,250 0 1,250
Operating Income 2,750 1,500 4,250
Interest Income (Expense) (1,525) 1,525 0
Net Income 1,225 3,025 4,250
–Consumption –1,150 –2,350 –3,500
Savings 75 675 750
–Investment –750 0 –750
Savings Surplus (Deficit) (675) 675 0

74
Finance Theory

Balance Sheet t = 1+

Person Person Person Person

#I #II Aggregate #I #II Aggregate
Assets Liabilities

Capital 2,000 0 2,000 2,200 0 220 Debt

Bonds 0 2,200 2,200
Capitalized
Wage Income 1,350 150 1,500 1,150 2,350 3,500 Net Worth
3,350 2,350 5,700 3,350 2,350 5,700

Flow Statement t = 2

Person #I Person #II Aggregate (I + II)

Wage Income 2,700 300 3,000

Production Income 2,000 0 2,000
Operating Income 4,700 300 5,000
Interest Income (Expense) (2,200) 2,200 0
Net Income 2,500 2,500 5,000
–Consumption –2,300 –4,700 –7,000
Savings 200 (2,200) (2,000)
–Investment (Liquidation) –(2,000) 0 –(2,200)
Savings Surplus (Deficit) 2,200 (2,200) 0

75
V. THE "DEFAULT-FREE" BOND MARKET AND FINANCIAL
INTERMEDIATION IN BORROWING AND LENDING

In Sections III and IV, we derived some of the important functions served by a capital
market in the efficient allocation of the economy's productive resources. By making a number of
abstractions from reality, we were able to derive these functional characteristics using a relatively
simple structure. The most important of these abstractions in terms of simplification was the
assumption of a perfect certainty environment. This assumption ensured that the future course of
interest rates were known in advance and that the promised payments on all claims would be met
at the time promised. Unfortunately, the perfect certainty assumption is also the least realistic of
the abstractions made, and therefore will be jettisoned beginning in Section VIII at the cost of
introducing a more complex structure. However, in this section, we continue (at least in part)
with the assumptions of Sections III and IV to analyze the "default-free", fixed-income securities
part of the capital market where maintaining the certainty assumption does the least violence to
reality.
Fixed income securities are claims with fixed or stated payments promised at specified
times. The most common type of fixed-income security is debt. However, what is "promised" is
not always paid, and the event of not meeting a promise on a fixed-income security is called
default. A default-free, fixed-income security is contained in that subset of fixed-income
securities where the promised payments will be met with (virtual) certainty. Strictly interpreted,
the only securities that fall in this class are debt issues of the federal government and its agencies
or debt issues which are guaranteed by the federal government and this is because the federal
government can always meet money-fixed obligations by "printing" money. However, in
practice, many state and some local government issues as well as some "gilt-edge" corporate debt
issues are treated as if they were default-free. These securities are not only important because
they represent a not insignificant fraction of the capital market ($800 billion of federal
government debt obligations are held by the private sector), but also because their prices provide
the base yield upon which other securities' prices are determined. For example, because the
promised payments on fixed income securities are also the maximum payments that their holders

76
Finance Theory

can receive, the promised yield on a fixed income security must be at least as large as the yield on
a corresponding, default-free fixed income security.

On the Pricing of Discount Bonds and the Term Structure of Interest Rates

We begin the study of default-free income securities by examining how prices are
determined for discount bonds which promise a payment of $1. As in Section III, let Pt(τ)
denote the price at date t of a default-free bond which promises a payment of $1 at date (t + τ).
Let rt denote the one-period rate of interest that can be earned between date (t - 1) and date t.
Suppose that there is a market in which these discount bonds are traded and that this market is
"open" for trading each period. Further, suppose that there are no transactions costs or taxes.
Consider two bonds with maturities τ1 and τ2 respectively at date t = 0. The return per dollar
from holding bond j over the next period is equal to the ratio of bond j's price next period to
its current price, i.e., P1(τj – 1)/P0(τj), j = 1,2. Suppose that P1(τ1 – 1)/P0(τ1) < P1(τ2 –
1)/P0(τ2). Then, any investor who plans to invest in the first bond now would be better off to
purchase the second bond now and wait (at least) until next period to purchase the first bond. To
see this, let the investor have $I to invest now. If he buys the first bond, then he will purchase
N1 = I/P0(τ1) bonds which will be worth $N1P1(τ1–1) next period. If instead he invests in the
second bond now, then he will purchase N2 = I/P0(τ2) bond now which will be worth $N2P1(τ2 -
1) next period. By hypothesis, N2P1(τ2–1) > N1P1(τ1–1). Hence, by following the second
strategy, the investor will have enough money next period to buy N1 of the first bonds plus he
will have [N2P1(τ2–1) – N1P1(τ1–1) left over. At these prices, the second bond is said to
dominate the first bond because independent of preferences or time horizon, every investor
would prefer the second bond to the first.
If P1(τ1 - 1)/P0(τ1) > P1(τ2 - 1)/P0(τ2), then by a similar argument, the first bond would
dominate the second bond. Since no investor would be willing to hold a dominated bond, a
necessary condition for equilibrium is that no bond dominate any other bond. Thus, in
equilibrium, we have that

77
 Robert C. Merton

(V.1) P 1( τ 1 - 1)/ P 0 ( τ 1 ) = P 1( τ 2 - 1)/ P 0( τ 2 )

for all maturities τ1 and τ2. Because these bonds are default-free, Pt(0) ≡ 1 for all t, and by
definition, Pt(0)/Pt-1(1) ≡ 1 + rt. Hence, we can rewrite equilibrium condition (V.1) as

(V.2) P 1( τ - 1)/ P 0( τ ) = 1 + r 1

for all maturities τ. Moreover, the same argument can be used to show that for any starting
date t, the one-period return per dollar on bonds of all maturities must be the same. Therefore,
condition (V.2) can be rewritten more generally as

(V.3) Pt+1 ( τ - 1) / Pt ( τ ) = 1 + r t+1

for all dates t = 0, 1, 2, ... and all maturities τ.

Consider now a specific bond which at t = 0 has maturity T. At date t = T, the bond
matures and will therefore have price PT(0) = 1. At date t = T-1, we have from (V.3) that its
price must satisfy PT-1(1) = PT(0)/(1 + rT) = 1/(1 + rT). At date t = T - 2, we have again from
(V.3) that its price must satisfy PT-2(2) = PT-1(1)/(1+rT-1) = 1/[(1 + rT)(1 + rT-1)]. Continuing in
this "backwards" recursive fashion, we can derive the price that this bond must have so that it
neither dominates nor is dominated by a one-period bond at any point in time during its
existence. The price formula is given by

⎡ T ⎤
(V.4) P0 (T) = 1/ ⎢ ∏ (1 + r j )⎥
⎣ j=1 ⎦

and this must hold for all maturities T.

As the reader will note from (II.37) and (II.38), (V.4) simply says that "the equilibrium
price for a default-free, discount bond is given by the present value formula using the current and
future one-period, market interest rates." Further, using (II.38), we can rewrite (V.4) in terms of
the average compound rate of return as

78
Finance Theory

P0 (T ) = 1/ (1 + RT ) .
T
(V.5)

Thus, given complete knowledge of the future course of one-period interest rates,
{r1,r2,...,rT,...}, one can use (V.4) to determine the current prices of default-free, discount bonds
of all maturities. However, the process can also be "reversed": namely, given a complete set of
current prices for default-free, discount bonds of all maturities, {P0(1),...,P0(T),...}, the future
course of one-period interest rates can be determined. From (V.4), we have that

(V.6) r T = [ P0 (T - 1)/ P0 (T)] - 1

for T = 1, 2,... . Note that the difference between (V.6) and (V.3) is that (V.6) specifies a
condition on the price ratio of two bonds with different maturities at the same point in calendar
time while (V.3) specifies a condition on the price ratio of the same bond at two different points
in calendar time. However, (V.3) and (V.6) can be combined to specify a relationship between
the dynamics or time series of a specific bond's price over time and the statics or cross-sectional
series of different maturity bond prices at the current time. This relationship can be written as

(V.7) Pt ( τ - 1)/ Pt -1 ( τ ) = P0 (t - 1)/ P0 (t) ,

for all future dates t = 1, 2, ... and all maturities τ. Thus, from a cross section of current bond
prices, one can deduce the dynamics of future bond prices and interest rates.
In describing the cross-sectional structure of current bond prices, it is the practice to quote
the average compound returns or yields on the different maturity bonds rather than their prices.
These yields are determined by the current prices using (V.5). I.e.,
-1/T
(V.8) RT = [ P0 (T) ] -1.

The curve generated by plotting the yield, RT, against maturity, T, is called the yield curve or
the term structure of interest rates. A "rising" term structure is one where RT+1 > RT for all T
and is illustrated in Figure V.1. A "U-shaped" term structure is one where either RT+1 > RT for

79
 Robert C. Merton
* * * *
0 < T < T and RT+1 < RT for T < T or RT+1 < RT for 0 < T < T and RT+1 > RT for T < T,
and is illustrated in Figure V.2. A "flat" term structure is one where RT = RT+1 for all T. Of
course, in general, the only restrictions on the shape of the term structure are that the current
bond prices implied by these yields satisfy (V.4) and that the future one-period interest rates
implied by these yields are non-negative.
As was discussed briefly at the end of Section II, one should not confuse "RT" with "rT"
in interpreting the yield curve. While the {rT} can be deduced from {RT}, the two are not
equal to one another, and indeed a plot of the rT versus T can look qualitatively quite different
from the yield curve. Moreover, if one buys a discount bond at a yield of RT(T > 1), then, in
general, its rate of return or growth in value in each period will not be the same and, indeed, can
be quite different from RT. Table V.1 illustrates these points by showing how the yields for
different maturity bonds at the current time correspond to the time pattern of one-period interest
rates. In addition, it also provides a comparison of the pattern of appreciation from an initial
$1000 investment in a fifteen-period discount bond with the pattern which would be generated if
each period the $1000 investment grew at the yield rate, R15.

80
Finance Theory

Figure V.1
A “Rising” Term Structure

Figure V.2
A “U” Shaped Term Structure

81
 Robert C. Merton

Table V.1
Interest Rates, Yields, and Investment Returns Comparisons

Actual Value Value of $1000

At One-Period of $1000 Initial Investment
Calendar Interest Yield, Initial at R=6.2% per
Time, t Rate, rt Rt Investment Period

1 2% 2.0% $1020 $1062

2 5 3.5 1071 1128
3 10 5.6 1178 1199
4 8 6.2 1272 1273
5 4 5.8 1323 1353
6 2 5.1 1350 1437
7 3 4.8 1390 1526
8 4 4.7 1446 1621
9 5 4.7 1518 1722
10 6 4.9 1609 1830
11 7 5.1 1722 1944
12 8 5.3 1860 2065
13 9 5.6 2027 2193
14 10 5.9 2230 2330
15 11 6.2 2475 2475

As Table V.1 along with Figure V.3 demonstrates, the actual appreciation pattern from
investing in a fifteen-period discount bond is very different from the hypothetical pattern
generated by a constant rate of growth at that bond's yield rate. Because the bond's price
dynamics must satisfy (V.3), the rate of return on the bond in each period must equal that period's
one-period interest rate, and hence, the observed erratic return pattern is a direct reflection of the
variability in those rates. Although the actual and hypothetical investments are (virtually) equal
at the end of period four, this is a coincidence of the particular pattern in the one-period rates. In
general, unless the term structure is flat, the values of the two investments will coincide only at
the maturity date of the bond.
Similarly, Table V.1 and Figure V.4 show that the cross-sectional pattern of yields is very
different from the time series of one-period interest rates. As with any average, the changes in

82
Finance Theory

the yields are less pronounced than the changes in the one-period rates. Moreover, the "turning
points" or the (approximately) flat points in the yield curve always occur after the "turning
points" in the one-period rates. So, for example, the local "peak" in the yield curve at the end of
period four occurs after the local peak in the one-period rates at the end of period three.
Similarly, the local "trough" in the yield curve between periods eight and nine occurs after the
trough in the one-period rates at the end of period six. Note too that the longer-term yields are
less sensitive than the shorter-term yields to a change in any one of the one-period rates. From
period one to period three, the one-period rates went from 2 percent to 10 percent while the
yields went from 2 percent to 5.6 percent. However, from period six to period fifteen, the one-
period rates increased steadily from 2 percent to 11 percent while the yields only went from 5.1
percent to 6.2 percent. Finally, because the yields are "geometric" averages, the T-period yield
will always be less than the (arithmetic) average of the one-period rates for the T periods (i.e.,
T

RT < ( ∑r t )/T) .
t=1

83
 Robert C. Merton

Figure V.3

84
Finance Theory

Figure V.4

85
 Robert C. Merton

The relationships between the {rT} and {RT} illustrated in Table V.1 are patterns that
hold in general. From (V.4) and (V.5), we have that

(V.9) (1 + RT +1 )/(1 + RT ) = [(1 + r T +1 )/(1 + RT ) ] 1/T +1 .

< >
Therefore, as was pointed out in (II.41), RT+1 = RT if and only if r T+1 = RT . Specifically, flat or
> <

turning points in the yield curve correspond to maturities where rT+1 = RT. Hence, if he pattern
*
of one-period rates between t = 0 and t = T is a rising one (i.e., r1 < r2 < ... < rT*), then from
(V.9), the yields for maturities in that region will also be rising (i.e., RT > RT-1 for T = 1,
*
2,...,T ). Moreover, for the peak in yields to coincide with the peak in one-period rates (i.e.,
RT*+1 < RT*), the one-period rate rT*+1 would have to satisfy
T*
log(1 + r T *+1 ) < [ ∑ log(1 + r t )]/ T * which for T* much larger than one would require that rT*+1
1

<< rT*. Hence, if the yield curve rises significantly over an extended number of periods, then
almost certainly, the peak in the yield curve will occur after the peak in the one-period rates. A
similar argument applies for the trough in the yield curve occurring after the trough in one-period
rates when the yield curve is declining.
From (II.37), we can derive the effect on the T-period yield from a change in one of the
one-period rates to be

1
(V.10) ∂ RT /∂ r t = {(1 + RT )/(1 + r t )}, t = 1,2,...,T.
T
Hence, the sensitivity of the yield curve between (T - 1) and T to the one-period interest rate for
that period can be written as

1
(V.11) ∂( RT - RT -1 )/∂ r T = {(1 + RT )/(1 + r T )} .
T
Inspection of (V.11) shows that the longer is the maturity, the less sensitive the yield curve will
be to distant future one-period rates. Indeed, in the limit as T → ∞, ∂(RT - RT–1)/ ∂rT → 0.
Therefore, virtually all yield curves will exhibit a "flattening" pattern for very long maturities.

86
Finance Theory

Since this pattern will occur for virtually all time paths in future one-period rates, great care must
be exercised in using the yield curve to draw inferences about the distant future one-period
interest rates. For example, such a pattern does not imply the existence of a stable, "long-run" or
"steady-state" one-period interest rate.
While we have formulated the term structure analysis here in discrete time with an (as of
yet) unspecified minimum time interval of "one period", it is common practice to study the yield
curve as if it were continuous and to assume that the one-period or "shortest" bond has an
infinitesimal length of time until maturity: namely dt. Using the notation developed in (II.10) of
Section II, let rc(t) denote the rate of interest between dates t and t + dt. Equation (V.3) can be
rewritten as

(V.12) Pt+dt ( τ - dt)/ Pt ( τ ) = 1 + r c (t)dt.

By using the same "backwards" recursive analysis which led to (V.4), we derive from (V.12) that
bond prices at date t = 0 must satisfy
T
(V.13) P 0(T) = exp [- ∫ r c(s)ds] ,
0

for all maturities T. If Rc(T) denotes the average continuously-compounded rate of return on a
discount bond that matures at time T in the future, then it follows that

(V.14) P0 (T) = exp[- Rc (T)T] ,

and from (V.13) and (V.14), the relationship between the yield curve and future interest rates is
given by
T
(V.15) R c(T) = [ ∫ r c(s)ds]/T
0

Hence, in the limiting case of continuous time, the average compound return is equal to a simple
arithmetic average of the future short rates.
To further explore the relationship between the yield curve and future short interest rates,
we differentiate (V.15) to obtain

87
 Robert C. Merton

(V.16) d Rc (T)/dT = [ r c (T) - Rc (T)]/T .

>
In an analogous fashion to (V.9) in discrete time, we have from (V.16) that dR c(T)/dT = 0 if
<

>
and only if r c(T) = R c(T) . Therefore, turning points in the yield curve correspond to maturities
<
* * *
{T } where rc(T ) = Rc(T ). As in the discrete time analysis, dRc(T)/dT tends to zero as T →
∞, and therefore, independent of rc, the yield curve "flattens out" for large T.
The curvature of the yield curve can be studied using (V.16) and its derivative which is
given by

d R c(T)/d T = {d r c( τ )/dT - 2d R c(T)/dT}/T .

2 2
(V.17)

* *
At turning point maturities {T }, the yield curve will have a local peak if drc(T )/dt < 0 and a
*
local trough if drc(T )/dt > 0. Hence, each turning point in the yield curve will always occur
after the corresponding turning point in future short interest rates. Points of inflection or zero
+ +
curvature in the yield curve will occur for those maturities {T } such that drc(T )/dt =
+
2dRc(T )/dT.

Problem V.l: Analyzing the Term Structure

Suppose that the yield curve at the current time is given by:

0≤T ≤T
2 3
R c(T) = R + AT - BT ,
3 4
= R + [AT - BT ]/T, T ≥ T

where A > 0, B > 0, and T ≡ 3A/4B. What is the future time pattern of short interest rates
implied by this yield curve, and how does this pattern compare with the shape of the yield curve?

From (V.15) and (V.16), we have that the time pattern of short rates implied by this yield
curve can be written as

88
Finance Theory

2
r c(t) = R + t (3A - 4Bt) , 0≤t ≤ T
=R , t ≥T .
From (V.16), we have that

dRc (T)/dT = T(2A - 3BT) , 0 ≤ T ≤ T .

*
Hence, the time path of short rates starts at r c (0) = R; rises monotonically until it peaks at t =

A/2B; it then declines monotonically until at t = T, it remains constant at R .

While the two patterns are similar, the time path of the short rates rises more steeply and
* *
peaks earlier than the yield curve (i.e., t < T ). Moreover, the peak level of the short rates
* 3 2 * 3 2
r c ( t ) = R + A /4 B is higher than the peak level of the yield curve Rc ( T ) = R + 4 A /27 B .
To examine the curvature of the yield curve, we have from (V.17) that

d R c(T)/d T = 2(A - 3BT), 0 ≤ T < T

2 2

3 3
= AT /2T , T > T.
+
Hence, the yield curve starts out convex until it reaches an inflection point at T = A/3B and

becomes concave on the interval (T , T ) .

+
While the first derivative of the yield

curve is continuous at T the second derivative is not, and Rc is again convex for (T,∞ ).
To examine the curvature of the time path of future short rates, we derive the second
derivative of the path to be

d r c(t)/d t = 6(A - 4Bt), 0 ≤ t ≤ T

2 2

= 0, t>T .
Like the yield curve, the time path of short rates starts out convex, reaches an inflection point at

t = A/4B where it becomes concave until t = T.

+
Although both the time path and the yield
curve reach their inflection points midway between the starting point and the peak, the inflection

89
 Robert C. Merton
+ +
point of the time path occurs earlier than the inflection point for the yield curve (i.e., t < T ).
As a form of summary, the analysis shows in continuous time what Figure V.4 illustrated for the
discrete-time analysis: Namely, that the yield curve and the future time path of interest rates can
differ significantly.
In summary, the yield curve or term structure is a plot at a given point in time of a cross-
section of discount bond yield which differ only with respect to their maturities. Although
inherently a static construct, the yield curve derived from equilibrium bond prices in an
environment of certainty has an exact relationship to the dynamics or time path of future interest
rates. Even in an environment where future interest rates are uncertain, the term structure is still
well-defined. In such an environment, there will be a set of prices for discount bonds {Pt(τ)} at
each point in time, and by the definition of yield, these prices can be used in (V.8) to uniquely
determine a set of yields {RT} which can then be plotted against maturity to form a yield curve.
Of course, once future interest rates are stochastic, the relationship derived between current
prices and future interest rates, (V.6), will no longer be valid. However, with some additional
assumptions, the yield curve can still be used to make inferences about the structure of the
stochastic processes which describe interest rate dynamics. Moreover, as will be seen, the yield
curve frequently provides sufficient information to solve problems involving the pricing of fixed-
income securities.

On the Pricing of the General Default-Free Fixed Income Securities

In preceding analyses, we studied the price relationships among default-free, discount

bonds with the same promised payment ($1) at maturity. Consider now a general default-free,
fixed-income security with a schedule of promised payments of $xt to be paid at the end of
period t, t = 1,2,...,T. xt can be either positive in which case the owner of the security receives
a payment of $xt, or negative in which case the owner must pay out, $|xt|. We denote the
equilibrium market price of this security at time t = 0 by V0(x1,...,xT).

90
Finance Theory

If there exists a set of default-free discount bonds with current equilibrium prices denoted
as before by {P0(τ)}, then the current equilibrium price of the general default-free, fixed-income
security must satisfy
T

(V.18) V 0 ( x1 , x 2 ,..., xT ) = ∑ xτ P0 ( τ ).
τ =1

The proof that (V.18) must hold in equilibrium is by contradiction. Namely, if (V.18) does not
hold, then we will show that the general security either dominates or is dominated by other
available securities, and a necessary condition for equilibrium is that no such dominance exists.
T

Define δ by δ ≡ V 0( x 1 , x 2 ,..., x T ) - ∑ xτ P 0( τ ). Suppose that V0 were larger than

τ =1
T

∑
τ
xτ P ( τ )
=1
0 and hence, δ > 0. If an investor purchases the general security for V0 then he will

receive in return a stream of payments of $xt at the end of period t for periods t = 1,2,...,T.
Consider an alternative investment which calls for the purchase of a group of discount bonds in
the following quantities: Buy N τ ≡ [ xτ + δ /( P 0( τ )T)] bonds each of which pays $1 at its

maturity date τ periods from now and do this for bond maturities τ = 1,2,...,T. The cost of
T T

acquiring these bonds is ∑ N τ P0 ( τ ) = ∑ xτ P0 ( τ ) + δ

τ =1 τ =1
and hence, is the same as the cost of

the general security. Because each of the t-period maturity bonds purchased will pay $1 at date
t, the investor will receive a stream of payments of $Nt at the end of period t for periods t =
1,2,...,T. By hypothesis, δ > 0 and therefore, Nt > xt for t = 1,2,...,T. So, for the same initial
cost, the investor will receive each period a larger payment from the alternative investment than
he would receive from the general security. Hence, every investor would strictly prefer the
alternative investment to the general security. The general security is dominated by the
alternative investment, and therefore, the hypothesized condition is not consistent with
equilibrium pricing.

91
 Robert C. Merton
T

Suppose instead that V0 were smaller than ∑

τ
xτ P ( τ )
=1
0 and hence δ < 0. The entity

that issued the general security (e.g., an individual, firm, or financial institution) is required to
make a payment of $xt at the end of period t to the owner of the general security for periods
t = 1,2,...,T. Suppose the entity purchases the general security in the market and finances this

purchase by issuing Nτ discount bonds of maturity τ for τ = 1,2,...,T. As was shown, the total
T

proceeds from issuing these bonds ∑

τ
=1
Nτ P (τ )
0 is equal to V0 and hence, the total transaction

does not change the current cash position of the entity. However, the net resultant of the
transaction is to replace the entity's liability to pay $xt at the end of period t with a liability to
pay $Nt at the end of period t for t = 1,2,...,T. By hypothesis, δ < 0 and hence, Nt < xt for
each t. By making the transaction, the entity reduces the amount it has to pay in every period.
Therefore, as long as δ < 0, the entity can make itself better off by purchasing the general
security and financing its purchases by issuing the appropriate quantities of discount bonds.
From the viewpoint of an issuer, the general security (as a means of raising money) is dominated
by the alternative of issuing discount bonds. Of course, form the viewpoint of a buyer, the
general security dominates the specific package of discount bonds {Nτ}. Thus, the hypothesized
condition that δ < 0 is not consistent with equilibrium pricing, and this completes the proof that
(V.18) must obtain in equilibrium.

On Arbitrage Opportunities: A Special Case of Dominance

The requirement that prices be such that no investment dominates any other investment is
frequently called a "No-Arbitrage" (or "No-Easy Money") condition although the two are not
strictly the same. An arbitrage opportunity is said to exit if there is a set of feasible transactions
which require no cash payments at any time, and the resultant of these transactions is to produce
positive cash receipts at one or more points in time. In effect, the existence of an arbitrage
opportunity implies that it is possible to get something of value for nothing.

92
Finance Theory

A simple example of an arbitrage opportunity would be as follows: Suppose that shares

of General Motors stock were selling for $54 a share on the New York Stock Exchange while at
the same time, these shares were selling for $55 on the London Stock Exchange. An investor
who simultaneously sold k shares of GM on the London Exchange for a total of $55k and
bought k shares of GM on the New York Exchange for a total of $54k would immediately
produce a positive cash receipt of $55k - $54k = $k. By delivering the shares purchased in New
York to cover the shares sold in London, the investor would eliminate any further liabilities
associated with these transactions, and hence, this set of transactions requires no cash payments
by him at any time. However, as a result of these transactions, the investor has immediately
increased his wealth by $k. Indeed, as long as the contemporaneous prices for GM on the two
exchanges are different, the investor can continue to increase his wealth by making these
transactions. The investor is truly getting something for nothing. Just as the laws of
thermodynamics rule out the existence of a perpetual-motion machine, so the laws of economics
rule out the existence of persistent arbitrage opportunities.
As a second, somewhat more-complicated example of an arbitrage opportunity, we
reexamine the analysis used to derive (V.18) with the additional institutional assumption that at
least one investor can buy or issue (sell) any of the available securities in arbitrary amounts.
Consider the following set of transactions: buy xτ units of a τ-period discount bond for
maturities τ = 1,2,...,T and simultaneously, issue (or sell) one unit of the general security. Let k
denote the number of "units" of this "package" taken by an investor where k > 0 means "buy
k xτ units of the discount bonds τ = 1,2,...,T and issue k units of the general security" and k <

0 means "issue | k | xτ units of the discount bonds and buy | k | units of the general security."
⎡T ⎤
At the time that the transactions are made (t = 0), there is a cash outflow of $ ⎢∑ kxτ P0 ( τ )⎥
⎣ τ =1 ⎦

⎡ T
⎤
and a cash inflow of $kV0. Hence, $kδ ≡ $k ⎢V 0 - ∑ xτ P 0( τ )⎥ is the current net cash flow
⎣ τ =1 ⎦
to the investor. So, by choosing the sign of k such that kδ > 0, the investor receives an
immediate, positive cash payment of $kδ as a result of these transactions. Note that in period t

93
 Robert C. Merton

(t = 1,2,...,T), the investor receives $kxt from the discount bonds which mature in that period
and pays out $kxt on the k units of the general security issued. Hence, for any k chosen, the
net cash flows associated with the investment package are zero in every future period. Just as in
the first example, the result of these transactions is to immediately increase the investor's wealth
by $kδ. By assumption, the magnitude of k is not bounded. Therefore, as long as δ ≠ 0 , the
investor can continue to increase his wealth without bound. So, either the investor ends up with
all of society's wealth or the prices of the discount bonds and the general security change so that
δ = 0. Clearly, the latter is the sensible conclusion, and therefore, by his actions, the investor will
"force" prices to adjust until δ = 0. Thus, under the hypothesized institutional conditions, prices
must satisfy (V.18).
Although subtle, the differences between a dominance situation and an arbitrage
opportunity are important. The price conditions required to rule out dominance are formally the
same as the ones that rule out arbitrage opportunities, and the existence of an arbitrage
opportunity necessarily implies a dominance situation. However, the existence of a dominance
situation does not necessarily imply an arbitrage opportunity. To see this, consider the case
where δ > 0 and therefore, a collection of discount bonds dominates the general security.
Suppose this dominance situation is recognized by a specific investor. If the institutional
structure permits, he can and will enter into a set of arbitrage transactions, and by his actions in
the market, he will unilaterally force prices to adjust until δ = 0. However, suppose that this
investor owns none of the general security and further suppose that institutional restrictions
prevent him from issuing the general security. Then, at least for this investor, δ > 0 does not
provide an arbitrage opportunity because the set of transactions required to institute arbitrage is
not feasible. The only action that he can take is simply not to purchase any of the general
security, and this action provides little, if any, pressure on prices to adjust so that δ = 0.
Of course, if it is feasible for some other investor to issue the general security and if this
other investor recognizes that the dominance situation exists, then this other investor can perform
the arbitrage transaction and prices will adjust. Or, as described in the dominance proof of
(V.18), if the investors who own the general security recognize that the dominance situation
exists, then these investors will sell their holdings of the general security, and their collective

94
Finance Theory

actions will tend to force prices to adjust until δ = 0. Thus, the significant difference between
arbitrage and dominance is the mechanism by which such opportunities are eliminated. In the
case of arbitrage, it takes only one investor who recognizes the opportunity to force prices to
adjust until the opportunity is eliminated. In the case of dominance generally, several investors
with specific endowments must recognize the opportunity for the same price adjustment to
obtain. For this reason, price relationships derived from a "No-Arbitrage" condition are less
likely to be violated than ones derived form a "No-Dominance" condition. However, it should be
pointed out that the occurrence of a significant dominance situation is an infrequent event
although it will occur far more frequently than a true arbitrage opportunity.
As was the case in the arbitrage derivation of (V.18), most arbitrage opportunities can be
exploited only if the arbitrageur can sell securities that he does not own. While in that
derivation, the term "issue" was used to describe all such sales, it is usually only used to describe
the sale of a security whose obligations to the purchaser are those of the seller. For example, if
General Motors sells a fixed income security which obliges General Motors to make the
specified payments to the purchaser, then General Motors is said to have "issued" that fixed
income security. Such sales are called primary (market) offerings, and are rarely, if ever, made
by individuals. The purchase or sale of already-existing securities whose obligations are not
those of the seller is called a secondary (market) transaction, and most arbitrage transactions are
of this type. A secondary-market sale of a security not owned by the seller is called a short-sale.
A short-sale is accomplished by borrowing the security from someone who owns it and
then selling it in the market. The terms of the "loan agreement" are typically as follows: (1)
Like a standard demand loan, either the borrower (short-seller) or the lender can terminate the
loan at any time. At the time of termination, the borrower must return the security borrowed to
the lender by either purchasing the security in the market ("covering" his short) or borrowing the
security from another lender. (2) During the time that the security is borrowed, the borrower
must reimburse the lender for all payments (including interest, dividends, and other distributions)
that he would have received from the security had he not lent it to the borrower. (3) The
borrower may be required to post and maintain sufficient collateral to ensure his ability to meet
his obligations, (1) and (2), to the lender. Unlike a conventional money loan, the lender is not

95
 Robert C. Merton

paid interest for lending his security. Hence, the lender earns a return equal to the one he would
have received had he remained the owner and not lent the security. However, because he is no
longer the owner of record, he forgoes any non-cash benefits of ownership (e.g., voting rights)
while the security is on loan. Hence, for this and other inconveniences associated with lending
the security including the risk that the short-seller may not meet his obligations, the lender may
require some additional compensation. The usual form of the compensation is to require that at
least some of the collateral for the loan be cash which in effect, provides the lender with an
"interest-free" loan. Alternatively, the borrower may pay a fee or premium for the loan.
In summary, the short-sale is an important transaction for the exploitation of arbitrage
opportunities. Therefore, in institutional environments which prohibit short-sales, one must rely
on the weaker mechanism of dominance to ensure that price relationships such as (V.18) will
obtain. Fortunately, the actual institutional structure that exists permits most securities traded in
organized markets to be sold short.
Thus, especially in environments which permit short-sales, one would expect the price
relationship between pure discount bonds and general fixed-income default-free securities to
satisfy (V.18). From (V.8), we can rewrite (V.18) as
T

(V.19) V 0 = ∑ xτ /(1+ Rτ )τ .
τ =1

From (V.19), one can evaluate any default-free security using a properly-constructed yield curve.
While (V.19) looks like a present value formula, nowhere in either the dominance or arbitrage
derivation of (V.18) was it required that the future time path of interest rates be known with
certainty. Hence (V.18), and therefore (V.19), provide the proper equilibrium price relationships
even when interest rates are stochastic.
Having established the fundamental price relationship between default-free discount
bonds and default-free fixed-income securities, we now demonstrate its use in a number of
specific applications.

96
Finance Theory

On Coupon Bonds and Estimating the Term Structure of Interest Rates

As the analysis leading to (V.18) demonstrates, it is sufficient to have a complete set of

current discount bond prices to determine the equilibrium price of any default-free fixed-income
security. It was also shown that such a set is sufficient to construct the term structure of interest
rates and forward prices. However, while discount bonds are frequently issued with maturities of
less than one year, they are rarely issued with longer maturities, and this is the case not only for
government debt, but for corporate debt as well. Therefore, one cannot generate the term
structure by simply observing the contemporaneous prices of discount bonds for all maturities
because such an array of bonds does not exist. However, by using the current prices of the
default-free bonds which are available, it is possible to estimate both the "missing" discount bond
prices and the term structure.
The most common form for intermediate and longer-term debt is the coupon bond. Like
the "Interest-Only" loans discussed in Section II, the coupon bond calls for a stream of periodic
and equal-in-size (coupon) payments and a single, lump-sum (principal) payment at maturity. If
Cj denotes the coupon payment per period for periods 1,2,...,Tj and Mj denotes the principal
payment at the maturity date Tj, then, from (V.18), the equilibrium price of coupon bond #j, Bj
must satisfy
Tj

(V.20) B j = ∑ C j P0 (t) + M j P0 ( T j ) .
t=1

Equivalently, the price of the coupon bond can be written in terms of yields as
Tj

B j = ∑ C j /(1+ Rt ) + M j /(1 + RT j ) .
t Tj
(V.21)
t=1

Again, it should be emphasized that (V.20) and (V.21) must be satisfied in equilibrium even if
interest rates are stochastic. In the special case of nonstochastic interest rates and a "flat" term
structure, (V.21) reduces to (II.34) and can be rewritten as
T T
(V.22) B j = C j [1 - 1/(1 + r ) j ]/r + M j /(1 + r ) j
where r is the per period rate of interest common to all periods.

97
 Robert C. Merton

To estimate the missing discount bond prices using available coupon bond prices, we
proceed as follows: Suppose there are n coupon bonds numbered in ascending order with
respect to their maturities (i.e., T1, T2 <... < Tn ≡ T where T is the maximum maturity of any of
the bonds). In equilibrium, the prices of these bonds must satisfy (V.20), which can be written as
the system of equations
(V.23) B1 = C 1 P0 (1) + C 1 P0 (2) + ... + ( C 1 + M 1 ) P0 ( T 1 )

B 2 = C 2 P0 (1) + C 2 P0 (2) + ... + C 2 P0 ( T 1 ) + ... + ( C 2 + M 2 ) P0 ( T 2 )

B n = C n P0 (1) + C n P0 (2) + ... + C n P0 ( T 1 ) + ... + C n P0 ( T 2 ) + ... + ( C n + M n ) P0 ( T n ) .

Because we know the terms {Cj,Mj,Tj} and current prices {Bj} of the coupon bonds, (V.23)
can be viewed as a system of n linear equations for the T (unknown) discount bond prices
{P0(1),P0(2),...,P0(Tn)}. (V.23) can be rewritten in compact vector-matrix notation as

(V.24) B = AP

where B denotes a n x 1 vector of the coupon bond prices {B1,...,Bn}; P denotes a T x 1

vector of the discount bond prices; and A denotes a n x T matrix whose elements aij are: for i
= 1,...,n, aij = Ci, j = 1,2,...,(Ti - 1); aij = (Ci + Mi) for j = Ti; aij = 0 for j = Ti + 1,...,T.
Because (V.23) is a linear set of equations, there are well-established procedures for
solving it when a solution exists. If the number of equations is fewer than the number of
unknowns (i.e., n < T), then clearly, there will not be a unique solution because not enough
information is available. If n = T, then a unique solution will exist provided that the n
equations in (V.23) are linearly independent (i.e., the rank of the matrix A in (V.24) is equal to
T). Such linear independence will occur if the n bonds chosen are sufficiently different with
respect to their terms. In this case, the solution for P is obtained by matrix inversion
-1
(P = A B). Although matrix inversion is a difficult operation to do by hand, there exist very
efficient computer programs which solve these equations with little difficulty even when n is
quite large.

98
Finance Theory

Problem V.2: Using Coupon Bond Prices to Estimate Discount Bond Prices

There are three coupon bonds with the following terms and current prices: (B1 = $961,
C1 = $100, M1 = $900, T1 = 2), (B2 = $968, C2 = $100, M2 = $900, T2 = 3), and (B3 = $879,
C3 = $50, M3 = $950, T3 = 3). What are the implicit current discount bond prices for periods
1,2, and 3, and what are the corresponding term structure yields? From (V.23), we have that

(i) 961 = 100 P0(1) + 1000 P0(2) + 0 P0(3)

(ii) 968 = 100 P0(1) + 100 P0(2) + 1000 P0(3)
(liii) 879 = 50 P0(1) + 50 P0(2) + 1000 P0(3) .

If we multiply equation (iii) by 2 and subtract equation (ii) from it, then we find that 790 = 1000
P0(3) or P0(3) = $0.79. Substitute $0.79 for P0(3) in (ii) to get 100 P0(1) + 100 P0(2) = 178
and subtract this from equation (i). The resultant is 783 = 900 P0(2) or P0(2) = $0.87. Finally,
substitute $0.87 for P0(2) in equation (i) to get 100 P0(1) = 91 or P0(1) = $0.91. Having
solved for the discount bond prices, we now use (V.8) to determine the term structure: Namely,
R1 = 9.89%; R2 = 7.21%; and R3 = 8.17%. Finally, using these discount bond prices, we can
value payments to be made during the first three periods on any default-free security. For
example, what is the current value for a default-free security with a stream of payments, x1 =
$800; x2 = –$600; and x3 = $2500? Using the above prices in (V.18), we have that V0 = 800 ×
.91 – 600 × .87 + 2500 × .79 = $2181. The reader should note that the only data required in this
problem were the terms and current prices for the coupon bonds. Nowhere was it assumed that
interest rates were nonstochastic.
The central purpose of Problem V.2 was to illustrate how one can compute the discount
bond prices needed to use (V.18) and to construct the term structure. However, the analysis used
also illustrates how an investor can "manufacture" discount bonds when none exists provided
that short-selling is permitted. For example, suppose the situation is as in Problem V.2 and an
investor would like to have a three-period discount bond. Consider an investing "package"

99
 Robert C. Merton

where he buys 2k units of bond #3 and sells short k units of bond #2. At the end of period 1,
he will receive coupon payments of $100k on the bonds that he owns, and he must pay $100k
entity which lent him the bonds for short sale. Hence, the net cash flow from the investment
package at the end of period 1 is zero. By the same analysis, the net cash flow from the package
at the end of period 2 is zero. At the end of period 3, he will receive $2000k in coupon and
principal payments on the bonds he owns, but he must pay $100k in coupon payments and $900k
to repurchase the bonds he has shorted. Hence, the net cash flow from the investment package at
the end of period 3 is $1000k. Thus, the pattern of returns from this investment package is
identical to those of a three-period discount bond with a promised payment at maturity of
$1000k. k is simply a scale factor chosen by the investor in the same way that he would choose
the number of discount bonds he wants to purchase. The cost of the package would be 2k × 879
– 968k or $790k. Thus, the formal mathematical manipulations used to deduce implicit
discount prices are the same as the ones used to determine the combination of purchases and
short-sales required to "manufacture" discount bonds when such bonds do not exist.
To complete the analysis of (V.23), we now examine the case where the number of bonds
in the sample exceeds the number of maturities (i.e., n > T). In this case, there are more
equations than unknowns, and hence, for a solution to exist, the "extra" equations must be
redundant. Specifically, a unique solution will exist if and only if the number of linearly
independent equations in (V.23) equals T, and the solution for each such linearly independent
subset of equations satisfies the other (n - T) equations. In terms of (V.24), this condition
implies that both the row and column ranks of A be equal to T. If different linearly independent
subsets of the n equations lead to different values for the discount bond prices, then the row
rank will exceed the column rank, and no solution will exist. The economic implication of
nonexistence is that not all the coupon bond prices satisfy (V.18), and therefore, either a
dominance or an arbitrage situation exists among the outstanding coupon bonds. This point is
demonstrated in the following problem.

100
Finance Theory

Problem V.3: Arbitrage Opportunities in Coupon Bonds

Assume the same environment as described in Problem V.2, but now add one more bond
with a market price and terms given by (B4 = $4000, C4 = $300, M4 = $4000, T4 = 3). If
there are to be no arbitrage opportunities, then the price of bond #4 must satisfy (V.20). I.e.,

(iv) 4000 = 300 P0(1) + 300 P0(2) + 4300 P0(3) .

Therefore, the system of equations corresponding to (V.23) is (i), (ii), (iii) from Problem V.2 and
(iv). From the solution of Problem V.2, (i), (ii), and (iii) are a linearly independent subset of this
system with a solution {P0(1) = .91, P0(2) = .87, P0(3) = .79}. Thus, if there is a solution to
this system, (iv) must be a redundant equation satisfied by the solution to (i), (ii), and (iii). But,
300 × .91 + 300 × .87 × .79 = 3931 which is not equal to B4 = 4000. Indeed, based upon
the prices of the other three bonds, bond #4 is "overpriced" in the sense that at these prices, bond
#4 is dominated by the purchase of some combination of the other three bonds.
To show this dominance, let k4 denote the number of units of bond #4 either owned or to
be purchased by an investor and let kj denote the number of units of bond #j in the proposed
dominating investment package, j = 1,2,3. If the {kj} are selected so as to satisfy the
conditions: (a) 100 k1 + 100 k2 + 50 k3 = 300 k4; (b) 1000 k1 + 100 k2 + 50 k3 = 300 k4; (c)
1000 k2 + 1000 k3 = 4300 k4, then the cash receipts from the proposed package in periods 1, 2,
and 3 will be identical to the cash receipts from k4 units of bond #4 in those periods. These
conditions are satisfied by: k1 = 0; k2 = 1.7 k4; k3 = 2.6 k4. However, the cost of acquiring these
identical streams of payments is not the same. The cost of acquiring the k4 units of bond #4 is
$4000 k4. The cost of acquiring the "package" is $961 k1 + $968 k2 + $879 k3 = $3931 k4. Thus,
the package dominates bond #4. Of course, if the institutional structure permits short-sales, then
these prices would imply an arbitrage opportunity where the arbitrageur would purchase 1.7 k4
units of bond #2 and 2.6 k4 units of bond #3 for each k4 units of bond #4 sold short. For each
such transaction, his wealth would increase by $69 k4.

101
 Robert C. Merton

As in the case for (V.20), price relationships deduced from the condition that no
dominance situations exist are relative pricing formulas. I.e., they specify conditions under
which a set of prices will be internally consistent with respect to one another. Because only a
subset of securities are examined, a set of prices that satisfies such relative pricing formulas need
not be one which will clear markets in equilibrium. Therefore, relative pricing formulas provide
necessary, but not sufficient, conditions for equilibrium.
For example, in the solution of Problem V.3, it was shown that bond #4 was "overpriced"
relative to the prices of bonds #2 and #3 in the sense that at these prices, anyone would prefer an
appropriate mix of these bonds to holding bond #4. Hence, the posited prices in that problem
cannot be equilibrium prices. However, if the price of bond #4 were changed so as to be
consistent with the other three bond prices (i.e., B4 = 3931), there is not sufficient information
given in the problem to determine whether or not these prices would clear the market.
In summary, we have shown how discount bonds can be estimated using coupon bond
prices, and how these estimates can be used to identify mispriced securities. While there are
many (virtually) default-free coupon bonds traded in the market, their diversity in terms of
coupon, principal, and maturity is usually not sufficient to generate a unique set of discount bond
prices for all maturities. Moreover, differences in terms, other than those discussed here, can
also cause errors in the price estimates obtained from (V.23). Some examples would be
differences in sinking fund and call provision and the tax treatment of the returns earned from
holding the bond. The latter is especially important in the case of municipal bonds whose
coupon payments are usually exempt from federal income taxes. Hence, precise estimates for the
{P0(τ)} can rarely be made. However, statistical techniques can be applied to the structural
equations (V.23) to estimate both the prices and the precision of the estimates.

Yield-to-Maturity and Duration for Coupon Bonds

A frequently suggested alternative to (V.23) for estimating the term structure and
identifying mispriced default-free securities is the yield-to-maturity method. The yield-to-

102
Finance Theory
*
maturity for a coupon bond r is defined as that value of r which causes (V.22) to obtain for
the current market price of the bond. I.e., it is that common per period rate of interest which
would obtain if: (i) the bond price is an equilibrium price; (ii) the term structure is "flat"; (iii)
future interest rates are nonstochastic.

*
By manipulating (V.22), we have that the yield-to-maturity for coupon bond #j, r j is equal to (x
– 1) where x is the real-root solution to the polynomial equation

(V.25) 0 = B j xT j+1 - ( B j + C j ) xT j - M j x + ( M j + C j ) .

Certainly, the yield-to-maturity method appears to be an attractive alternative to (V.23) in terms

of data requirements and computational simplicity. Both methods require only current bond
prices and their terms to estimate the term structure. However, to identify RT using (V.23)
requires a minimum of two different coupon bonds whereas a single T-period bond can be used
to compute that bond's yield-to-maturity. Because Bj, Cj, and Mj are all positive, one can be
assured by Descartes' Rule of Sign that there is only one real-root solution to (V.25), and there
exist very fast and accurate numerical methods for finding the root of such a polynomial
equation. Moreover, unlike (V.23) which requires a simultaneous solution of a system of
equations, the yield-to-maturity equation can be solved separately for each bond. Undoubtably,
these attractive computational features provide the genesis of the standard practice of quoting
coupon bond prices as "priced to yield 100 r*j percent." There is no harm in such a convention
unless it is misused in the resolution of substantive issues. Specifically, these features of the
yield-to-maturity method are attractive only if it provides valid estimates for the term structure
and correctly identifies mispriced securities.
To determine the conditions under which this method does provide valid estimates, we
begin by examining pure discount bonds (i.e., Cj = 0). Inspection of (V.25) shows that the yield-

to-maturity on a pure discount bond is simply its yield R T j

as defined in (V.8). So, trivially,

the yield-to-maturity method applied to pure discount bonds provides a valid estimate for the
term structure. The yield method also works in identifying mispriced securities when comparing

103
 Robert C. Merton

pure discount bonds of the same maturity. However, for pure discount bonds with different
maturities, the one with the higher yield-to-maturity need not be the better buy. To illustrate this
point, consider the time pattern of interest rates and yields presented in Table V.1. By
construction, the discount bond prices and their yields displayed there are "fair" in the sense that
bonds of all maturities will have the same holding period returns. Yet, inspection of Table V.1
shows that the yield-to-maturity on a two-period bond R2 equals 3.5 percent while the yield on a
four-period bond R4 equals 6.2 percent. Indeed, if the price of a four-period bond were such
that its yield were 6 percent and other bond prices were unchanged, then for the same time
pattern of interest rates, the four-period bond would be dominated by an initial investment in the
two-period bond followed by a "rolling-over" of one-period bonds for periods three and four.
That is, by investing $1000 in the four-period bond when it is priced to yield 6 percent, the value
of the investment at the end of four periods would be $1262. However, by investing $1000 in the
alternative, the value at the end of four periods would be $1272.
In a similar fashion, it is straightforward to show that the yield-to-maturity method cannot
be used to identify the "better buy" when comparing coupon bonds with different maturities.
Moreover, as the following problem demonstrates, it cannot in general be used to compare
coupon bonds with the same maturity.

Problem V.4: Bond Swapping with Coupon Bonds

Investment strategies which attempt to improve the returns on a portfolio of fixed-income

securities by exchanging bonds currently in the portfolio for other bonds with the same maturity
and risk are called bond swapping strategies. Suppose that a portfolio of default-free fixed-
income securities contains bond #1 which is a 15-year bond with an annual coupon C1 = $50; a
principal or "face value" M1 = $4865; and a current market price of $1545. Suppose further that
bond #2 which is a 15-year bond with an annual coupon C2 = $200 and a face value M2 =
$1000 is currently selling for $1545. As the investment manager of this portfolio, should you
"swap" bond #1 for bond #2?

104
Finance Theory
*
Solving (V.25) for bond #1 and bond #2, we have that their yields-to-maturity are r1 =
*
.10 and r2 = .12 or 10 percent and 12 percent, respectively. Thus, if the manager uses the

yield-to-maturity method for selecting bonds, then he will choose to swap bond #1 for bond #2.
However, is this the correct decision?
To answer this question requires additional information. Suppose that the future time
path of interest rates is as described earlier in this section in Table V.1. Because the initial
investment required to acquire either bond is the same, clearly, the proper choice is the bond
which provides the larger cumulative increment to the value of the portfolio. In a similar fashion
to the derivation of the annuity formula in Section II, we can use the actual time path of interest
rates described in Table V.1 to determine the accumulated sum at the end of fifteen years,
j
S15 from holding bond #j to maturity and reinvesting all coupon payments received in the

interim. So, for example, the $50 received from bond #1 at the end of year 1 will be deposited
for year 2 at 5 percent and then reinvested along with accumulated interest at 10 percent for year
15

3, and so on. At the end of year 15, this payment will have grown to $50 ∏ (1 + r t ) . Doing
t= 2

this for all payments for both bonds, we find that the accumulated sum at the end of fifteen years
1 2
for bond #1 is S15 = $6127 and for bond #2, S15 = $6049. Therefore, bond #1 is actually the
better investment.
Hence, unlike the yield on pure discount bonds, the yield-to-maturity on coupon bonds
does not provide a ranking for comparing bonds of the same maturity. Moreover, while the yield
on a discount bond is always equal to the actual average compound return earned from holding
the bond to maturity, the yield-to-maturity on a coupon bond does not. To see this, note that
bond #2 is formally equivalent to a 15-year pure discount bond which pays $6049 at maturity and
has current market price of $1545. From (V.8), the average compound return on such a bond is
1/15
given by (6049/1545) – 1 or 9.5 percent which is significantly different from its 12 percent
yield-to-maturity. Thus, even if bond #2 had been the better investment, in general, its yield-to-
maturity will not be equal to the average compound return from holding it to maturity. This

105
 Robert C. Merton

significant discrepancy also demonstrates that the yield-to-maturity method as an alternative to

solving (V.23) can produce significant errors in estimating the term structure.
In summary, the yield-to-maturity method is not a reliable one either for making bond
swapping decisions or for estimating the term structure. The correct method is to estimate the
pure discount bond prices using (V.23) and then to evaluate individual coupon bonds using
(V.21).
The reason that the yield-to-maturity on a coupon bond fails to provide either the correct
return on the bond when held until maturity or the correct ranking of alternative bond
investments can be traced to the derivation of the present value formula in Section II. In that
derivation, it was essential that the interest rates used be the ones at which payments received
could be reinvested. Therefore, (V.22) from which the yield-to-maturity is derived, is a valid
present value formula only if the coupon payments received can be reinvested each period at rate
rj* . Hence, unless the reinvestment rates each period happen to equal rj* , the average compound

return from holding a coupon bond until maturity will not equal its yield-to-maturity. Indeed, as
was demonstrated by Problem II.2 in Section II, the choice among claims cannot in general be
made without reference to these reinvestment rates. Because the yield-to-maturity method makes
no reference to such reinvestment rates, it is perhaps not surprising that it cannot provide an
unambiguous ranking among investments.
While the preceding analysis provides essentially a negative report on the yield-to-
maturity method, it was pursued in detail because this method is frequently used and mis-used.
Also, in the study of corporate finance in Section VII, the same issues will arise again with
respect to the internal rate of return method for making capital budgeting decisions.
Two other yield terms frequently used in connection with coupon bonds are the coupon
+
rate and the current yield. The coupon rate r is defined to be the ratio of the coupon payment
+
per period to the principal or face value of the bond (e.g., for bond # j , r j ≡ C j / M j ). The

current yield r is defined to be the ratio of the coupon payment per period to the current
market price of the bond (e.g., for bond # j, r j ≠ C j/ B j ) . The current yield is the rate of

return that would be earned from holding the bond for one period if the price of the bond does

106
Finance Theory

not change. Using (V.22), the relationship among the coupon rate, the current yield, and the
yield-to-maturity for bond #j can be expressed by

(V.26) ( r*j - r j )/( r*j - r +j ) = ( r j / r +j )(1 + r*j )-T j

So, for example, the current yield will equal the yield-to-maturity if and only if either Tj = ∞
(i.e., the bond is a perpetuity) or Bj = Mj (i.e., the current yield equals the coupon rate). If
+ * + + * +
r j > r j , then r j > r j > r j , and if r j < r j , then r j < r j < r j . As with the yield-to-maturity,

neither the coupon rate nor the current yield is particularly useful for estimating the return from
holding a coupon bond.
All pure discount bonds with the same maturity date have the same time pattern of
payments: Namely, all payments are made on the maturity date. However, depending upon the
relationship between the relative size of the coupon and principal payments, the time pattern of
payments for coupon bonds with the same maturity date need not be the same. For example, in
Problem V.4, both bonds had fifteen-year maturities. However, bond #1 had a $200 per year
coupon and a $1000 principal whereas bond #2 had a $50 per year coupon with a $4865
principal. Thus, a $1545 investment in bond $1 would receive a larger fraction of its total
payments relatively earlier than the same investment in bond #2. The duration of a default-free
coupon bond is the (value-weighted) average time of the payments received on the bond and is
defined for bond #j by

Tj
D j ≡ ∑ tδ t
j
(V.27)
t=1

where δ t ≡ C j P 0(t)/ B j , for t = 1,2,...,T j - 1 and δ T j ≡ ( C j + M j )P0( T j )/ B j . Hence, δ t is

j j j

the fraction of bond #j's total value attributable to the payment received in period t, t = 1,...,Tj.
Because coupon bond #j is formally equivalent to a collection or portfolio of pure discount bonds
with $ δ t B j invested in pure discount bonds which mature at date t, Dj is equal to the (value-
j

weighted) average maturity of this portfolio of bonds. While duration does provide more
information about the time pattern of payments than the maturity, it does not appear to have

107
 Robert C. Merton

much operational importance for the evaluation of coupon bonds. For example, it is not true in
general that comparing two coupon bonds with the same durations, the one with the higher yield-
to-maturity is the better buy.
An alternative definition of duration sometimes used is
Tj

D j’ ≡ ∑ t γ t
j
(V.28)
t=1

where γ t ≡ C j(1+ r*j ) / B j , for t = 1,2,...,Tj – 1 and γ T j ≡ ( C j + M j )(1+ r*j ) j/ B j . This

j -t j -T

measure differs from the original because it replaces the actual present value of the individual
payments with the discounted value using the yield-to-maturity of the bond. Two attractive
features of this measure are: (1) it can be computed without knowledge of discount bond prices
{P0(t)}. (2) It does provide a measure of the sensitivity of the bond's price to a change in its
yield-to-maturity. Specifically,(V.29) ( r j / B j )∂ B j /∂ r j = - D j’ .
* *

Because it is equal to the (negative of the) price elasticity of a bond with respect to a change in its
yield-to-maturity D'j is sometimes used as a measure of the relative price variability of a

coupon bond with respect to a change in interest rates. However, it is not a reliable measure
because actual changes in interest rates do not affect the yield-to-maturity on different bonds in
the same way. That is, a bond with a longer duration than another bond may have a smaller
percentage price change in response to a change in interest rates because the effect of this change
on its yield-to-maturity may be smaller than the effect on the second bond's yield-to-maturity.
For a short, but excellent, discussion on the use and misuse of duration as a measure of price
variability, see Cox, Ingersoll, and Ross (1979).

Financial Intermediation and Interest Rate Spreads

We have assumed throughout this section that the fixed-income securities are traded in
organized markets. That is, to raise money, individuals or firms issue fixed-income securities
directly in the market, and to invest money, they purchase such securities directly in the market.

108
Finance Theory

However, there is an alternative to direct market participation: Namely, fixed-income securities

can be issued to or purchased from a financial intermediary. A financial intermediary is defined
as an economic organization whose principal function is to purchase financial securities and
finance these purchases by issuing financial securities. Probably the best-known type of financial
intermediary is a bank which makes ("purchases") loans and finances them by ("issuing")
deposits. Although the operational differences have become progressively less distinct, the
banking function is further specialized by both the type of loan made and the form of deposits
issued. For example, commercial banks specialize in making short-term loans to business firms
and individuals and finance them principally by demand deposits. Savings and loan associations
and mutual savings banks specialize in making long-term loans (principally mortgages) and
finance them by time deposits. Other examples of financial intermediaries are finance
companies, insurance companies, and investment companies.
Whether a specific financial security is best handled by a market or financial intermediary
will, or course, depend upon the relative costs associated with the two alternatives. The market
system works best when there are a large number of both buyers and sellers of the security
willing to transact in minimum lot sizes sufficient to cover the costs of maintaining a market. In
general, such securities would have to be of standard form and available in reasonably large
quantities. Further, information about the issuer which is relevant to the evaluation of the
securities would have to be available to a large number of potential participants at a reasonable
cost.
The advantages of using financial intermediation come when there are important
economies of scale. For example, by its geographical location, a bank may have significantly
lower costs in gathering information about the local real estate market than would a nonlocal
entity. Significant information asymmetries in general will favor financial intermediation. If the
economic lot size required to support a market is large, then the financial intermediary may
provide divisibility otherwise unavailable. It may also provide added flexibility over a market by
allowing nonstandardized contracts. As will be shown later, a financial intermediary may
provide more-efficient risk-spreading at a lower cost than could be achieved with markets alone.
In the light of these differences between the two alternatives, it is not surprising that virtually all

109
 Robert C. Merton

borrowing by individuals is done through financial intermediaries rather than by issuing claims
directly in the market and the majority of fixed-income securities held by individuals are claims
against financial intermediaries.
The no-arbitrage pricing formulas derived for default-free fixed-income securities imply
that there is a single interest rate for each period rt. In fact, it is not uncommon to find fixed-
income securities with the same maturities and promised payments that sell for different prices
and hence, different yields. These persistent differences in promised yields are called interest
rate spreads, and they occur both for fixed-income securities traded in markets and for similar
securities available through financial intermediaries. As mentioned earlier, one reason for these
differences is that the observed securities are not all default-free. When there are different
probability assessments of receiving the promised payments, then promised yields will be
different. A second reason for these differences is that the terms (other than the maturity and
promised payments) are different. Two examples would be differences in sinking fund and call
provisions.
A third reason is that the tax treatment of the returns earned on the securities is different.
Coupon bond payments on municipal bonds are exempt from Federal and (sometimes) State
income taxes. Returns earned from price appreciation on a bond are taxed at a different (capital
gains) rate than coupon payments. One series of US Government bonds (appropriately called
"flower bonds") provided a means for reducing Federal estate taxes. A more subtle example of a
tax difference occurs for demand deposits issued by banks. Instead of paying the market rate of
interest on such deposits, banks frequently provide the service and convenience of a checking
account at "no charge." Because interest income is taxable but service charges for a checking
account are in general not tax-deductible, the implicit interest received in the form of these "free
services" is, in effect, tax-free. In general, one must include not only the explicit cash payments
but also the value of any "payments in kind" when comparing returns on fixed-income securities.
A fourth reason for the differences is a difference in the transactions costs associated
with different fixed-income securities. Dealers and market-makers who provide the services of
an orderly market for trading these securities are compensated for these services by buying at one
price (the bid price) and selling at a higher price (the ask price). The average compensation per

110
Finance Theory

"round trip" trade is the difference between the ask price an the bid price which is called the bid-
ask spread. Because there are now two "prices" for a fixed-income security, two identical
securities could have different observed transaction prices if the last trade for one were a
purchase and the last trade for the other were a sale. If the marginal cost of making a market in
fixed-income security #j is higher than the marginal cost of making a market in fixed-income
security #i, then security #j is said to have lower marketability then security #i. Other things the
same, a security which has lower marketability will have a larger bid-ask spread. Because a
larger bid-ask spread implies a greater cost to the investor making transactions in that security,
equilibrium promised yields on less-marketable securities will be higher. Interest rate spreads
caused by costs are especially common for fixed-income securities available through financial
intermediaries as Problem V.5 illustrates.

Problem V.5: Interest Spreads on Consumer Installment Loans

A consumer goes to a bank to obtain a 36-month loan of $3000 to buy an automobile. As

is standard for such loans, the terms call for a series of equal monthly payments to repay the loan
with interest (i.e., it is an annuity type loan). It is given that future market interest rates are
nonstochastic and the term structure is "flat" at a level of 10 percent per year (i.e., 0.7974% per
month). The cost to the bank of closing the loan is $60 and the month cost to the bank of
servicing the loan is $0.65. If there is no chance that the consumer will default on the loan, what
is the smallest monthly payment required by the bank so that it would make the loan? What
would be the corresponding "quoted" interest rate on this loan?
Because the bank can always invest its funds at 10 percent per year in default-free fixed-
income securities, it will only make the loan on terms that will generate (at least) a 10 percent
return on its funds and cover all costs. If $x denotes the monthly payment by the consumer,
then the payment received by the bank per month net of servicing costs is $y ≡ $(x – .65). To
receive this stream of payments, the bank must initially pay out $3000 to the consumer and $60
in closing costs or a total of $3060. Because the reinvestment rate is the same each period, we

111
 Robert C. Merton

can use the present value formula for an annuity derived in Section II to determine the monthly
payments y which will generate a 10 percent annual return. Substituting r = .007974, N = 36,
and AN = 3060 in (II.19), we have that

y = r AN /[1 - 1/(1 + r )N ]
(V.30)
= $98.12 .

Therefore, the monthly payment made by the consumer x is equal to $98.77. As discussed in
Problem II.2, the practice is for the bank to quote the terms of the loan in the form of an annual
interest rate based upon the amount of money borrowed by the consumer. This "quoted" interest
*
rate is the yield-to-maturity r (annualized) on an annuity which pays $x per month for 36
*
months and has a present value of $3000. I.e., r is the solution to equation (V.30) where Bj =
*
$3000, Cj = x and Mj = 0. Solving this equation, we have that r = .009488 or .9488% per
* 12
month. The annualized interest rate implied by this monthly rate is given by (1 + r ) – 1 or 12
percent. Although the quoted rate is 200 basis points higher than the market rate (100 basis
points equals 1 percent on an interest rate), the bank only earns the market rate of 10 percent on
its funds. The difference in rates just covers the cost of creating and servicing the loan. Of
course, if there is a chance that the consumer will default on the loan, then the additional costs
associated with repossessing the automobile and selling it would have to be covered by the
promised monthly payments, and the spread between the quoted and market rates would be even
larger.
This completes our study of default-free fixed-income securities. In Sections II-V, we
have emphasized the intertemporal aspects of financial markets and instruments with little or no
explicit consideration of uncertainty. In Sections VI and VII, corporate investment decisions are
examined in such an environment. Beginning in Section VIII, the balance of this Volume will be
devoted to the role of uncertainty in financial theory.

112
 Robert C. Merton

Table III Summary of Selected Financial Instruments Markets

Obligation Secondary Maturities Denominations Volume Volatility (Range Quotation Basis Forward; Futures Market
Market over 2 years)
Treasury Bills U.S. Government Excellent 3 month $10,000 - $6 billion per wk. (high) (low) Discounted interest One-week forward market
Obligations secondary market. 6 month $15,000 In 3- and 6-month 9.37 4.76 based on actual days in sometimes available on bid
1 year $50,000 maturities 360-day year. side.
$100,000 - $3-5 billion per day
$500,000 in secondary market. Futures traded on
$1,000,000 International Monetary
Market.
U.S. Agency Paper Obligations of U.S. Good secondary 30 days to 40 $1,000 to $87 billion Varies with each Discounted or interest- 3-5 day forward market for
agencies established market years $100,000 outstanding in May, issue and maturity. bearing. Interest based some “when issued”
by Congressional acts 1976. on 30-day month or securities some trading for
360-day year. delayed delivery; standby
commitments available.

GNMA futures traded on

Chicago Board of Trade.
Prime Commercial Promissory notes of No secondary 30-270 days $5,000 to $5 million $50.5 billion 12.50 5.00 Discounted or No forward market.
Paper issuing companies market. $100,000 basic outstanding in occasionally interest
(industrial & trading unit. May, 1976 bearing. Based on No futures market.
financial) actual days in 360-day
year.
Certificates of Obligation of bank Good secondary 1-12 months $100,000 and up; $70.6 billion 12.66 5.56 Yield basis. Interest No forward market.
Deposit accepting the deposit market and occasionally $500,000 minimum outstanding in June, paid on actual days in
up to 18 months trading unit; 1976 in CDs of over 360-day year. Interest No futures market.
$1,000,000 most $100,000. & principal paid at
common trading unit. maturity.
Bankers Obligation of bank Good secondary 30-180 days. Issued in odd $19.5 billion 12.16 4.94 Discounted. Interest No forward market.
Acceptances against which draft is market. 90 days is most denominations; outstanding in May, based on 360-day year.
drawn and which common traded in $100,000 to 1976. No futures market.
accepts draft primary market $1,000,000 lots.
maturity.
Federal Funds Obligations of the No secondary Usually Negotiated among $26.9 billion— 13.55 4.73 Par basis—interest There is a forward market
bank borrowing market. overnight. participants, average weekly paid based on 360-day or one to two weeks.
funds generally $1 million volume for July, year. Interest &
units. 1976 principal paid at No futures market.
maturity.
Long Term & U.S. Government Limited secondary 1-10 year notes $10,000 minimum. $188 billion 3-5 year Price basis; quoted in No forward market.
Intermediate Term obligation market—good for and 10-40 year outstanding in June, 8.69 6.71 dollars per hundred
Government some short terms bonds. 1976. dollars face value. No futures market.
Securities but very thin for Long term
long terms

113
Finance Theory

114
VI. THE VALUE OF THE FIRM UNDER CERTAINTY

In Sections III and IV, it was shown that the maximization of the current value of the firm is
the appropriate primary objective for good management. It is therefore natural to begin the study of
corporate finance by developing first techniques for determining the value of the firm and then
examine how that market value is affected by these investment and financing decision variables
which are under management's control. In this section, valuation formulas are derived in a certainty
environment where the future cash flows of the firm are known, and in Section VII, the capital
budgeting or the firm's investment decision problem is studied within this same framework. In
Sections XIV and XV, the investment decision by firms is reexamined in the context of uncertainty.
Valuation techniques can be separated into three categories:

(i) Rules of thumb intended to facilitate comparisons of value among similar assets.

(ii) Approaches based upon the economic theory of market value under certainty.

(iii) Approaches which explicitly recognize uncertainty and take into account risk in a
market context.

While category (i) techniques are frequently used in practice (especially in security analysis),
they are less useful for corporate financial decisions because of the difficulty in determining the
impact of alternative management decisions on market value. Moreover, the use of such techniques
can be "dangerous" unless the user understands the set of implicit assumptions upon which their
valid application depends and the associated limits within which they can be relied upon.
As will be demonstrated, most Rules of Thumb are simplified abstractions of the techniques
in category (ii). Since these techniques also permit the analysis of alternative management decisions
on firm value, we begin the study of value with these techniques. However, the reader is warned
that because they assume certainty, the valuation formulas derived in category (ii) are themselves
significant abstractions, and care must be exercised in applying them in practice. Category (iii) to be
examined in Section XIV is the least abstract of the three and therefore, the most rigorous.
However, these techniques are also the most complicated and require more information and analysis
to implement. Which technique to use will depend upon the situation and the judgment of the
manager. One important purpose of this course is to help develop this judgment.

115
 Robert C. Merton

There are four basic approaches used to determine value in a certainty context:
(I) The value of the firm is the present value of the stream of dividends paid by the firm.
(The "Dividend-Discount" approach.)
(II) The value of the firm is the present value of the cash flows generated by the firm.
(The "Discounted Cash Flow" approach.)
(III) The value of the firm is the present value of the earnings generated by the firm. (The
"Discounted Earnings" approach.)
(IV) The value of the firm is the present value of earnings generated from assets currently
in place plus future investment opportunities. (The "Growth Opportunities"
approach.)

As an aid to the reader, a glossary of notation used in the analysis to follow is presented on
the next page.
To determine which (if any) of the four statements of values, I-IV, are correct, we start from
first principles. If Z(t) is the return per dollar from investing in the equity of the firm between time
t and t+1, then, by definition

d(t + 1) + S(t + 1)
(VI.1) Z(t) ≡
S(t)

where d(t+1) is the dividend per share paid at time (t+1) and S(t+1) is the price per share (ex-
dividend paid at time t+1). From the identity (VI.1), we derive a price restriction from the
(arbitrage) condition under certainty that all securities must yield the interest rate. I.e., that

d(t + 1) + S(t + 1)
(VI.2) 1 + r(t) = Z(t) ≡
S(t)

where r(t) is the one-period rate of interest from time t to t+1.

Consider a firm which will remain in business for T periods (from now) and then liquidate.
To deduce the value of the stock today, we first go forward in time and then work backwards to
today.

116
Finance Theory

At time T in the future, the firm will pay its last dividend, d(T), per share and the ex-
dividend price per share at that time will be the salvage value (per share) of the firm, SALV, which
is assumed to be paid out as either a liquidating dividend or return of capital.

Glossary of Notation

V(t) ≡ market value of the firm at time t

n(t) ≡ number of shares of the firm's stock outstanding at time t
S(t) ≡ price per share of stock at time t (ex-dividend paid at time t)
V(t) ≡ n(t)S(t) if the firm is all equity-financed
D(t) ≡ total dividends paid by the firm at time t
d(t) ≡ dividend per share = D(t)/n(t–1)
Z(t) ≡ return per dollar to the investor in the firm
REV(t) ≡ total revenues in period t = stream of cash receipts
O(t) ≡ total operating cash outflow in period t
π(t) ≡ after-tax profits in period t
DEP(t) ≡ depreciation in period t
CGS(t) ≡ cost of goods sold in period t
τ(t) ≡ taxes paid in period t
I(t) ≡ gross investment (both new and replacement) in period t
i(t) ≡ net (new) investment in period t
I(t) ≡ i(t) + DEP(t)
X(t) ≡ "gross" profit or net cash flow in period t
X(t) ≡ π(t) + DEP(t)
Pt(s) ≡ price of a default-free discount bond at time t which pays $1 at time t+s (i.e., s
periods in the future).
r(t) ≡ short-term, one-period riskless interest rate for period t

117
 Robert C. Merton

Except for some tax implications, we could assume that d(T) includes this payment in which case
SALV = 0. From (VI.2), we have that

d(T) + S(T) d(T) + SALV

(VI.3) 1 + r(T - 1) = Z(T - 1) = =
S(T - 1) S(T - 1)
or

d(T) +SALV
(VI.3') S(T - 1) = .
[1 + r(T - 1)]

Consider an investor who at time (T–2) is going to buy the stock; the total return in dollars
for holding one share for one period [from (T–2) to (T–1)] will be d(T–1) + S(T–1), and again to
avoid arbitrage, we have from (VI.2) that

d(T - 1) + S(T - 1)
(VI.4) 1 + r(T - 2) = Z(T - 2) =
S(T - 2)
or

d(T - 1) + S(T - 1)
(VI.4') S(T - 2) =
[1 + r(T - 2)]

Substituting for S(T–1) from (VI.3') into (VI.4'), we have that

d(T - 1) d(T) + SALV

(VI.5) S(T - 2) = + .
[1 + r(T - 2)] [1 + r(T - 2)][1 + r(T - 1)]

At time (T–3) from now, we have from (VI.2) that

d(T - 2) + S(T - 2)
(VI.6) 1 + r (T - 3) =
S(T - 3)

or from (VI.5) and (VI.6),

118
Finance Theory

d(T - 2) + S(T - 2)
S(T - 3) =
[1 + r(T - 3)]
d(T - 2) d(T - 1)
(VI.6') = +
[1 + r(T - 3)] [1 + r(T - 3)][1 + r(T - 2)]
d(T) +SALV
+
[1 + r(T - 3)][1 + r(T - 2)][1 + r(T - 1)]

Proceeding inductively in this backwards fashion, we arrive at the price per share today (time zero)
which ensures that an investor buying the stock at any time and selling at any other time will earn a
fair return and no arbitrage opportunities will be created. I.e.,

d(1) d(2) d(T) + SALV

S(0) = + + ...+
[1+ r(0)] [1+ r(0)][1+ r(1)] [1+ r(0)]...[1+ r(T - 1)]
T
d(t) SALV
S(0)= ∑ t
+ T
t=1
( ∏[1+ r(s - 1)]) ∏[1+ r(s - 1)]
(VI.7) s=1 s=1
T
d(t) SALV
=∑ t
+
t=1 [1+ R(t) ] [1+ R(T) ] T
T
= ∑ P0(t)d(t) + P0(T) SALV
t=1

where R(t) and P0(t) are as defined in Section V. Indeed, (VI.7) follows directly as a special case
of valuation formulas (V.18) and (V.19) in Section V.
If it is assumed that the firm is financed entirely by equity (of a single-homogeneous class)1,
then we have that the current market value of the firm is V(0) ≡ n(0)S(0) where n(0) is the number
of shares currently outstanding. From (VI.7), we have

T
n(0)d(t) n(0)SALV
(VI.8) V(0) ≡ ∑ t
+
t=1 [1+ R(t) ] [1+ R(T) ] T
for notational convenience, we rewrite (VI.7) and (VI.8) as

1
An assumption maintained until we reach Section IX.

119
 Robert C. Merton

∞
d(t)
(VI.7') S(0) = ∑
t =1 [1 + R(t) ] t
and
∞
d(t)n(0)
(VI.8') V(0) = ∑
t =1 [1 + R(t) ] t

where it is understood that a finite-lived firm will have d(t) = 0 for t > T, and any
salvage value is incorporated in d(T).
Returning to our four approaches to valuation: From (VI.8'), (I) is valid provided that it is
more carefully stated to say that the current market value of the firm is equal to the present
(discounted) value of the stream of dividends paid by the firm to the current shares outstanding.
Total dividends paid by the firm at time t, D(t), are equal to n(t–1)d(t).2
∞
D(t) ≡
So, in general, V(0) ≠ ∑ t
unless n(t) n(0). I.e., unless the firm neither issues any
t =1 [1 + R(t) ] t

new shares (to raise additional capital) nor purchases any shares ("share repurchase") for treasury
stock.
Note: Even though the value of the firm (or an individual share) is written as the present
value of future dividends, the investor will earn the market return on his investment over any sub-
period of time even if no dividends are paid during that time. E.g., if d(3) = 0, then the return from
S(3)
period 2 to period 3, will be ≡ Z(2) = 1 + r(2).
S(2)
To work out the dynamics of how an all-equity-financed firm's value changes through time,
it is important to distinguish between the change in an individual investor's wealth from the return
earned by the firm and the change in the firm's total value. By definition,
V(t +1) ≡ n(t +1)S(t +1)= n(t)S(t)+ [n(t +1)- n(t)]S(t +1). Substituting (partly) for S(t+1) form

(VI.2), we have that

2
Note it is "n(t–1)" because we have assumed that dividends paid at time t go only to shares
outstanding as of time (t–1).

120
Finance Theory

V(t + 1) = n(t)[(1+ r(t))S(t) - d(t + 1)] + m(t + 1)S(t + 1)

(VI.9)
= (1+ r(t))V(t) - D(t + 1) + m(t + 1)S(t + 1)
where m(t+1) ≡ n(t+1) – n(t) = number of new shares issued by the firm at the (ex-dividend) price
S(t+1). [If m(t+1) < 0, then this corresponds in absolute value to the number of shares purchased
by the firm from shareholders.] From (VI.9), the change in firm value, ∆V ≡ V(t+1) - V(t), can be
written as

(VI.10) ∆V = [r(t)V(t) - D(t + 1)] + [m(t + 1)S(t + 1)]

= r(t)V(t) + {m(t+1)S(t+1)–D(t+1)}
_______________ ____________________
Total change in Net new financing by
shareholder's the firm or net new
wealth from capital raised by the
investing in firm
company

So, for example, AT&T could have a beginning-of-the-year market value of $20 billion; investors
could average a 10% return for the year (or $2 billion); AT&T could pay out dividends (and interest)
of $1.5 billion; and it could issue $4 billion worth of new shares (and debt). From (VI.10), the
change in firm value would be ∆V = $2 billion + {$4-1.5}billion = $4.5 billion while the net gain to
investors would be $2 billion. Hence, the change in the market value of the firm can be larger,
smaller, or equal to the change in shareholders wealth. From (VI.9), we also have

1
(VI.11) V(t) = { V(t + 1) + D(t + 1) - m(t + 1)S(t + 1) } .
[1 + r(t)]

Having established the validity of the stream-of-dividends approach, what can be said about
the other three methods? Using REV(t+1) to denote the total revenues [or stream of cash receipts
during period (t to t+1)] and O(t+1) to denote total cash (operating) outflow, the basic cash flow
accounting identity can be written as
(VI.12) REV(t+1) + m(t+1)S(t+1) = O(t+1) + D(t+1)
121
 Robert C. Merton

Total cash inflow Total cash outflow

REV(t) and O(t) can be expressed in terms of their component parts as

REV (t ) = π (t ) + τ (t ) + CGS (t ) + DEP(t )

and
O(t ) = I (t ) + τ (t ) + CGS (t )
where the terms are defined in the glossary.
Moreover, from the accounting identity (VI.12), we have that

(VI.13a) D(t + 1) − m(t + 1) S (t + 1) = REV (t + 1) − O(t + 1)

(VI.13b) D (t + 1) − m(t + 1) S (t + 1) = π (t + 1) + DEP (t + 1) − I ( t + 1) = X (t + 1) − I (t + 1)

(VI.13c) D(t + 1) - m(t + 1)S(t + 1) = π (t + 1) - i(t + 1)

From (VI.11c) and (VI.13a) - (VI.13c), we have that

1
(VI.14a) V(t) = {V(t +1)+ REV(t +1) - O(t +1)}
[1+ r(t)]
1
(VI.14b) V(t) = {V(t + 1) + X(t + 1) - I(t + 1)}
[1 + r(t)]
1
(VI.14c) V(t) = {V(t + 1) + π (t + 1) - i(t + 1)}.
[1 + r(t)]

Let V(T) denote the value of the firm at time T in the future; then, by employing the same
backward technique used in deducing the value of a share of stock in (VI.7), we can work backward
to solve (VI.14a) - (VI.14c) which can be rewritten as
T
[REV(t) - O(t)] V(T)
(VI.15a) V(0) = ∑ t
+
t=1 [1+ R(t) ] [1+ R(T) ] T
122
Finance Theory

T
[X(t) - I(t)] V(T)
(VI.15b) V(0) = ∑ +
t=1 [1 + R(t) ] [1 + R(T) ] T
t

T
[ π (t) - i(t)] V(T)
(VI.15c) V(0) = ∑ +
t =1 [1 + R(t) ] [1 + R(T) ] T
t

lim V(T)
Provided that { } = 0, we can rewrite (VI.15a) - (VI.15c) for a firm that is
T → ∞ [1 + R(T) ] T

going to continue indefinitely as

∞
[REV(t) - O(t)]
(VI.15a') V(0) = ∑
t=1 [1+ R(t)] t

∞
[X(t) - I(t)]
(VI.15b') V(0) = ∑
t =1 [1 + R(t) ] t

∞
[ π (t) - i(t)]
(VI.15c') V(0) = ∑
t =1 [1 + R(t) ] t

From (VI.15a'), we see that the value of the firm can be written as the present value of the cash flows
generated by the firm and hence, approach (II) is a valid description of value.

From (VI.15'), it is not a valid claim that the current value of the firm can be written as the
∞
π (t)
present discounted value of future earnings (i.e., V(0) ≠ ∑ [1 + R(t) ]
t =1
t
) because in general, to

generate a specific earnings flow, it is necessary to make capital expenditures in the future.
Working in net terms, if it is necessary to make (net new) investment expenditures i(t) in the ith
period, then the present value of this opportunity cost is i(t)/[1+R(t)]t. Summing over all t, we get
∞
i(t)
the total additional cost required to generate the stream of earnings {π(t)} to be ∑ [1 + R(t) ]
t =1
t
.

123
 Robert C. Merton

Subtracting these costs from the present value of the earnings will give the value of the firm which
is verified in (VI.15b') in gross terms or in (VI.15c') in net terms.
∞
π (t)
Note: ∑ [1 + R(t) ]
t =1
t
may either overstate or understate the correct market value because

i(t) can be negative or positive although i(t) ≥ – DEP(t).

In a contracting industry, one might expect to find that gross investment may not be as large
as required for replacement (i.e., i(t) < 0), and hence, capacity would decline over time. In a stable
or stagnant industry, gross investment might just match replacement requirements (i.e.,
i(t) ≈ 0), and capacity would remain about constant over time. In an expanding industry, gross
investment would probably exceed replacement requirements (i.e., i (t) > 0), and capacity would
∞
π (t)
increase over time. So, unless the economy as a whole is stagnant, ∑ [1+ R(t)] t will be a biased
t=1

estimate for market value. Nonetheless, approach (III) is valid if interpreted in the sense of
(VI.15c').
Although equivalent to the other three statements of value, the "current earnings and future
investment opportunities" approach (IV) is probably the most interesting. It is an especially useful
form for the investor planning to invest in the firm, and is the most natural approach for a (single)
owner planning to take over the firm.
To avoid notational complexities, let us assume that r(t) = r and R(t)= r , for all t (i.e., a
"flat" term structure).
To determine the value of the firm, the take-over investor considers three things:

(1) the competitive ("normal" or "alternative") rate he can make in the market which is r.
(2) the earnings potential of the existing assets of the firm.
(3) the opportunities (if any) for the firm to invest in real assets that will yield more than the
competitive rate of return. (Due to special advantages of the firm.)

Clearly, the take-over investor is not concerned with dividends patterns because he can choose any
pattern he wishes.

124
Finance Theory

To evaluate the earnings potential of the existing physical (tangible) assets, one can use the
regular discounted cash flow formula. Using the annuity formula (Section II, formula (II.16)), we
can compute an equivalent perpetual constant flow. Call it X ce , and the value of the firm's tangible

assets is
X ce .
r
To evaluate the (intangible) assets associated with future investment opportunities, first,
consider those projects beginning in period t requiring investment in that period of I(t). Second,
using the discounted cash flow method, evaluate the projects as of date t . Third, using the annuity
formula, convert this value into an equivalent perpetual annuity with a constant flow of F(t)
F(t)
dollars at the end of each year. Define: r*(t) ≡ . Then r* (t) is the average rate of return per
I(t)
period on projects taken in period t (and F(t) is the (equivalent) dollar return on projects
undertaken in period t). The present value at the beginning of period t of these projects is
F(t) r* (t)I(t)
= .
r r
Note: r*(t) is an average rate of return. Since one can always earn at least r by buying
market securities, one would not (voluntarily) take investments yielding less than r . Therefore

r*(t) ≥ r.
* *
(t)I(t) r (t) - r
The "goodwill" difference between worth and cost is r - I(t) = I(t)[ ], and the
r r
*
present value of this "goodwill" is r (t) - r -t
I(t)[ ](1 + r ) ; and the present value of all such
r
∞

∑ r* (t) - r
1
I(t)[ ].
(1+ r) r
"goodwill" for all the future is t=1

The current value of the firm will be the sum of the value of current assets plus the current value of
"goodwill." I.e.,

125
 Robert C. Merton

∞ *
1 r (t) - r ].
(VI.16) V(0) = X ce + ∑ t
I(t)[
r t=1 (1 + r ) r

To show that (VI.16) is equivalent to the other formulations for value, note that by definition of
*
r (t ) and X ce,

X (1) = X ce
X (2) = X ce + r *(1) I (1)
X (3) = X ce + r *(1) I (1) + r *(2) I (2) = X (2) + r *(2) I (2)
t -1
X (t ) = X (t -1) + r *(t -1) I (t -1) = X ce + ∑r *(s) I (s).
s =1

From (VI.15b'), we have that

∞
[ X (t ) − I (t )] X (1) − I (1) ∞ [ X (t ) − I (t )]
V (0) = ∑ = +∑
t =1 [1 + r ]t (1 + r ) t =2 (1 + r )t

X ce − I (1) ∞ 1 t −1
= +∑ [ X ce + ∑ r *( s ) I ( s ) − I (t )]
(1 + r ) t = 2 (1 + r )
t
s =1

X ce ∞ t −1 r *( s ) I ( s ) ∞ I ((t )
V (0) = + ∑∑ −∑
r t = 2 s =1 (1 + r )t t =1 (1 + r )
t

Note:

∞ t -1 * ∞ ∞
( s) I ( s) 1 1
∑ ∑ r (1 + r )
t = 2 s =1
t
= r *(1) I (1) ∑
k =2 (1 + r ) k
+ r
*
(2) I (2) ∑
k =3 (1 + r )
k
+ ...
∞
1
+ r *( s ) I ( s ) ∑
k = s +1 (1 + r )
k
+ ...
∞ ∞ ∞
1 1 1 1 1 1 1
∑
k = s +1 (1 + r )
k
=
(1 + r ) s
∑
k = s +1 (1 + r )
k -s
= s ∑
= •
(1 + r ) j =1 (1 + r ) r (1 + r ) s
j

So,

∞ t −1
r *( s ) I ( s ) ∞ r *( s ) I ( s )
∑∑
t = 2 s =1 (1 + r )t
=∑
s =1 r (1 + r )
s
or

126
Finance Theory

∞ t -1 * ∞ *
r (s) I ( s) r (s) I (s)
∑ ∑
t = 2 s =1 (1 + r )
t
= ∑
s =1 r (1 + r )
s
or

∞ *
V (0) = X ce + ∑
1
I (t )[ r (t ) - r ] which is (VI .16).//
t =1 (1 + r )
t
r r
____________ __________________________
Value of
tangible assets Value of future opportunities

Inspection of (VI.16) demonstrates two important points: (1) a firm can have positive value without
any physical assets; (2) the current value of the firm will only be affected by future investment
opportunities if those opportunities have rates of return on physical assets that exceed the market
rate.
To summarize, the four major approaches to valuation (appropriately interpreted) are equally
valid, and in fact, equivalent. Because each follows from the other using the basic accounting
identity, none is "more primal" than any other. Which one uses is more a matter of convenience.
The above analysis is precise and without controversy. If the world were certain, then there
would be nothing more to do in terms of valuation formulas. However, the future is uncertain, and
the impact of uncertainty on valuation is non-trivial. As will be shown in Section XIV, it is possible
to develop precise valuation formulas under uncertainty, but these (or the underlying assumptions)
are subject to controversy.
Before going into the use of these formulas in the firm's investment decisions, we conclude
this section with a brief discussion of "growth" and "glamour" stocks which gives a precise
definition for such stocks and may clear up some misconceptions about what growth stocks are.

Growth Stocks

127
 Robert C. Merton

A common rule of thumb used to value a firm is to compute the average or "normal" price-
to-earnings ratio ("PE") for companies within the same industry (or risk class), and to estimate the
value of the firm by assuming that it should have the same price-to-earnings ratio. I.e., if there are
n firms in the industry, and π i (0) and V i (0) denote current earnings and firm value for the ith

V i(0) 1 n
firm, then PE i ≡ and the industry average is PE ≡ ∑ PE i . To find the "fair value" for
π i(0) n i=1

the firm, we set V(0) = ( PE ) • π (0). A variate of this "quick-and-dirty" method is to compute the

1 n 1 n π i(0) π (0)
average earnings-to-price ratio, EP = ∑
n 1
EP i ≡ ∑
n i=1 V i(0)
and set V(0) =
EP
. The crude

justification for this method goes as follows: Earnings are what is available to shareholders; if the
current earnings of the firm are reasonable estimates of future earnings and if the firm is expected
to remain in business indefinitely, then the shareholders can reasonably expect to receive in
payment (or in equivalent increase in share value) π (0) each period indefinitely. Viewed in this
light, ownership of shares is essentially the same as owning a perpetual annuity or consol bond
(page II-17, formula (II.18)). From page II-17, if A∞ is the value of such an annuity; y is the
y
payment per period; k = required rate of return (or discount rate) on the stream, then A∞ = .
k
Applying the analog to the ith firm, we have that
π i(0) = V i(0) 1
EP i ≡ k i and = PE i = . If it is assumed that, on average, the required return on
V i(0) π i(0) ki

all firms in the same industry (or risk class) is the same, then k ≡ EP is the required return for the
π (0) π (0)
stream of the firm, and therefore, V(0) should equal = .
k EP
In the light of our previous analysis, is there a rationale for this approach, and if so, under
what conditions is it valid?
From the analysis of the stream-of-earnings approach to valuation, we found ((VI.15c')) that
∞
π (t) - i(t)
the firm's value could be written as V(0) = ∑ . Under what conditions will
t=1 [1 + R(t) ] t

128
Finance Theory

π (0)
V(0) = where k is the return per period on the firm? If current earnings are an estimate of
k
≡
future earnings ( π (t) π (0)) and the firm requires no net new investment to generate these
t
≡
earnings (i.e., i(t) 0) and if the term structure is essentially "flat" (at least, in "real" terms), then
t
π (0)
V(0) = where r is the per period rate of return. In simple terms, if the firm's investment
r
(gross) is essentially matched by depreciation (i.e., maintenance); if current earnings are
representative of future earnings; if the reinvestment rate is reasonably constant, then the rule of
thumb method is reasonable.
Further, if one interprets the earnings used in computing the PE or EP as a kind of "long-
term average earnings," then even for cyclical-type companies, the rule of thumb has some validity.
Thus, by using the discounted cash flow method to convert the flows from current tangible assets

π (0)
to an equivalent equal annual flow { π (0)}, then provided the other conditions hold, V(0) = .
r
In early empirical studies, when earnings were "smoothed" in this fashion to eliminate transient
earnings, the PE rule of thumb was at times a reasonably good forecaster of value except for an
important subset of stocks.
These stocks had unusually high price-to-earnings ratios, and such stocks have traditionally
been identified as "glamour" or "growth" stocks.

Note: It should be noted that the type of stocks in this category have a consistently higher-
V(0)
than-average PE ratio. I.e., many stocks may have current PE ratios, , which are very
π (0)
V(0)
high (or even undefined, if π (0) ≤ 0), but have long-run which are not "out-of-line." Such
π (0)

129
 Robert C. Merton

stocks are not what is meant by growth or glamour stocks. Rather growth stocks are stocks with
V(0)
that are unusually high persistently.
π (0)
In earlier times, such stocks were thought to be outside the traditional mode of analysis and
1
were simply excluded from such discussions. The problem was that EP = was treated as an
PE
estimate for the required rate of return k, and often, when "normal" companies would have PE's
of 10 or 12 implying a k = 10% - 8.5%, growth stocks would have PE's of 25-50 implying a
k = 4% - 2% which would often be below the riskless rate r. Did this mean that investors in such
stocks were "fools," and that investment in such stocks could only be justified on the basis that
somehow investors would be willing to pay even more for them later, independent of the "rational,"
implicitly-low return?
An alternative, rational explanation can be found in the previous analysis of growth
V(0)
opportunities. From (VI.16), we can rewrite the expression for PE = (noting that
π (0)

π (0) = X ce ) as

V(0) 1 1 ∞ 1 *
r (t) - r) ]
PE = = + ∑
π (0) r π (0) t=1 (1 + r )t
I(t)[
r
(VI.17)
1
> if r* (t) > r.
r

Thus, if a company has some special advantages (e.g., patents, superior distribution capabilities,
monopoly, etc.) so that it can reasonably be expected to find investment opportunities in the future
which yield (non-competitive) rates of return r* (t) which exceed the market required return r,
then from (VI.17), it is quite rational to bid the price of the firm beyond the normal PE associated
with the profits generated by the assets currently in place. Moreover, from the derivation of
(VI.16), the investor will earn a rate of return r on such investments. The magnitude of the
difference between the PE ratio and 1/r will depend on the size and number of investment
opportunities that have returns that exceed r (i.e., I(t)) and the size of the spread between r* (t)
(the average return on projects) and r (the required rate of return by investors).
130
Finance Theory

Examples of companies that have at times been termed "growth" stocks are IBM, Coca Cola,
Polaroid, Xerox, and several drug companies.

Identification of "Growth" Stocks

Questions for thought

(Q.1) Is a company whose total asset size and earnings are growing over time (in a steady trend) a
growth stock?
(Q.2) Is a company whose earnings per share are growing over time a growth stock?
(Q.3) Is a company whose earnings per share are growing over time at a rate less than the required
rate of return r not a growth stock?
(Q.4) Is a company whose earnings per share are growing over time at a rate greater than the
required rate of return r a growth stock?
(Q.5) As an investor, what are the main questions you would want answered in deciding whether a
firm was a growth company or not?

Example: The Constant-Growth Case

Although not exactly empirically relevant, the constant-growth case displays some of the
qualitative characteristics of growth stocks. Assume that the firm does have investments such that
* *
r (t) > r and further that the average return on investments per period, r (t), is the same in every
≡ *
period. I.e., r*(t) r > r. From the previous analysis, we have that
t

X(t) = X(t - 1) + r* I(t - 1) ; X(1) = π (0) . Assume that the firm's investment policy leads to a total
investment each period which is a constant fraction of that period's gross earnings, i.e., I(t) = δX(t)

where 0 ≤ δ ≤ 1 . Then,

131
 Robert C. Merton

X(t) = X(t - 1) + r* δX(t - 1) = X(t - 1)[1 + r* δ ] = π (0)[1 + r* δ ] t -1 and

X(t) - X(t - 1) *
= r δ = rate of growth of earnings. Substituting for
X(t - 1)

I(t) = δX(t) = δ π (0)[1 + r* δ ] t -1 into (VI.16), we have that

∞
⎛ r * − r ⎞ δ π (0)[1+ r*δ ] t-1
π (0)
+∑⎜
⎜ r ⎟⎟
V(0) =
r t=1 ⎝ ⎠ [1+ r ] t
π (0) ⎡ δ (r * − r) ∞ 1+ r*δ t-1⎤
= ⎢1+
r ⎣
∑[
(1+ r) t=1 1+ r
] ⎥
⎦

1+ r*δ
From p. II-15, (y ≡ )
1+ r

∞ ∞
y∞ - 1 1 ⎡ 1+ 2 ⎤
∑ = y t −1 = ∑ y j = = =⎢
y - 1 1- y ⎣ r - r* δ ⎥⎦
, provided that r > δ r* .
t=1 j=0

So,

π (0) ⎡ δ ( r* - r) 1+ r ⎤
V(0) = ⎢1+ • ⎥
r ⎣ (1+ r) r − r *δ ⎦

or

(1- δ )π (0)
V(0) =
r - r*δ
(VI.18)
(1- δ )π (0)
= where g ≡ r*δ = rate of growth of earnings.
r-g

D(0)
Note that D(0) = (1 - δ )π (0) , and therefore, V(0) = , a version of the "dividend-discount"
r-g
model.

Question: What is the interpretation of the g ≥ r case?

132
Finance Theory

What is the model telling you?

133
VII. THE FIRM'S INVESTMENT DECISION UNDER CERTAINTY: CAPITAL
BUDGETING AND RANKING OF NEW INVESTMENT PROJECTS

The most important decisions for a firm's management are its investment decisions.
While it is surely possible to get the firm into "trouble" through poor financing decisions or
improper management of working capital, the value of the firm is principally determined by the
prospects for its investments. Investments by the firm take two forms: (i) internally-generated
projects which, if undertaken, create new assets; and (ii) the acquisition of external already-
existing assets from other firms by either direct purchase of the assets or the acquisition of the
whole firm by merger, consolidation, or takeover.
Mergers and acquisitions are important topics for financial management and will be
discussed in Section XV. However, with the exception of a few specialized firms, the primary
function of the business firm is to find and undertake profitable new projects, and it is this form
of investment which is the topic of this section. The capital budgeting problem is how to select
those physical investments or projects so as to maximize the value of the firm. Much of the
formal apparatus has already been developed in Sections II, VI, and to some extent in Section V.
However, to put these tools in a more specific framework, we examine the various traditional
capital budgeting methods used to evaluate projects.
Before proceeding, we begin with some definitions:
A project is defined by the series of net cash flows it generates at the end of each period,
{X(1),X(2),...,X(N)}. These flows {X(t)} can be either positive or negative. If X(t) is positive,
then the project provides a net flow of cash into the firm at the end of period t, and if X(t) is
negative, then it causes a new flow of cash out of the firm. Since most projects require an initial
outflow, it is a common convention to denote this flow by "–I0" where I0 is the (positive)
outflow or initial investment in the project. For symmetry, we will also denote –I0 by "X(0)",
the net cash flow at the end of the "zeroth" period (or the beginning of the first period).

X(t) = [Revenues–Costs–Depreciation] × (1 – tax rate) + Depreciation –

Investment (in the project)
134
 Robert C. Merton

= After-tax Operating Profits – net new investment (in the project)

Let k denote the cost of capital to the firm (measured in percent per period) where the
cost of capital is the (external) rate of return required by investors for providing funds to the firm
and it reflects all the market opportunities available to investors. In a world of certainty (which is
the formal setting for this section), the cost of capital is simply the market rate of interest, r.
However, we follow tradition of using "k" rather than "r" to include the possibility in a quasi-
uncertainty sense (made rigorous in Section XIV) that different risk projects will have different
required returns (and in particular, required rates different from the riskless interest rate).
Following the practice of Section II, to simplify the analysis, it is assumed that the
explicit opportunity cost to investors for investing in the firm, k, is constant over time. If k
were changing over time, then in an analogous fashion to R(t) in Sections II and V, we could
t
define K(t) by [1+ K(t) ] ≡ ∏[1+ k(t)], and use "[1 + K(t)] " everywhere in the formulas when
t t

j=1

t
"[1 + k] " appears.

Independent Projects are project such that the firm can decide to do both or either one or neither.
(Note: this definition has no implications of statistical independence among projects.)

Mutually Exclusive Projects are projects such that the firm can only do one or the other, but not
both.

Traditional Methods of Project Selection

I. Pay-Back Method

135
Finance Theory

If I0 is the initial investment, then the payback period is that value of T such that
T

I 0 = ∑ X t . I.e., it is the minimum length of time until the net cash flows sum to the value of the initial
t=1

investment. The payback method says rank all (independent) projects from the shortest to the
longest and then take (invest in) all projects with a payback period less than or equal to some
*
given time, T . When choosing among mutually exclusive projects, select the one with the
smaller payback period.

II. Present Value Method (Review Section II)

The (net) present value of a project is

N N
X(t) X(t)
PV = - I 0 + ∑ t
= ∑ t
.
t =1 (1 + k ) t =0 (1 + k )

As described in Section II, the present value rule says rank all (independent) projects from the
highest to the lowest, and then take all investments with positive (or as a matter of indifference,
zero) present value. When choosing among mutually exclusive projects, select the one with the
largest present value.

Note: If the cost of capital were changing over time, then the present value of the project will be
.
N
X(t)
PV = - I 0 + ∑ and the method is still applicable
t =1 [1 + K(t) ] t

III. Internal Rate of Return Method (Review Section V on Yield-to-Maturity)

The internal rate of return for a project, i, is that discount rate such that the present
value of the project (computed at that rate) is zero. I.e., i is the solution to

136
 Robert C. Merton

N
X(t)
0= - I0 + ∑ .
t =1 [1 + i ] t

It is called an internal rate because, unlike k (the cost of capital), which is an (external) market
(opportunity cost) rate, i depends only on the nature of the time-flow patterns of the project and
is completely unrelated to any market rate.
The internal rate of return rule says rank all (independent) projects from the highest to
the lowest, and then take all investments whose internal rate of return is greater than some
* *
specified rate i (usually taken to be the cost of capital, i.e., i = k). When choosing among
mutually-exclusive projects, select the one with the largest internal rate of return.

IV. Profitability Index Method

⎡ T X(t) ⎤
Profitability Index = PI = ⎢∑ t⎥
/ I0
⎣ t=1 (1 + k ) ⎦
Method: Rank all (independent) projects form the highest to the lowest and take all investments
with profitability index greater than one. When choosing among mutually exclusive projects,
select the one with the largest profitability index.

Evaluation of these Methods:

Problems with Payback

1. Neglects the time value of money (no discounting)

2. Neglects all flows beyond the payback period (implicit "infinite" discounting)
Therefore, misses future negative or positive flows.
A related method sometimes used is the "Modified" Payback Method. The modified payback
period is defined as the minimum T such that
T
X(t)
I0= ∑
t =1 (1 + k )t

137
Finance Theory

Present Value
In perfect capital markets and certainty, the value of the firm is equal to the present value of all
its future flows discounted at the (market-determined) cost of capital. Hence, the present value
rule maximizes the value of the firm. It is sometimes called a conservative rule because the firm
always has available investments which will earn k: namely, it can buy its own stock. Thus, the
firm should never take negative PV projects. In uncertainty, the rule can be modified according
to the "risk-adjusted" method to be discussed later: Namely,

N
α t X (t)
PV = - I 0 + ∑
t =1 (1 + r )t

where α t = a certainty equivalent and X (t) is the expected cash flow. In general, present value is
the most appropriate of these four traditional techniques.

Internal Rate of Return

While the present value method assumes that the flows can be reinvested at the cost of capital
(which is always possible), the internal rate of return assumes that the flows can be reinvested at
the internal rate of return i.

Technical Problems that Can Arise with Internal Rate of Return

1. There may be either more than one value of i or no value of i which makes the
present value of the project zero.
2. If the cost of capital is varying over time, the "cut-off" rule of taking only projects
with i = k is not well-defined.

Example: Present Value vs. Internal Rate of Return

Assume k = .05

138
 Robert C. Merton

Present Value
Project A (Mutually Exclusive of B)

Year End Net Cash Flow Discount 1 Present Value

t
(1+k)

0 -1,000,000 1 -1,000,000

1 3,150,000 .952 2,998,800

2 -3,307,500 .907 -2,999,992

3 1,157,630 .864 1,001,192

Present Value of A = Sum of PV = 0

Project B (Mutually Exclusive of A)

Year End Net Cash Flow Discount 1 Present Value

t
(1+k)

0 -1,000,000 1 -1,000,000

1 3,210,000 .952 3,055,920

2 -3,433,800 .907 -3,114,457

3 1,224,080 .864 1,057,605

Present Value of B = Sum of PV = -4,548

So, by the Present Value Method, A is preferred to B.

Example (continued)

Internal Rate of Return

Project A: Let x = 1 + i

139
Finance Theory

3,150,000 3,307,500 1,157,630

PV = - 1,000,000 + - 2
+ 3
=0
x x x

or find the roots of the cubic equation:

- x 3 + 3.15 x 2 - 3.3075x + 1.15763 = 0

( x1 , x 2 , x3 ) are x1 = 1 + i1 = 1.05
3 roots: x 2 = 1 + i 2 = 1.05 so i A = .05
x 3 = 1 + i 3 = 1.05
Project B: Let x = 1 + i

3,210,000 3,433,800 1,224,080

PV = - 1,000,000 + - 2
+ 3
=0
x x x
or find the roots of the cubic equation:
− x3 + 3.21x 2 − 3.4338 x + 1.22408 = 0
( x1 , x 2 , x3 ) are x1 = 1 + i1 = 1.04
3 roots: x 2 = 1 + i 2 = 1.07
x 3 = 1 + i 3 = 1.10
B B B
So there are three internal rates of return i 1 = .04; i 2 = .07; i 3 = .10

140
 Robert C. Merton

Example (continued)

"Switching Points" at k = .05 take A over B

at k = .08 take B over A

141
Finance Theory

As noted in Section VII, this is not a paradox because different interest rates imply different
"worlds" with different alternatives. Thus, which technology to use (e.g., wood bridge versus a
steel bridge) rarely can be answered with knowledge of the technology only. The switching
problem (or multiple-roots problem) occurs when there is more than one positive root which
makes the present value equal to zero. One can use the following rule to check to see whether
n n -1 n-2
more than one such root can occur: if x + a1 x + a 2 x + ... + a n = 0, then (Descarte's rule of
signs) the number of positive roots either is equal to the number of variations of signs of the ai's or is
less than this number of variations by an even integer.
In the example, both projects had three sign changes, and hence, either three or one
positive roots.
It should be noted that from the tables in this example, both the payback and the modified
payback methods would have picked Project B over Project A.

More on Present Value versus Internal Rate of Return

If X(0) = – I0 < 0 and all X(t) ≥ 0, for t = 1,2,... for all the projects being considered
and if the projects are independent, then the Present Value Rule and Internal Rate of Return Rule
will lead to the same answer with respect to which projects will be taken. To see this, note that a
plot of present value versus cost of capital will look like:

142
 Robert C. Merton

Hence, if i > k, then the present value will be positive. However, even in the case of a single
positive root, the rankings of projects by the two methods can be different. Hence, danger lurks
for evaluating mutually exclusive projects using i or in using it in the case of capital rationing as
the following example illustrates.

Example: Suppose that you have $1000 and you can purchase either Project A or Project B.
Given that the only investment alternative available in future years for any money received will
be to stuff it in a mattress or bury it in a coffee can (i.e., k = 0), which should you take?

Project A: Pay $1000 today (i.e., I0 = 1000) and you receive no payments until the end of
fifteen years when you will receive $4,177 (i.e., x(1) = x(2) = x(3) = ... = x(14) = 0 and x(15) =
4177).

Project B: Pay $1000 today (i.e., I0 = 1000) and you receive $214 at the end of each year for
fifteen years (i.e., x(1) = x(2) = x(3) = ... = x(14) = x(15) = 214).
We know by Descarte's rule of signs that both bonds have only one positive root. Hence, the
internal rate of return for both is unique. Using the present value tables and the formula for an
annuity, the internal rate of return on A is iA = .10 and on B is iB = .20. Clearly, on a IRR
basis, B is preferred to A. What about present value? At k = 0,

4177
PV A = - 1000 + = 3,177 and
(1 + 0 )15
15
214
PV B = - 1000 + ∑ t
= - 1000 + (15x214)
1 (1 + 0 )
= 2210

Clearly, by the Present Value Rule, A is preferred to B. Which is "more" correct? Fist, note
that since all interim payments cannot be invested to earn a positive return, it is easy to compute
how much money we will have at the end of fifteen years from each project: for Project A, we
have $4177 and for Project B, we will have only $3210. Since they both cost the same, which do

143
Finance Theory

you prefer? Further, since we know the final amounts, we can compute an actual average
15 4177
compound return per year for both. I.e., (1 + R A ) = has the solution RA = .10. So, the
1000
15 3210
true return per year from A is 10%, and (1 + R B ) = has the solution RB = 8.2%. So, the
1000
true return per year from B is 8.2% NOT 20%. Hence, Present Value is a better ranker. Note:
the internal return, iB, is a number and need not bear a close relationship to the actual returns
earned. E.g., 20% versus 8.2%.
In bond evaluation, yield-to-maturity is just an internal rate of return calculation, and
therefore, as noted in Section V, the same warnings apply to comparing alternative bond
investments by yield-to-maturity even when the bonds have the same maturity date.

Imperfections and Capital Budgeting

If the firm is a "perfect competitor" for capital (i.e., the firm's cost of capital is unaffected by the
scale of its investments) and capital markets are "reasonably" perfect, then the correct capital
budgeting decision rule is present value. However, in the face of certain imperfections, this
decision rule may require modification.

Capital Rationing: an examination of all the decision rules given shows that each assumes that
there is no budget constraint for profitable investments. I.e., each period, the firm looks over all
available project proposals and selects all projects with positive present value. This done, then a
budget is established to determine how much capital is needed (and from which sources it will be
raised) to carry out the program. If the estimates of the cost of capital and the cash flows are
accurate, then there should be little problem in raising the necessary (additional) funds in the
capital market. Further, this procedure is optimal relative to the (efficiency) criterion of
maximizing market value. Note: the procedure to be described is contrary to the one an
individual consumer would follow in allocating his income (and wealth) over various
consumption goods at different points in time.

144
 Robert C. Merton

However, in certain situations, there may be a (predetermined) absolute limit to the

amount that can be invested by the firm in any one period. This situation is called capital
rationing. It may occur for the firm in countries where there are no (or poorly-organized) capital
markets; or for divisions of firms where (incorrectly determined) decentralization rules dictate a
fixed budget for each division prior to the examination of the projects available; it is not an
infrequent case in the public sector where resources are at times allocated (prior to specific
knowledge of projects) on the basis of "last year's" allocation (of I0)".
Under capital rationing, it is sometimes suggested that the Profitability Index (or
"Benefit/Cost" ratio) is a better rule than present value. While it is true that the profitability
index gives the most Present value per dollar of initial investment which is highly suggestive of
what one should do in a constrained situation, it does not reflect future budgetary constraints.
Thus, a plan may satisfy the current budget constraint, but violate all future constraints.
The best technique in this situation is to maximize present value subject to the budget
constraint in each year using mathematical programming techniques. While such a procedure is
not optimal relative to (unconstrained) maximizing of market value, it does produce a feasible
program. Moreover, the "shadow prices" or dual variables will give an explicit estimate of the
marginal costs of the rationing. These values can often be used to argue for the elimination of
the constraints, particularly if the costs are high. Always, ask yourself: why the constraint? How
much is it costing? Is it rational?

Rising Cost of Capital. It is typically assumed that the cost of capital is a constant function of the
amount of investment, in each period. However, if k depends on the scale, then programming
techniques must be employed.

145
Finance Theory

Application of Present Value: The Replacement Problem

The product decisions are already made and the decision is to choose between two
alternative machines to produce the product. Technical change is neglected and the optimal
horizon for the product run is given.

I. Replacement time for each machine is known.

Same product for T years: which machine? Life of machine A is T1 years. Life of machine B
is T2 years.

146
 Robert C. Merton

1 2
Machine A costs IA and has operating costs per year C A ,C A ,...,C TA1 and has salvage value with
(2 T 1 - T) years (to go), of S.

1 2
Machine B costs IB and has operating costs per year C B ,C B ,...,C TB 2 . Assume replace each
machine with the same machine.

The present value of the costs of machine A over its life is

T1 t
PA= I A + ∑
CA
t
t =1 (1 + k )

For machine B
T2 t
PB = I B + ∑
CB
t
t =1 (1 + k )

If we choose machine A, then it must be replaced at time T1. At that time, the present value of
costs will be
T -T 1
Ct S
I A+ ∑ (1+ kA )t - (1+ k )T -T , because it is not used for its full life
1
t=1

If we choose machine B, then it must be replaced at time T2. At that time, the present value of
costs will be PB again because it is used for its full life. To decide which machine to use, we
compare the present values of costs for the entire product life, which are, today,

1 ⎡ T −T1 t
CA S ⎤
PA' = PA + ⎢ IA + ∑ − T −T1
⎥
(1 + k )T 1
⎢⎣ t =1 (1 + k ) t ( ) ⎥⎦
1 + k

and

147
Finance Theory

1 ⎡ 1 ⎤
PB' = PB + PB = PB ⎢1 + ⎥ ,
(1 + k ) (1 + k ) 2
T2 T
⎢⎣ ⎥⎦

' '
and choose the smaller one between PA and PB .
A common situation is when the length of the product run is anticipated to run
indefinitely into the future (formally, T = ∞). Hence, if a machine has life of length n, we
anticipate replacing the machine every n years and making an "infinite" number of
replacements. If P is the present value of costs of the machine over one cycle, i.e.,
n
P= I +∑
Ct
t
, then, as above, the present value of costs over the product life will be
t =1 (1 + k )

P P P P
P' = P + + + + +K
(1 + k ) (1 + k ) (1 + k ) (1 + k )
n 2n 3n 4n

⎡ ⎤
⎢ 1 1 1 1 ⎥
= P ⎢1 +
'
1
+ 2
+ 3
+ 4
+ K⎥
⎢⎣ ⎡⎣(1 + k ) ⎤⎦ ⎡⎣(1 + k ) ⎤⎦ ⎡(1 + k ) n ⎤ ⎡(1 + k )n ⎤
n n
⎣ ⎦ ⎣ ⎦ ⎥⎦

∞
1
or P' = P ∑X j
where X ≡
(1 + k )
n
j =0

N -1
N
X -1 1
From (II.14), we have that ∑ X = as N → ∞ for X < 1.
j
= So substituting for
j =0 X -1 1- X

X, we have

⎡ ⎤
⎢ ⎥ ⎡ ⎤
⎥ = P ⎢ (1 = k ) ⎥
n
1
P' = P ⎢
⎢ 1 ⎥ ⎢⎣ (1 + k ) − 1 ⎦⎥
n

⎢1 − 1 + k n ⎥
⎣ ( ) ⎦
'
So to determine which machine to choose, take the one with the smaller P .

148
 Robert C. Merton

An alternative representation (due to Lewellen) is to convert the present value

calculations into a constant annual cost (flow) comparison. This approach would be useful for
comparison between the choice between buying and maintaining the machine or renting (or
leasing) the machine with a service contract from another firm, i.e., what is the (maximum)
constant annual payment that you would be willing to pay at the end of each year for renting the
machine and having it serviced? Clearly, this represents an annuity payment problem. The
present value of the annuity is P; the (maximum) rate to be paid is k; find the annual payments
implied: From Section II, we have that the formula for a N-year annuity is

1
y(1 - N
)
(1 + k )
AN =
k
where y = the annual payment.
Corresponding to the perpetuity (N = ∞), we have

y
A∞ =
k
⎡ ⎤
⎢ ⎥
⎢ k ⎥
so the annual flow (cost) will be y = kP = P
⎢ 1 ⎥
⎢1 - n⎥
⎣ (1 + k ) ⎦

Again, we can select between two machines by choosing the one with the smaller y, and if a
leasing contract is available for less than y, then take it.

II. Optimal Replacement Time

In the previous analysis, it was assumed that the machines' replacement times were
known and that they corresponded to their physical lengths of life. Rarely is this the case.
Normally, the decision to replace a machine is an economic one. As before, assume that we will
always be replacing old machines with new machines and that replacement goes on indefinitely.

149
Finance Theory

Let I = initial investment and C1,C2,C3,...,CT be the annual operating costs up to the end of the
physical life of the machine, T; let S1,S2,S3,...,ST-1,ST (= 0) be the salvage value of the machine
at the end of each year. From the assumptions of the problem, an optimal replacement time
solution will be the same for all time, i.e., it will never be optimal to replace every two years for a
while and then switch to replacing every three years, etc.
Let τ = length of time between replacements (0 < τ ≤ T). To "convert" the problem to
the type of the previous section, consider that each of T different replacement strategies defines
a "different" machine (which it does in the economic sense). Thus, let P(τ) = present value of
costs for one cycle for the machine replaced every τ years. Then
τ
P(τ ) = I + ∑ Ct - Sτ .
t =1 (1 + k )t (1 + k )τ

Let P(τ) be the present value of costs of the machine (replaced every τ years) over the product
life. Then

⎡1 1 ⎤
P (τ ) = ⎢ - τ
⎥ P(τ ) .
⎢⎣1 (1+ k ) ⎥⎦

*
The optimal replacement time, τ , will the be τ such that

P( τ * ) ≤ P( τ ) for all possible τ = 1,2,3,...,T

We can also write the conditions in terms of equalized annual costs by:

yτ = kP( τ ) and then τ * is such that

yτ * ≤ yτ for all possible τ .

*
Note: τ will depend on k, the structure of the operating costs, and the salvage values, and
hence will be different for different cost of capital, etc.

150
VIII. FORWARD CONTRACTS, FUTURES CONTRACTS AND OPTIONS

In this section, the assumption of certainty is removed and we begin the study of financial
instruments in an uncertain environment. Futures and option securities permit investors to
modify the patterns of returns which would otherwise be received from the underlying securities
and in particular, to eliminate or hedge against the uncertainties of price changes in these
securities. Forward or futures contracts have long been traded on basic commodities and in
recent years, have been widely expanded to include financial securities (see chart on the last page
of Section V). Although most organized markets use futures contracts, the more-intuitive
contract is the forward contract. As an introduction, therefore, we analyze the forward contract
in the context of a forward loan agreement. The reader may find it helpful to review the analysis
of default-free fixed income security pricing in Section V before proceeding to the example
It is a common practice for borrowers to obtain a commitment for a loan in advance of
actually receiving the money. For example, a firm may undertake a project that does not require
investment until some future date, and therefore, the firm may have no need for funds now.
However, to ensure the availability and terms of a loan sufficient to finance this future
investment, the firm may enter into an agreement with a bank now to borrow the money at a
specified future date. While the terms of such agreements are variable, a typical example would
be a τ-period term discount loan where the bank agrees to lend $L to the firm at date T and the
firm agrees to pay back to the bank $M at date T + τ.
Such an agreement is an example of a forward contract. Specifically, it is a forward
contract where the firm agrees to deliver to the bank at date T a τ-period discount bond on
which the firm promises to pay $M at maturity and the bank agrees to pay the firm $L (the
delivery price) on delivery. Although not necessary, typically no money changes hands at the
time that the forward contract is made. Under this assumption and the assumption that the loan
is default-free, what is the equilibrium value for M .
The spot price for an item is defined as the price for that item delivered immediately. For

example, the current spot price for a discount bond which pays $1 at date T is P0 ( T ) , and

at date t, the spot price for that same bond will be P t (T - t). The forward price associated

151
Finance Theory

with a forward contract is defined as that delivery price which makes the value of the forward
contract equal to zero at the time that the contract is made.
Because in the case at hand, nothing is paid by either party to the other for making the
contract, the terms must be such that the value of the forward contract (not to be confused with
the forward price of the contract) is zero at the time it is made. Otherwise, if the value of the
contract to the lender (borrower) were positive, then the value of the contract to the borrower
(lender) would be negative, and the borrower (lender) would be giving something of value away
for nothing. Hence, the relationship between L and M must be such that L is equal to the
forward price of the contract. If F t (L, M,T,τ ) denotes the value of the forward contract (to the
lender) at time t then, to avoid arbitrage from (V.18), it must satisfy

(VIII.1) F t (L, M,T,τ ) = - L Pt (T - t) + M Pt (T + τ - t)

for t = 0,1,2,…,T. Of course, the value of the forward contract to the borrower at time t is
- F t (L, M,T,τ ). From the condition that F 0 (L, M, T,τ ) = 0, we have that
M = L P 0 (T)/ P 0 (T + τ ), and therefore, (VIII.1) can be rewritten as

(VIII.2) F t = L P 0 (T){ [ P t (T + τ - t)/ P 0 (T + τ )] - [ P t (T - t)/ P 0 (T)] }

From (VIII.2), the value of the forward contract to the lender at date t is proportional to the
difference between the return per dollar from holding a (T + τ)-period discount bond for the
period [0,t] and the return per dollar from holding a T-period discount bond for the same
period. The proportionality factor is equal to the value at the time the contract is made of a
discount bond which pays $L at its maturity date T. Thus, while F 0 = 0 , the value of the
contract at date t , Ft will not equal zero if the holding period returns on the two bonds are not
the same. This can happen if interest rates are stochastic, and will happen whenever the ex-post
time path of interest rates is different from what was expected ex-ante. Of course, if Ft is
positive, then, ex-post, the borrower is worse off than if he had not entered into the agreement
because the value of the contract to him, –Ft , is negative. However, just as the lender is
committed to making the loan on the terms agreed upon, so the borrower is equally committed to

152
 Robert C. Merton

take the loan on these terms. If the borrower had the choice of not taking the loan, then the
agreement would not be a forward contract but rather an option contract.
Because $L is the T-period forward price for a default-free discount bond which pays
$M at date (T + τ), L/M is the T-period forward price for a default-free discount bond which
pays $1 at date (T + τ). If we define TP0(τ) to be the forward price for delivery at time T of a
discount bond which pays $1 at date (T + τ), then TP0(τ) = L/M and therefore,

(VIII.3) T P 0(τ) = P 0(T + τ)/ P 0(T) , τ = 0 ,1,... .

Since P 0 (T) is the amount one would pay today for one dollar delivered at date T and
P 0 (T + τ ) is the current price of the bond, from (VIII.3) the forward price TP0(τ) is equal to the
current price of the bond measured in units of dollars paid at date T. With a complete set of
these forward prices for all T and τ, the forward price associated with a forward contract for

any default-free, fixed-income security can be computed. If sV0 denotes the s-period forward

price for a default-free security which pays $ xt at date t, t = 1,2,...,T then

T
(VIII.4) sV 0 = ∑ s P0 (t - s) x t
t=s+1

for s = 0,1,..., T - 1, where it is assumed that the security is delivered at date s but ex-the-period

s-payment (i.e., after the payment of $ xs has been made).

The only data required to compute forward prices for all default-free securities are the
current spot prices of discount bonds or equivalently, the current term structure of interest rates.
Indeed, the more-common practice is to quote a forward yield rate, TR0(τ) rather than the forward
price TP0(τ) where in an analogous fashion to (V.8) in Section V, TR0(τ), is defined by

153
Finance Theory

−1/ τ
(VIII.5) T R 0 (τ ) ≡ ⎡⎣ T P 0 (τ ) ⎤⎦ − 1

Consider the more general case of a forward contract which calls for delivery of one unit
(e.g., share) of a security or commodity at date T at a price L. Let X(t) denote the spot price
of the commodity at time t. If shortsales are permitted; if there are no shortage or transactions
costs; and if the commodity or security provides no payouts prior to T, then it follows from the
condition of no arbitrage that the value of the forward contract is

(VIII.6) F t (L, T) = X(t) - LP t (T - t)

From the definition of the forward price at time t , Lt , we have from (VIII.6) that

(VIII.7) L t = X(t)/ P t (T - t) .

The dollar gain on a forward contract entered into at time t between t and t + 1 is given by

⎡ X(t + 1) Pt+1 (T - t - 1) ⎤
(VIII.8) F t +1 ( Lt ,T) - F t ( Lt ,T) = X(t)⎢ - ⎥.
⎣ X(t) Pt (T - t) ⎦

The change in the forward price is given by

X(t) ⎡ X(t + 1) P t+1 (T - t - 1) ⎤

(VIII.9) L t+1 - Lt = ⎢ - ⎥.
Pt +1 (T - t - 1) ⎣ X(t) Pt (T - t) ⎦
Combining (VIII.8) and (VIII.9), we have that

(VIII.10) F t +1( Lt ,T) - F t ( L t ,T) = P t+1(T - t - 1)[ L t+1 - L t ] .

The change in the value of a contract entered into at time t = 0 between t = 0 and t = T, the
delivery date, is given by

154
 Robert C. Merton

T -1

F T ( L 0 ,T) - F 0( L 0 ,T) = ∑ [ F t+1( L 0 ,T) - F t( L 0 ,T)]

t=0

(VIII.11) = X(T) - L 0
= LT - L 0

In preparation for the analysis of futures contracts, consider the following investment
strategy in forward contracts: enter into a forward contract at time t. At time t + 1, settle the
contract and put the proceeds into a discount bond which matures at time T . Now enter into a
new forward contract. The initial value of the investment is zero. The increment to value
between t and t + 1 is given by (VIII.10). This increment (invested in a discount bond) will be
worth at time T , { F t +1 ( Lt ,T) - F t ( Lt ,T)} / Pt +1 (T - t - 1) = Lt +1 - Lt . The total value of this
investment strategy at time T will, therefore, be given by
T -1

∑{ F
t =0
t +1 ( Lt , T) - F t ( L t , T) } / Pt +1 (T - t - 1)
(VIII.12) T -1
= ∑ [ Lt +1 - Lt ] = LT - L0
t=0

which is identical to (VIII.11), the increment from holding a single forward contract for the entire
period until delivery.
If a person is long in a futures contract, then he is required to purchase at date T one unit
of the security (or commodity) at the then futures price (denoted by) f(T), and he also will

receive in cash at date t + 1, the difference between the futures price at that date and the futures
price at date t [i.e., f(t + 1) - f(t)] for t = 0,...,T - 1 . [Note: If f(t + 1) - f(t) < 0, then he
must pay out | f(t + 1) - f(t) | in cash.]
If a person is short in a futures contact, then he is required to sell at date T one unit of
the security (or commodity) at the then futures price, f(T), and he also must pay in cash at date
t + 1, the difference between the futures price at that date and the futures price at date

155
Finance Theory

t, f(t + 1) - f(t), for t = 0,...,T - 1. [Note: If f(t + 1) - f(t) < 0 , then he will receive
| f(t + 1) - f(t) | in cash.]

The futures price at date t, f(t), is defined to be that price such that the value of a

futures contract is zero. It follows immediately that the futures price at the delivery date T is
equal to the spot price at that date. I.e., f(T) = X(T).
Note that holding a futures contract long until the delivery date is quite similar to the
examined investment strategy of "rolling over" forward contracts which from (VIII.12) was
shown to be equivalent to simply holding a forward contract until delivery.
Consider the analogous investment strategy in futures contracts: At the beginning of each
period t (t = 0,...,T - 1), enter into N t futures contracts. At the end of the period (i.e., t + 1 ),
you will receive $ N t [f(t + 1) - f(t)]. Invest this money in bonds that mature at date T . [So, at
date T , you will have $ N t [f(t + 1) - f(t)]/ P t +1 (T - t - 1) from this transaction.] Adjust your
position so that you have N t +1 contracts for the period t + 1 to t + 2.
The accumulated sum at date T , V T , from this strategy will be

T -1

(VIII.13) V T = ∑ N t [f(t +1) - f(t)]/ P (T - t - 1) .

t +1
t =0

Consider the case where we choose N t = 1. In this case, from (VIII.13), the accumulated
sum will be
T -1 T -1
1
V T = ∑[ f (t + 1) - f (t )] + ∑[ f (t + 1) - f (t )[ -1]
t =0 t =0 P t +1(T - t -1)
(VIII.14) T -1
= f (T ) - f (0) + ∑[ f (t + 1) - f (t )][1- P t +1(T - t -1)] / P t +1(T - t -1).
t =0

Thus, by inspection of (VIII.14), unlike a forward contract, the dollar return from entering into a
futures contract and remaining long one contract until the delivery date will not, in general,
produce a dollar return equal to the difference between the futures price at the time of initial
entry, f(0), and the price at delivery, f(T).

156
 Robert C. Merton

However, there is one case where a strategy in the futures contracts will exactly replicate
the outcome of a forward contract. Suppose that changes in interest rates over the life of the
futures contract are known with certainty. Then, a feasible strategy would be to set
N t = P t+1 (T - t - 1). [Note: since this strategy requires one to know at time t , the price that a
bond will have at time t + 1, it is not feasible in a world of uncertain interest rates.]
Substituting for N t in (VIII.13), we have that

T -1

V T = ∑[f(t + 1) - f(t)]
(VIII.15) t=0

= f(T) - f(0).

Suppose that simultaneously with following this strategy, we also go short one forward contract
at time t = 0. Because both the forward contract and the futures contract have zero value, these
positions require zero investment. From (VIII.11) and (VIII.15), the accumulated value from
'
these combined positions, V T , can be written as

'
V T = f(T) - f(0) - [ L T - L 0 ]
(VIII.16)
= L 0 - f(0)
'
because f(T) = X(T) and from (VIII.7) LT = X(T). But, to achieve V T requires no
investment. Hence, if L0 ≠ f(0), then an arbitrage opportunity would exist. Thus, to avoid
arbitrage, we have that, for nonstochastic interest rates,

(VIII.17) f(t) = Lt ,

and from (VIII.7), that

(VIII.18) f(t) = X(t)/ P t (T - t) .

Moreover, for most practical cases of relatively short-lived futures contacts (i.e., T not too
large), the uncertainty about the bond price at time t + 1 viewed from time t will be small, and
therefore (VIII.17) should be an excellent approximation.

157
Finance Theory

The reader is warned that while the relation between the futures and forward price derived
in (VIII.17) will hold whenever interest rates are nonstochastic, the relation between the futures
and spot price in (VIII.18) will only obtain under the posited assumptions about the security or
commodity underlying the contract. That is, if the underlying item is a commodity which is not
being stored, then f(t) < X(t)/ P t (T - t) can be a stable result because it is not possible to
shortsell the commodity spot. Similarly, if there are payouts on the security prior to the delivery
date or storage costs for the commodity, then (VIII.18) need not obtain without creating an
arbitrage opportunity.

Options: Insurance for the Value of Risky Securities

A Put Option gives it owner the right to sell a specified number of shares of stock at a
specified price per share (the "exercise price") on or before a specified date (the "expiration
date".) If the option is not exercised on or before the expiration date, then it expires and becomes
worthless.
If T denotes the expiration date and if S(t) denotes the stock price per share at date t,
then the value of the put option per share on its expiration date is Max [0, E - S(T)] where E
denotes the exercise price.

158
 Robert C. Merton

Put Option Viewed as a Term Insurance Policy

General Example (3/6/79)

Asset Insured Stock IBM

Asset's Current Value Stock Price, $S $303.875

Term of Policy Time until expiration 7 months and 14

of the put days (10/20/79)

Maximum Insurance Cover- Exercise Price $300.00

age [Face Value of Policy of the Put, $E
(maximum loss to insurer)

Amount of the Deductible $[S – E] $3.875

(maximum loss to insured)

Insurance Premium Put Price/per share $15.25

Important Differences
• Early Exercise and Marketability
• Dividends

Three Ways of Reducing Risk

1) Diversification: "Mixing" less-than-perfectly correlated risky assets
2) Substituting the riskless security for risky assets
3) Insurance: options
If an investor holds a risky security and reduces his risk by the purchase of a put option on
that risky security, then such an investment strategy is called a "Protective Put" or "Insured
Equity" strategy.
Figure VIII.1 illustrates the basic payoff structure to a "Protective Put" strategy for the
case when the risky security is IBM stock. Note that the payoff structure is a nonlinear function
of the price of IBM stock, and therefore, this method of reducing risk is fundamentally different

159
Finance Theory

from the alternative method of reducing risk which is to reduce one's holdings of IBM and invest
in the riskless security.
Reducing the risk of a portfolio of stocks by the purchase of put options can be
accomplished by either purchasing a put option on each individual stock within the portfolio or
by purchasing a put option on the portfolio itself. The pattern of returns achieved by these
alternate approaches to the Protective Put Strategy will be somewhat different.
The following table provides the simulated return experience from following a protective
put strategy where one purchases a put option on each stock in the portfolio.

160
 Robert C. Merton

*
Summary Statistics for Rate of Return Simulations

Stocks Mixed With Commercial Paper Strategies Versus "Protective Put" Strategies
July 1963- June 1977

Protective Protective
1/ 2/ 3/
Semi-Annual: Stocks Put (E = S) Put (E = .gS)

Average Rate of Return 4.6% 4.5% 4.7%

Standard Deviation 13.7% 7.9% 10.4%
Highest Return 49.1% 35.1% 40.8%
Lowest Return –16.4% –1.8% –7.0%
Average Compound Return 3.7% 4.2% 4.2%
Growth of $1000 $2,829 $3,209 $3,218

1/ 1/
75% Stocks 50% Stocks Commercial
Semi-Annual: 25% Paper 50% Paper Paper

Average Rate of Return 4.2% 3.8% 3.1%

Standard Deviation 10.0% 6.9% 1.0%
Highest Return 37.9% 26.8% 5.9%
Lowest Return –10.8% –5.3% 1.7%
Average Compound Return 3.8% 3.6% 3.1%
Growth of $1000 $2,812 $2,717 $2,339

*Source: "The Returns and Risks of Alternative Put Option Portfolio Investment
Strategies," by Robert C. Merton, Myron S. Scholes, and Mathew Gladstein
(Journal of Business, January 1982).

1/
Equal-dollar Weighted Portfolio of 30 Dow Jones Industrial stocks rebalanced semi-
annually. Returns include reinvesting all dividends. No provisions for taxes or transaction costs.

2/ 1/
Same as footnote 1 plus a six-month put option with exercise-price-equal-to-initial-
stock price for each share of each stock.

161
Finance Theory

3/ 2/
Same as except exercise price is equal to 90% of initial stock price.
The following table provides the simulated return experience from following a protective
put strategy where one purchases a put option on the whole portfolio. The portfolio chosen was a
value-weighted portfolio of all New York Stock Exchange stocks and the particular Protective
Put Strategy examined was to purchase a one-month put on the portfolio with an exercise price
equal to the initial value of the portfolio times one plus the one-month interest rate.

Summary Statistics for Rate of Return Simulations

January 1927 - December 1978

30-Day U.S.
NYSE Protective Treasury
Per Month: Stocks Put Bills
Average Rate of Return 0.85% 0.55% 0.21%
Standard Deviation 5.89% 3.55% 0.19%
Highest Return 38.55% 30.14% 0.81%
Lowest Return -29.12% -7.06% -0.24%
Average Compound Return 0.68% 0.49% 0.21%
Growth of $1000 $67,527 $21,400 $3,604
Average Annual Compound Return 8.47% 6.04% 2.55%

A call option gives its owner the right to buy a specified number of shares of stock at a
specified price per share (the "exercise price") on or before a specified date (the "expiration
date"). If the option is not exercised on or before the expiration date, then it expires and becomes
worthless. The value of the call option per share on its expiration date is Max [0, S(T) - E].
As an exercise, show that the value at the expiration date of a protective put strategy
levered by going short (i.e., borrowing) in a riskless discount bond with face value of $E and
maturity date equal to the expiration date of a put is exactly equal to the value of a call option on

162
 Robert C. Merton

the same stock with exercise price and expiration date the same as for the put. Having shown
this, you will have proved that the purchase of a call option is equivalent to buying the stock;
levering the position by borrowing; and insuring the risk by purchasing a put option.

On the Relationship Between Risky Debt and Options

Consider a firm with two classes of liabilities: equity and debt. Assume that there is a
single, homogeneous class of debt with the following terms:

1. The debt is a "pure" discount loan where the firm promises to pay $M ("face
value") for each bond on the maturity date T. If there are n bonds outstanding,
then the total promised payment to the debtholders is $B ≡ nM on the maturity
date T.

2. In the event that the firm does not make the promised payment ("default"), then
the firm is turned over to the debtholders, and each bondholder will receive his
pro rata share of the "reorganized" firm. The original equityholders will receive
nothing in that event.

Let V(t) denote the market value of the firm at date t (which, by definition, will
always be equal to the sum of the market value of debt plus equity).
On the maturity date of the debt, if the value of the firm exceeds the amount of the
promised payment (i.e., V(T) > B), then it is in the interest of the equityholders (who elect
management) to have the debt paid. Thus, the value of the debt issue in that event will be B,
and the value of equity will be V(T) - B.
On the maturity date of the debt, if the value of the firm is less than the amount of the
promised payment (i.e., V(T) < B), then the firm cannot make the promised payment. Because
corporate equity enjoys limited liability, the equityholders cannot be compelled to contribute the
"short fall" to pay the bondholders, and it is, clearly, not in their interests to do so. Thus, the firm
will default, and the value of the debt issue in that event will be V(T), and the value of equity
will be 0.
In summary, on the maturity date, the

163
Finance Theory

(VIII.19) value of debt issue = Min [B,V(T)] = B - Max [0, B - V(T)]

and the

value of equity = V(T) - Min [B,V(T)]

(VIII.20) = V(T) - B + Max [0, B - V(T)]
= Max [0,V(T) - B].

Note: If the debt issue were default-free, then the value of the debt at maturity would always
equal B, the promised amount.

Inspection of the above value formula shows, therefore, that risky corporate debt "looks
like" a combined position of buying a default-free discount bond with face value B and maturity
T and issuing (short-selling) a put option on the firm's value with an exercise price = B and an
expiration date T. If there was not limited liability or equivalently, if the equityholders had
chosen to get leverage by personal (unlimited liability) borrowing where the aggregate face value
of the loan were B, then the value of the equity at maturity would be V(T) - B [which could, of
course, be negative].
Inspection of the value formula for equity (VIII.20) shows that equity levered by
corporate borrowing "looks like" a combined position of levering equity with an unlimited
liability loan and purchasing a put option on the firm's value with an exercise price = B and
expiration date T.
Thus, if one buys corporate debt, then one is not only lending money, but is also issuing
insurance to the equityholders against declines in the asset value of the firm below $B.
Similarly, if one issues corporate debt, then one is not only borrowing money, but is also
purchasing insurance against a decline in the value of the firm's assets below $B.
Most kinds of insurance of guarantees of the value of a security can be viewed as options.
Hence, the theory of option price determination has much broader application beyond simply
evaluating puts and calls. Some examples would be deposit insurance and loan guarantees.

164
IX. THE FINANCING DECISIONS BY FIRMS: IMPACT OF CAPITAL
STRUCTURE CHOICE ON VALUE

The capital structure of a firm is defined to be the menu of the firm's liabilities (i.e, the
"right-hand side" of the balance sheet). A great variety of types of securities can be and are used
in firms' capital structures. In addition to common stock equity, some typical examples are bank
loans, commercial paper, secured bonds, debentures, convertible bonds, income bonds, preferred
stock, and warrants. While the traditional treatment of capital structure is to examine each of
these types of securities separately, the modern approach (as was suggested in Section VIII)
views all these types as part of a unified theory of contingent claims pricing. That is, each of
these "hybrid" securities can be represented as a "mixture" (albeit at times a complicated one) of
pure" "default-free" debt and "pure" (as if 100%-financed by common stocks) equity. Beyond
simply providing a unified theory of pricing, this approach avoids many of the pitfalls and
misconceptions about the costs and benefits of different capital structure choices. Therefore, the
analysis in this section of the capital structure choice and its impact on the total market value of
firm will focus almost exclusively on the choice between debt and equity in providing the firm's
external financing.
As background, the reader should become familiar with the meaning (and effects) of
financial leverage and with the distinction between financial risk and asset (or business) risk. It
will be helpful to develop a feel for typical debt-to-asset ratios in various industries.
In the study of the firm's investment decisions in Section VII, it was assumed that all
external financing was done by issuing equity, and therefore that the firm had the simplest
structure possible: namely, all claims on the firm are homogeneous equity. The fundamental
question explored here is: Given the investment decision of the firm, does the financial structure
of the firm "matter"? That is, for a fixed investment policy, will a change in the firm's mix
between debt and equity cause a change in the market value of the firm?
As one might expect, the answer to this fundamental question depends upon the assumed
environment. Since, by hypothesis, the investment policy of the firm is fixed, the total cash flow
generated by the firm will not be affected by the capital structure choice. Thus, if the capital
structure matters (and thereby, a change in it will cause a change in the market value of the firm),

165
 Robert C. Merton

then, from the valuation formulas derived in Sections VI and VII, a capital structure change must
cause a change in the cost of capital. [The only exception would be if the capital structure choice
changes either the tax liabilities or the level of government subsidies to the firm, a topic
addressed later in this section.] Since, as shown in Section VII, the cost of capital is used in
determining the (optimal) investment decision, if the capital structure matters, then it is necessary
for the manager to consider simultaneously both the investment and financing decision in making
an overall optimal set of decisions for the firm.
Why should the financial structure matter? With the exception of certain tax features, it
is necessary to assume some type of uncertainty to give this question serious meaning because
otherwise, debt and equity are essentially indistinguishable. Possibilities are:
1. Does the issuance of debt create a new set of securities which were previously not
available? (i.e., how substitutable is personal leverage for corporate leverage?)
2. Are there costs to bankruptcies?
3. Are there tax features unique to corporate debt?
4. What are the effects on control of the firm?
5. Does the existence of outstanding debt "induce" changes in investment policy?
Other factors which are often considered by managers in deciding on the debt/equity ratio are:
1. growth rate of future sales
2. stability of future sales
3. the competitive structure of the industry
4. the asset structure of the industry
5. lender attitudes toward the firm and its industry.

To analyze the problem, we begin by studying the impact of capital structure in a specific
environment and use this as a benchmark for insights into why financing decisions might affect
value. This environment includes the following assumptions:

(A.1) No income taxes (to be modified later).

166
Finance Theory

(A.2) The debt-to-equity ratio is changed by issuing debt to repurchase stock or

by issuing equity to pay off debt. Moreover, a change in the capital
structure is affected immediately and there are no transactions costs.
(A.3) The firm finances all investment by external means (i.e., dividend policy is
to pay dividends equal to earnings).
(A.4) The expected values of the (subjective) probability distributions of future
(operating) earnings for each firm are the same for all investors
("homogeneous investor beliefs").
(A.5) No growth of earnings: the expected value of operating earnings for all
future periods are the same.
(A.6) All investments that the firm considers are from the same risk class, i.e.,
the business risk characteristics are independent of the number of projects
taken, and are taken as constant.

A "Benchmark": The "Pure Equity" Case (100% Financing by Equity)

Let X ≡ average expected dollar return per period for the firm
Let k 0 ≡ cost of capital for 100% equity financed firm
= expected rate of return required for the firm's particular risk characteristics, and it
is assumed to be constant over time.
Then, the market value of this firm is:

X
V0 =
(IX.1) k0

Let F = (expected) annual interest payments on debt outstanding

B0 = market value of debt outstanding

167
 Robert C. Merton

E = (expected) annual earning available to shareholders

= X -F

S0 = market value of equity outstanding

F
ki ≡ = cost of debt = required (expected) return by investors to
B0
hold this amount of debt in the firm

E
ke ≡ = cost of equity = required (expected) return by investors
S0
to hold this amount of equity in the firm

V 0 = S 0 + B0

X F+E +
k0 ≡ = = k i B 0 k e S 0 or
V0 V0 V0

⎛ B0 ⎞ ⎛ S0 ⎞ ⎛ B0 ⎞ ⎛ ⎞
(IX.2) k 0 = ⎜⎜ ⎟⎟ k i + ⎜⎜ ⎟⎟ k e = ⎜⎜ ⎟⎟ k i + ⎜⎜ S 0 ⎟⎟ k e .
⎝V 0 ⎠ ⎝V 0 ⎠ ⎝ B0 + S 0 ⎠ ⎝ B0 + S 0 ⎠

k 0 is called the "weighted" cost of capital and is the relevant number to use in the investment

(capital budgeting) decision of Sections VI and VII.

The question "Does financial structure `matter'?" can be restated as "does k 0 change for
different mixes of debt and equity (given a fixed investment policy)?" Or, for a given level of business or
asset risk, does changing the financial risk of equity change the total value of the firm?

168
Finance Theory

"Extreme" Classical Theory: The "Net Income" Approach

Assumption: k e and k i are constants with k e > k i

Example: k e = 10% and k i = 5%; net operating earnings = $1000

Case 1: B 0 = $3000 which implies F = $150 and so,

⎛ B0 ⎞ ⎛ S0 ⎞
Net Operating Earnings $ 1000 k 0 = ⎜⎜ ⎟⎟ k i + ⎜⎜ ⎟⎟ k e
⎝V0 ⎠ ⎝V0 ⎠

⎛ 3000 ⎞ ⎛ 8500 ⎞
minus interest payments (150) =⎜ ⎟ (.05) + ⎜ ⎟ (.10)
⎝ 11500 ⎠ ⎝ 11500 ⎠

1000
Earnings available to equity $ 850 = ≈ 8.7%
11500

÷ k e = .10
______

B0 = 3000
Market Value of Equity ( S 0 ) $8500 Leverage Factor ≡
S 0 8500

+ Market Value of Debt ( B0 ) 3000 = ≈ .35

Market Value of Firm ( V 0 ) $11500

Case 2: B0 = $6000 which implies F = $300, so

⎛ 6000 ⎞ 7000
Net Operating Earnings $1000 k0 = ⎜ ⎟ ( .05) + (.10)
⎝ 13000 ⎠ 13000

1000
minus interest payments (300) = ≈ 7.7%
13000

Earnings available to equity $ 700

÷ k e = .10 ______

169
 Robert C. Merton

B0 = 6000
Market Value of Equity ( S 0 ) $7000 Leverage Factor ≡
S0 7000

+ Market Value of Debt ( B0 ) 6000 ≈ .86

Market Value of Firm ( V 0 ) $13000

Classical Approach (generally) as is illustrated in figure above

1. Assumes that there is an optimal capital structure and hence, through the
appropriate choice of leverage, the value of the firm can be increased.
2. Assumes that beyond some point of leverage, k e rises at an increasing rate.
3. Assumes that beyond some point of leverage, k i may rise.

170
Finance Theory

Modigliani-Miller (as is illustrated in figure above)

Assume perfect capital markets: equal information; no transactions costs; investors are
rational and believe everyone else is; free access to borrowing and lending; no taxes; firm debt is
default-free. Their basic propositions are:

(1) The total market value of the firm and its cost of capital, k 0 , are independent of its capital
structure (the total market value of a firm is calculated by capitalizing the expected stream of
operating earnings at a discount rate appropriate for its risk class).

171
 Robert C. Merton

(2) The required expected return on equity, k e , is equal to the capitalization of a "pure" equity
stream, plus a premium for financial risk which equals the difference between the "pure" equity
capitalization rate and k i , times the leverage factor ( B0 / S 0 ).

(3) Therefore, the "cut off" rate for asset selection (investment policy), k 0 , is independent of
the financing decision.

Graphically, under the assumption that the debt has no risk of default (hence, k i is constant),
the M - M result is:

Proposition 2 can be derived formally from equation (IX.2) as follows:

⎛ B0 ⎞ ⎛ S0 ⎞
k0 = ⎜ ⎟ ki + ⎜ ⎟ ke
⎝V 0⎠ ⎝V0⎠
⎛V 0⎞ B0 ( B0 + S 0 ) ⎛ B0 ⎞
ke = ⎜ ⎟ k0 - ki = k0 - ⎜ ⎟ k i
⎝ S0 ⎠ S0 S0 ⎝ S0 ⎠

or

⎛ B0 ⎞
k e = k0 + ⎜ ⎟ ( k0 - k i )
⎝ S0 ⎠
(IX.3)

172
Finance Theory

(First) Proof of M–M result by arbitrage

Suppose the borrowing and lending rates are equal and the same for all investors and
corporations, i.e., k i = constant ≡ r, rate of interest; (the debt is default-free). Consider two

companies with identical anticipated earnings, i.e., X 1 = X 2 = X. Suppose that company #1 has no
debt (financed completely by equity) and company #2 has some debt.

Company #1 Company #2

Earnings: X 1= X Earnings: X 2= X

1 2
Debt: B0 = 0 at rate k i = r Debt: B0 > 0 at rate k i = r

1 1 2
Equity: S 0 (= V 0 ) Equity: S0

1 2 2 2
Firm Value: V0 Firm Value: V 0 (= B0 + S 0 )

Consider an investor who currently owns S 2 dollars of stock in company #2. If α = % of total

shares of company #2 held by this investor, we have that S 2 = α S 0 .

2

Suppose that V 02 > V 10 , i.e., by the "right" choice of leverage, company #2 will have a larger value

than #1. If the investor continues to hold his present portfolio, his return in dollars, Y 2 , will be his

fractional claim, α , times the portion of earnings available to shareholders, X - rB 0 . I.e.,

2

Y 2 = α (X - r B0 )
2
(IX.4)

The investor could sell out his present holdings and choose an alternative portfolio as follows:
Step 1: sell his current holding for S 2 (= α S 0 ) dollars.
2

Step 2: borrow ( α B0 ) dollars.

2

Step 3: with the proceeds from steps 1 and 2, buy shares in company #1.
If S 1 = number of dollars invested in company #1's shares, then

S 1 = α S 0 + α B0 = α ( S 0 + B0 ) = α V 0 .
2 2 2 2 2

173
 Robert C. Merton
1
Step 4: as an owner of $ S 1 worth of shares of company #1, he will have claim on ( S 1 / S 0 ) percent
1 1
of #1's earnings. Because #1 has no debt, S0 = V 0 and so, he has claim on

( S 1 / S 10 )X = ( S 1 / V 10 )X dollars of return.

The return in dollars on his new portfolio, Y 1, will be

⎛ S1 ⎞ αV02
⎜V ⎟
2
( )
Y1 = ⎜ 1 ⎟ X − r αB0 = 1 X − rαB02 , from step 3
⎝ 0⎠ V0

⎛ V2 ⎞ ⎛ V 2 − V1 ⎞
(IX.5) ( )
= α⎜⎜ 01 X − rB 02 ⎟⎟ = α X − rB 02 + αX⎜⎜ 0 1 0 ⎟⎟ or
⎝ V0 ⎠ ⎝ V0 ⎠

⎛ V 2 − V1 ⎞
Y1 = Y2 + αX⎜⎜ 0 1 0 ⎟⎟
⎝ V0 ⎠

2 1
so for V 0 > V 0 , Y 1 > Y 2 . Hence, we have demonstrated that for the same number of dollars
2 1
invested in either case, if V 0 > V 0 , then the investor can earn a higher return in the second portfolio
than in the first, for every possible outcome of earnings, X. Therefore, rational investors will

"switch" to portfolio #2 from portfolio #1 until V 0 ≤ V 0 . The argument goes through in precisely the
1 2

1 2
same way, if it was assumed that V 0 > V 0 . Therefore, to avoid arbitrage or dominance,

1 2
(IX.6) V 0 =V 0 .

A second proof of the M - M proposition using the same notation and company data as in
the first proof is as follows (where firm #1 has no debt and firm #2 has some debt):

Hold as an investment: $ S 1 of shares of company #1 = α S 0 = α V 0 . The return

1 1
Case I:
from the investment will be = αX .

174
Finance Theory

Alternative Investment:
Transaction Investment Return

(1) (sell $ S1 of #1 and) buy

the same fraction α of
(
α S02 ≡ α ⎡⎣ V02 − B02 ⎤⎦ ) (
α X - rB02 )
the shares of #2

(2) buy α percent of the α B02 rα B02

bonds of firm #2

Total α V02 αX

1 2 1 2
So, if V 0 > V 0, the investor gets the same return for α(V 0 - V 0) fewer dollars invested.

Case II: Hold α% of shares of firm #2: $S2 of shares of #2 =

α (S02 = α (V02 -B02 ). The return will be α (X-rB02 ).

Alternative Investment:
Transaction Investment Return

(1) buy fraction α of

the shares of #1
(
α S10 ≡ α V01 ) αX

(
(2) borrow α B0 dollars
2
) − α B02 - α rB02

Total (
α V01 − B02 ) (
α X - rB02 )
If V0 > V0 , then the investor gets the same return for α (V0 -V0 ) fewer dollars. Hence, to
2 1 2 1

1 2
avoid dominance or arbitrage, V0 = V0 .

175
 Robert C. Merton

Thus, given their assumptions, M-M demonstrate, by a powerful arbitrage argument, that
the capital structure or financing decision among alternative instruments does not affect the
market value of the firm or its (average) cost of capital for determining which assets to purchase.
The intuition is that a purely financial transaction for a fixed amount of real assets should not
affect any "real" decisions or values. Or, if personal borrowing is a perfect substitute for
corporate borrowing, then M-M holds because investors will not pay more for firms that borrow
for them if they can do it themselves. The classical view of the capital structure simply assumes
that leverage "matters." M-M showed why it does not. To disagree with the conclusions of M-M
one must therefore disagree with their assumptions.

Items that could affect the M-M Conclusion

1. Tax deductibility of interest payments by the firm.

2. The risks of personal versus corporate leverage (limited liability and bankruptcy).
3. Cost of borrowing may be higher for the investor than the firm.
4. Institutional restrictions may prevent institutional investors from "levering". Margin
requirements restrict individuals.
5. Transactions costs in establishing the arbitrage position.
6. Moral Hazard: management makes decisions in the best interests of the shareholders that
may conflict with the interests of the bondholders, and therefore, reduce the overall
market value of the firm.

Of these items, by far, the most important are (1) and (2). In the proofs of M-M, it is
assumed that both corporate debt and personal borrowing is default-free, and therefore, investors
who levered equity by personal borrowing could exactly replicate the payoffs to investors who
held shares levered by corporate borrowing. As was already demonstrated in Section VIII, this
will no longer be the case when there is a possibility of default on the debt of the firm. To
review the difference between personal and corporate borrowing, we maintain the assumption

176
Finance Theory

that personal borrowing is default-free and briefly reexamine the case of a pure discount term
loan:
Consider a personal term loan (with no interim interest payments) with a face value of B
dollars due at time T in the future. Let the firm (unlevered) have a current value of V0. Then
T
the value of levered equity, Ep, would be Ep = V0 – B/[1 + R(T)] , and at time T in the future
if the firm is worth V(T), the payoffs to the debt and levered equity will be

Consider a corporate loan with the same terms except limited liability for the shareholders. The
payoffs are:

177
 Robert C. Merton

By inspection, the payoff to the debtholders in the corporate case is less favorable than in the
personal case. Correspondingly, the payoff to the equityholders in the corporate case is more
favorable than in the personal case. Therefore, the current value of the corporate-levered equity
will exceed the current value of the personal-levered equity, i.e., Ec > Ep. Correspondingly, the
current value of the corporate debt will be less valuable than the current value of the personal
debt, i.e., Dc < Dp.
Inspection of the payoffs to Ep versus Ec shows that, in the event that V(T) < B,
Ec(T) – Ep(T) = B – V(T). As noted in Section VIII, the corporate-levered equityholders are
"insured" against losses that would occur for the personal-levered equityholders if V(T) < B. So,
we can write the value of the corporate-levered equity as

Ec = E p + g
= V 0 - B / [1+ R(T) ]T + g

where g is the value of this "downside insurance" (i.e., the put option insurance premium).
Similarly, the corporate debtholder is not only lending money but, in addition, is "insuring" the
equityholder, i.e., we can write the value of corporate debt as
T
Dc = D p - g’ = B / ([1 + R(T) ] - g’

where g′ is the "liability" associated with issuing the put insurance (its cost). If V0 is the value
of the unlevered firm and if g = g′, then M-M holds even when bankruptcy is possible. That is,
even in the presence of default possibilities, M-M will hold if either put options on the stock exist
or if these options can be created by low-transaction cost investment strategies. Of course, if
there are significant "dead-weight" losses to the firm's liability holders from a bankruptcy (e.g.,
attorney fees, disruption of the operations of the firm), then corporate leverage can matter.
This analysis should serve to underscore once again (as noted in Section VIII) that one
cannot compare the "true" or "economic" cost of the debt of one firm with that of another by
simply comparing promised yields on debt. That is, by definition, the promised yield (for the
period) is simply [B/Dc] – 1 which can be rewritten (for T = 1) as

178
Finance Theory

g(1 + R)
(IX.7) Promised yield = R +
B - (1 + R)g

where g is the value of the (implicit) put option. If the value of the put option on one firm is
larger than the value of a corresponding put option on the other, then it is entirely possible that
the debt with the higher promised yield could have a lower economic cost than the debt with the
lower promised yield.
The analysis also makes clear why the promised yield on a personal loan will be lower
than on a (comparable) corporate loan because in the former, the investor pledges all his assets
and in the latter (with limited liability) he pledges only his share of the corporate assets.

Effect of Corporate and Personal Taxes on the M-M Result

The federal tax law allows corporations and individuals to deduct interest payments from
their income before computing taxes. This tax shield is a subsidy to borrowers and may induce
corporations and individuals to borrow when they otherwise might not.

Taxation in the M-M Model

Let X = operating income (before interest and taxes) and XT = after-tax earnings before
interest; Tc = corporate tax rate; Tp = personal tax rate; Bc = "long-run" amount of debt
outstanding and R = interest payments. V0 = value of unlevered firm and V = value of levered
firm. Let k0 = required pre-tax expected return on the unlevered firm. Then

X (1 - T c )(1 - T p ) (1 - T c ) X
(IX.8) V0= = .
(1 - T p ) k 0 k0
On the levered firm,

X T = (X - R)(1 - T c ) + R = (1 - T c )X + T c R

179
 Robert C. Merton

(1 - T p )r Bc
If the debt is riskless, MM argue that R = rBc, then the value of the debt is = Bc ,
(1 - T p )r

and they show, by their arbitrage argument, that

(1 - T c ) X R
V= + Tc
(IX.9) k0 r
= V 0 + T c Bc

In essence, because of the tax subsidy, the levered firm is equivalent to the unlevered firm plus a
certain number (TcBc) of riskless bonds. MM assume that all earnings are paid out as dividends
which are taxable at Tp. Moreover, they assumed that the magnitude of the tax shield is certain
which need not be so if Bc is "pegged" to V. The latter is not of substantive importance because
the value of the tax shield can be shown to equal TcBc, even if there is a possibility of default.
Alternatively to corporate borrowing, let all the flows be riskless and let Y = after-tax
income to an investor who maintains a fixed total leverage (corporate + personal borrowing =
constant) position. Let Bp = amount of personal borrowing. Then,

(IX.10) Y = [( X -r Bc )(1 - T c )(1 - T p ) - r B p (1 - T p )]

and the value of this stream will be

r[ Bc (1 - T c )(1 - T p ) + B p (1 - T p )]
V( Y ) = V 0 -
(IX.11) (1 - T p )r
= V 0 - Bc (1 - T c ) - B p

If Bc + Bp = constant, then dBc = –dBp and

dV
(IX.12) = - (1 - T c ) + 1 = T c as MM claim .
dBc

However, suppose that one pays capital gains on the income of the firm, then

180
Finance Theory

(IX.13) Y = [ ( X -r B c )(1 - T c )(1 - T g ) - r B p (1 - T p )] ,

and if we capitalize at (1 – Tp)r, then

r[ B c(1 - T c )(1- T g ) + B p(1- T p )]

(IX.14) V(Y ) = V 0 − and
r(1- T p )
dV (1 - T c )(1 - T g ) <
=1- 0 depending on T c , T g , T p
(IX.15) dB c (1 - T p ) >
> 0 if T p < T c

In summary, while the theoretical and empirical evidence is hardly conclusive on whether
or not capital structure matters, it is probably a reasonable conclusion that generally, the effects
of capital structure on the firm's cost of capital will not be large enough to make a capital
budgeting project worth undertaking when it would not have been undertaken if financed entirely
by equity. There are, of course, exceptions to this general rule especially when projects are
subsidized by government and the subsidy takes the form of below-market interest rate loans,
loan guarantees, or tax exemption for corporate debt.
In completing this section, we present another example of the care that must be exercised
in computing the cost of borrowing. The example is that of a bank loan with compensating
balances and line fees.

Problem IX.1. On the Cost of Bank Borrowing

Loan Commitment = Maximum that can be Borrowed = "Line" ≡ L

Principal Amount Borrowed = Gross Borrowings ≡ B
Stated Interest Rate on Loan ≡ R = r + δ
where r = "prime" rate and δ ≡ amount "over prime" charged.

181
 Robert C. Merton

CB ≡ required (by the bank) amount to be kept on deposit in free balances in the
form of noninterest-bearing demand deposits. ("Compensating Balances")

= c1L + c2B (i.e., a fraction of the line plus a fraction of the principal)
P ≡ penalty charged for not maintaining sufficient compensating
balances
__
= Rp[CB–CB]
where Rp ≡ penalty rate and CB ≡ compensating balances actually maintained.

D≡ amount of noninterest-bearing demand deposits maintained by firm

d≡ amount of noninterest-bearing demand deposits which would have been maintained by
the firm even if there were no loans.

Of each $1 deposited, $.16 must be maintained at the Federal Reserve, so that only $.84
represent free-balances.
Therefore, CB = .84D or D = CB/.84 ≈ 1.19CB
Fee is payable to the bank for the unused part of the line [i.e., L–B]. Let
RL ≡ rate paid as a line fee
M ≡ actual amount of money available for corporate purposes

(IX.16) M = B - D+d

I ≡ $ charges paid for money

(IX.17) I = R • B + R L (L - B) + R p [ CB -CB]

Let RT ≡ the "true" interest rate cost of borrowing = I/M

182
Finance Theory

(IX.18) { }
RT = ⎣⎢ R + RPC2 − RL ⎦⎥ B + ⎣⎢ RL + RPC1 ⎦⎥ L − RPCB / M

(IX.18') { }
RT = ⎣⎢ R + RPC2 − RL ⎦⎥ B + ⎣⎢ RL + RPC1 ⎦⎥ L − RP .84 D / [ B − D + d ]

Should the firm maintain the compensating balance or pay the penalty? [i.e., which D should be
chosen for d ≤ D ≤ 1.19CB + d]. Holding fixed the amount of money available for corporate
purposes, M, how is RT R affected by the choice of D?

{ }
dRT = [ R + R p c2 − RL ] dB − .84 R p dD M −
R
M
[ dB − dD ]

But dM = 0 ⇒ dB = dD. Therefore ,

dR
= ⎡( R − R L ) − (.84 − c 2 ) R p ⎤⎦ / M
dD M fixed ⎣
(IX.19)

dR T
< 0 if R p >
(R − RL ) ⇒ D optimum = 1.19CB + d
dD M (.84 − c2 )

dR T
= 0 if R p >
[ R − RL ] ⇒ indifference w.r.t. choice of D
dD M (.84 − c2 )

dR T
> 0 if R p >
[ R − RL ] ⇒ D =d
dD M (.84 − c2 ) optimum

A Numerical Example: Prime = r = 18%; δ = 2% so that R = 20%

183
 Robert C. Merton

Compensating Balance Requirement: 10% of the Line plus 10% of the Principal

[i.e., c1 = c2 = .10]

Compensating Balance Penalty Rate: Rp = [R – RL]/[.84 – c2]

Line fee: RL = 0.5%; Payments of interest and fees once a year.

Line = L = $10,000,000; d = 0.

Since Rp is such that for fixed M, the level of deposits has no effect upon RT , assume that D

__
is chosen such that D = CB/.84 (i.e., no penalties)

Amount for Corporate Purposes (M) "True" Interest Cost RT

$1,000,000 53.49%
2,000,000 37.81
3,000,000 32.59
4,000,000 29.97
5,000,000 28.41
6,000,000 27.36
7,000,000 26.61
8,000,000 26.05
9,000,000 25.62
10,000,000 25.27

This RT should be compared with "stated" rate of R = 20%

∂RT
Note: ≈ 1.25
∂r

[i.e., an increase in prime of 100 basis points will cause (at least) a 125 basis point increase in the

cost of the loan.]

184
X. THE INVESTOR'S DECISION UNDER UNCERTAINTY: PORTFOLIO
SELECTION

By assuming certainty and perfect exchange markets in Sections II-VII, the optimal
consumption and investment decisions by households were derived; a rational criterion function
for the firm was deduced and rules for investment choice by firms were established. Beginning
with Section VIII, and for the balance of these Notes, the certainty assumption is dropped. The
introduction of uncertainty substantially complicates decision making by all economic units. As
a result, the structure of the capital market and the types of financial instruments and
intermediaries required for an efficiently functioning economy are greatly expanded. As in the
certainty case, we begin with the analysis of the individual household or consumer allocation
problem. To do so, we postulate that the criterion of choice for individuals satisfies the von
Neumann-Morgenstern Expected Utility Maxim. That is, in choosing among uncertain
alternatives, each person's rankings of those alternatives can be represented by computing the
expected value of some utility function of the random variable payoffs to these alternatives.
In making an allocation of his wealth, the investor has many assets to choose from, and
within limits of divisibility and transactions costs, he can choose mixes or combinations of these
assets to form alternative portfolios. The solution to the general problem of selecting the best
asset mix is called portfolio theory. It takes as given the menu of available assets where assets
are operationally-defined by their joint probability distribution of end-of-period values. Thus,
strictly defined, portfolio theory has nothing to say about where these distributions come from or
about why some assets exist and others do not. However, we will use the theory in an
equilibrium context to deduce certain properties of these distributions; to determine what
information about the distributions is required by investors to make optimal decisions; and to
answer (at least in part) the question of why certain types of assets exist.
The basic formulation of the portfolio selection problem is as follows: Assume that the
investor has a von Neumann-Morgenstern utility function for end-of-period wealth, U(W) and
assume further that U is strictly concave (such investors are called "globally risk averse"). Let W 0

185
 Robert C. Merton

denote his initial wealth and suppose that there are n different assets or securities available in units
o
called "shares". Let Pi denote the current price per share of asset i , i = 1,2,..., n. which is known.
The investor can buy or sell all the shares he wants at the current price (i.e., he acts as a "price taker").
th
Denote by N i the number of shares (or units) of the i security that he chooses to purchase. His set
n

of feasible choices is determined by his budget constraint: W o = ∑ N i Pi .

0
Suppose that he has a
i=1

1
probability distribution for the end-of-period price per share, Pi , for each asset. Then, his end-of-
n o

∑N P
N i Pi
period wealth will be the random variable W 1 = i
1
i . Define wi ≡ to be the fraction of
i=1 Wo
th
his wealth invested in the i security and define the (random) variable return (per dollar invested)
1
Pi
in the i asset to be Z i ≡
th
o
. Then, we can write the expression for W 1 as
Pi

n n
⎛ N i Pio ⎞ ⎛ P1i ⎞ n
W 1 ∑ N i Pi ∑ ⎜ ⎜ ⎟ ⎜ o ⎟ W o W o ∑ wi Z i
⎟ ⎜ ⎟
1
(X.1) = = =
i=1 i=1 ⎝ W o ⎠ ⎝ P i ⎠ i=1

Note: by definition of wi and the budget constraint:

∑w = 1 .
i=1
i

So the investor's problem of portfolio selection can be written as either

n n
Max
(X.2) E{U( ∑ N i P1i )} subject to W 0 = ∑ N i Poi
{ N 1 , N 2 ,..., N n } 1 1

or equivalently
n n
Max
(X.2′) E{U( W o ∑ wi Z i )} subject 1 = ∑ wi .
{ w1 , w2 ,..., wn } 1 1

186
Finance Theory

Suppose that one of the available securities is "riskless" and offers a return per dollar of R with
1
th Pn
certainty. If, by convention, we choose the n security to be riskless, then we have that Z n = o
=R
Pn

with certainty. Note that 1 = ∑1 wi = ∑1 wi + wn or wn = 1 - ∑1 wi where m ≡ n - 1 .

n n -1 m
Define:
W1
Z ≡ ∑1n wi Z i = = return per dollar on the portfolio. With a riskless asset, we can rewrite Z as
Wo
Z = ∑1n wi Z i = ∑1m wi Z i + wn Z n = ∑1m wi Z i + (1 - ∑1m wi ) Z n = ∑1m wi ( Z i - Z n ) +

Z n = ∑1 wi ( Z i - R) + R . Hence, the investor problem becomes:

m

⎛ ⎡ m ⎤⎞
(X.3) Max E { U ⎜
⎜ W o ⎢ ∑ w i( Z i - R) + R ⎥ ⎟⎟ }
{ w1 ,w2 ,...,wm } ⎝ ⎣ 1 ⎦⎠
where { w1 ,..., wm } are "free" decisions variables since the budget constraint was substituted out. To
solve (X.3), the usual calculus technique gives us that the maximum

∂(EU)
= 0 = E{U ′( W 1 ) ( Z 1 - R)}
∂ w1
∂(EU)
= 0 = E {U ′( W 1 )( Z 2 - R)}
(X.4) ∂ w2
M M M
∂(EU)
= 0 = E {U ′( W 1 )( Z m - R)}
∂ wm

or

(X.4') E{U ′( W 1 ) Z i } = R E{U ′( W 1 )} i = 1,2,..., m .

{ w*1, w*2 ,..., w*m } wn = 1 - ∑1 wi )

* m *
Call (and by the budget constraint, the optimal portfolio
*
(proportions) which will be the solution to (X.4). Suppose some wi < 0 . This means that the investor

will short-sell security i . To short-sell, the investor borrows shares today and sells them (using the
proceeds of the sale to purchase other securities). He must return the same number of units of that
security at the end of the period. Unlike borrowing money, the short-seller has a liability for returning
*
the specified number of shares instead of a specified number of dollars. If wn < 0 , then we call it

187
 Robert C. Merton
th
borrowing. Since the n asset is riskless, the liability can be denoted as a specified dollar amount.
Let Z ≡ ∑1 wi ( Z i - R) + R = return per dollar on the optimal portfolio.
* *m

m
E{U ′( W 1 ) Z * } = E{U ′( W 1 )( ∑ w*i ( Z i - R) + R)}
1
m
= ∑ w E{U ′( W )( Z - R)}+RE{U ′( W )} .
*
i 1 i 1
1

But from (X.4), E{U’ ( W 1 )( Z i - R)} = 0, 1 = 1,2,..., m. So

(X.5) E{U ′( W 1 )Z * } = RE{U ′( W 1 )} .

Example: Quadratic Utility

b
U(W) = W - W 2 0 ≤ W ≤ 1/b, b > 0
2
1 1
= W≥ .
2b b
Suppose that there are just two securities: asset #1 has (uncertain) return Z 1 and asset #2 has a
certain return Z 2 = R. If w1 = w = the fraction of his wealth invested in the "risky" asset #1 and W2
= 1–W1 = 1–W = fraction of his wealth invested in the "safe" asset #2, then the investor's
portfolio problem can be described by:

Max E{U (W 1)} = Max E{U (W o[ w( Z 1 - R) + R ])}

{w} {w}
b
= Max E{W o[ w( Z 1 - R) + R] - W o2[ w 2( Z 1 - R) 2 + 2 wR( Z 1 - R) + R 2]}
{w} 2
*
Maximizing, by the usual calculus, gives the condition that w = w , the optimal portfolio, when

dEU 2
0= = E{ W o ( Z 1 - R) - bW o2 [ w* ( Z 1 - R ) + r( Z 1 - R)]}
dw

188
Finance Theory

or by dividing by W o and rearranging terms

E( Z 1 - R) = bW o w* E{( Z 1 - R )2 } + b W o RE( Z 1 - R) or

* (1 - bW o R)E[ Z 1 - R]
(X.6) w= 2
bW o E[( Z 1 - R ) ]

1
If W o R > (i.e., if it is not possible to achieve the "satiation" level of wealth for certain), then a
b
*
necessary condition for him to hold some amount of the "risky" asset (i.e., w > 0) is that
E( Z 1 - R) > 0 or E( Z 1 ) > R, i.e., the expected return on the "risky" asset must exceed the certain
* *
return R . Note: the expected return on the portfolio = E(Z) = E[ w ( Z 1 - R) + R] > R for w > 0

only if E( Z 1 ) > R .

Mean-Variance Portfolio Selection and the Effects of Diversification

While in general, the expected utility maxim requires knowledge of the complete joint
distribution of asset returns to determine the optimal portfolio, under certain conditions, it is
sufficient to know only the first two moments of the joint distribution. That is, the criterion
function for choice can be written as a function of just the expected value (mean) of end-of-
period wealth and the variance of end-of-period wealth V[E( W 1 ) ,Var( W 1 )] where V is an
increasing function of its first argument and a decreasing function of its second argument).
Under these conditions, the choice problem is called the Mean-Variance portfolio selection
problem. While beyond the scope of these Notes, it can be shown that if the time interval
between successive portfolio revisions is small, the optimal portfolio choice can be well-
approximated by the mean-variance problem's solution. For the balance of these Notes, we shall
focus exclusively upon environments in which the mean-variance criterion is appropriate.

189
 Robert C. Merton

The mean-variance model is the first step in introducing uncertainty quantitatively into
the ranking of portfolios or investments. Classical methods of ranking investments use a single
parameter measure such as (expected) rate of return or (expected) present discounted value.
Although such one parameter measures make ranking quite easy (highest to lowest), they clearly
do not reflect differences among alternatives due to uncertainty. With a two parameter ranking,
there is no simple ranking like highest to lowest. The purpose of the mean-variance model is to
determine optimal portfolios and to make explicit, the tradeoff between risk and return.

A digression on some characteristics of probability distributions

1. The first moment or expected value or the mean of the random variable X which can
take on value x1 with probability P1 ; value x 2 with probability P2; ...; value x n with
probability P n is defined as

n
E(X) = X = ∑ Pi xi
i=1

2. The second (central) moment or variance of the random variable X is defined by

n
σ = E[(X - X ) ] = ∑ Pi ( xi - X ) .
2 2 2
X The standard deviation is a measure of the
i=1

dispersion of possible outcomes around the expected value, as is illustrated in Figure X.1.

190
Finance Theory

Figure X.1.a Figure X.1.b

In Figure X.1, the distribution in (b) is more disperse than in (a), and (b) has a larger
standard deviation (and variance) than (a). Note: in the special case when there is only
one possible outcome for X, call it y, then E(X) = X = y and

σ 2X = 0 = σ X . An alternative useful formula for σ 2X is:

n n
σ = ∑ P i( x i -X ) where X = ∑ Pi x i
2 2
X
i=1 i=1
n n n n
= ∑ P i( x i2 -2 x i X + X 2 ) = ∑ P i xi2 - 2X ∑ Pi x i + X 2∑ P i
i=1 i=1 i=1 i=1
n
= ∑ P i xi2 - 2 X 2 + X 2 = E[ X 2 ] - ( X )2
i=1

3. The covariance between two random variables X and Z is defined by

σ XZ = Cov(X, Z) = E{(X - X )(Z - Z )}

n m
= ∑ ∑ Pij ( xi - X )( Z -Z )
j where Pij is the probability that
i=1 j=1

191
 Robert C. Merton

X = xi and Z = Z j . It is a measure of the co-relationship between variations in possible

outcomes of the two random variables X and Z around their expected outcome.
Note: (1) the covariance is symmetric, i.e.,

Cov(X, Z) = E{(X - X )(Z - Z )}

= E{(Z - Z )(X - X )}
= Cov(Z, X)
(2) the covariance of a random variable with itself is the variance, i.e.,

Cov(X, X) = E{(X - X )(X - X )} = E{(X - X )2 } = σ 2X

(3) if X and Z are independent, then Cov(X, Z) = 0. I.e., if they are

X Z
independent, then Pij = P i P j where

P iX = prob{X = X i } and P j = prob{Z = Z j } Hence,

Z

n m
Cov(X, Z) = ∑ ∑ Pij ( xi - X )( Z -Z ) =
j

[∑ ][ ]
i=1 j=1
n m

P X
i ( x - X ) ∑ P ( Z -Z )
i
Z
j j
i=1 j=1

n n n

but ∑ PiX ( xi - X
i=1
)= ∑ P
i=1
X
i xi - X ∑ PiX =
i=1
X - X =0
so Cov(X, Z) = 0.
4. The correlation coefficient between the random variables X and Z is defined by

= σ XZ . Roughly, its absolute magnitude represents the percentage of the

Cov(X, Z)
ρ XZ ≡
σ XσZ σ XσZ
variation in X explained by the variation in Z . Its sign says whether X and Z tend to
move in the same or opposite directions. Note: ρ XZ = ρ ZX ; | ρ XZ | ≤ 1 . If ρ XZ = 1, then X

and Z are perfectly positively correlated and if ρ XZ = − 1 , then X and Z are

perfectly negatively correlated. If X and Z are independent, ρ XZ = 0 and X and Z

are said to be uncorrelated. Of course, ρ XX = 1 .

192
Finance Theory

Characteristics of Portfolios and the Effects of Diversification

Consider the case of n assets with (random) variable returns (per dollar) Z i . As done in the general
th
case, if wi = fraction (of each dollar invested in the portfolio) invested in the i asset, then we can

define Z = ∑1 wi Z i as the (random) variable return (per dollar invested) on the composite security or
n

portfolio, with proportions ( w1 , w2 ,..., wn ) invested in each asset. The expected return on the
n n n

portfolio, E ( Z ) = Z i , is E ( Z ) = E[∑ wi Z i ] = ∑ wiE (Z i) = ∑w Z i i, the weighted sum of the

1 1 1

expected returns on the individual assets. Let σ i be the variance of the return on the i
2 th
asset (i.e.,
2
σ i2 = E{[ Z i - Z ] } and σ ij be the covariance between the returns on the ith and jth assets (i.e.,

σ ij ≡ E{( Z i - Z i )( Z j - Z j )} and σ ii = σ i2 ). Then, the variance of the portfolio can be computed to be

n
σ 2 = E[(Z - Z ) ] = E[( ∑ wi Z i - Z ) ] .
2 2

i=1

n n
Note : ∑wi( Z i - Z i ) = ∑wi Z i - Z
i=1 i=1

(∑ ( ) (∑ )
2
⎡∑wi( Z i - Z i )⎤
n n n
= wi Z i - Z i ) w j( Z j - Z j )
⎣ i=1 ⎦ i=1 j=1
n n
=∑∑ wi w j( Z i - Z i )( Z j - Z j )
i=1 j=1
n n n
Hence, E { [ ∑ wi ( Z i - Z i ) ] 2
}= E [ ∑ ∑ w w ( Z i j i - Z i )( Z j - Z j )]
i=1 i=1 j=1
n n
= ∑ ∑ wi w j E[( Z i - Z i )( Z j - Z j )]
i=1 j=1
n n
= ∑ ∑ wi w j σ ij = σ 2 .
i=1 j=1

193
 Robert C. Merton

Example: The effects of diversification

Suppose there are two risky assets with (random) variable returns Z 1 and Z 2 , and suppose that
E( Z 1 ) = Z 1 = E( Z 2 ) = Z 2 = m , i.e., they have the same expected return; suppose that

E[( Z 1 - Z 1 )2 ] = σ 12 = E[( Z 2 - Z 2 )2 ] = σ 22 = v 2 , i.e., they have the same variance of return. Let

σ 12 = E{( Z 1 - Z 1 )( Z 2 - Z 2 )} = covariance of the returns which can be written in terms of the

correlation coefficient, ρ 12 , between asset #1 and #2, as σ 12 = ρ 12 σ 1 σ 2 = ρ 12 v . Let Z = return

2

on the portfolio mix of #1 and #2: Z = w1 Z 1 + w2 Z 2 (for w1 + w2 = 1 ). The expected return on Z

can be computed as

Z = E(Z) = E( w1 Z 1 + w2 Z 2 ) = w1 E( Z 1 ) + w2 E( Z 2 ) = w1 m + w2 m
= ( w1 + w2 )m = m .

Hence, in this example, for any mix ( w1 , w2 ) , the expected return on the portfolio will be the same,
namely m. What about the variance of Z ?
2 2
σ = E[(Z - Z ) ] = E[( w1 ( Z 1 - m) + w2 ( Z 2 - m) ) ]
2

= E[ w12 ( Z 1 - m )2 + 2 w1 w2 ( Z 1 - m)( Z 2 - m) + w22 ( Z 2 - m )2 ]

= w12 E[( Z 1 - m )2 ] + 2 w1 w2 E[( Z 1 - m)( Z 2 - m)] + w22 E[( Z 2 - m )2 ]
= w12 σ 12 + 2 w1 w2 σ 12 + w22 σ 22
= w12 v2 + 2 w1 w2 ρ 12 v 2 + w22 v2 = v 2 [ w12 + 2 w1 w2 ρ 12 + w22 ].

Hence, as the relative proportions ( w1 , w2 ) are varied, the variance of the portfolio is changed.

Since w1 + w2 = 1, to see the effect on σ , we first substitute w1 = w and w2 = 1 - w1 = 1 - w, to get

2

σ 2 = v 2 [ w2 + 2w(1 - w) ρ 12 + (1 - w ) ] = v 2 [1 - 2(1 - ρ 12 )w(1 - w)] .

2

So σ , as a function of the "mix parameter" w, is a parabola. Since the expected return is the same
2

for all mixes, a risk-averse investor would want to choose the portfolio with the smallest variance

194
Finance Theory
*
(dispersion). Formally, we can calculate the "variance minimizing" mix, w , by using the
calculus as follows:

min σ
2
= min [ v 2 [1- 2(1- ρ 12 )w(1- w)]]
w w
dσ 2
1
|w=w* = 0 = - v 2 [2(1- ρ 12 )(1- 2 w* )] or w* =
dw 2
which is not exactly a big surprise because of the symmetry of the problem. However, this
general technique is applicable even if the individual variances are not equal. Figure X.2
presents the graph of σ
2
for various correlation coefficients and mixes. From above, the minimum
1
variance (corresponding to w = ) for a given ρ 12 is
2
2
v
σ min
2
= (1 + ρ 12 )
2

195
 Robert C. Merton

Figure X.2

As Figure X.2 shows, as long as the two assets are not perfectly positively correlated, the investor
can lower the variance of his return by mixing. This phenomenon is called the diversification
effect. Note that the less positively correlated are the returns, the greater the effect. If they are
independent (i.e., ρ 12 = 0 ), then the variance is halved. Although negative correlation is even better,
the existence of such assets in the real world is rare. Although this example was very specialized, the
diversification effect holds generally. Hence, risk-averters will tend to diversify if they act rationally.
While diversification is not a new idea (or rule), our systematic approach will allow us to measure
quantitatively how much diversification is provided by adding securities to a portfolio and how much the
investor should diversify. Later, these quantitative results will lead to a number of new qualitative
insights.

Constructing Portfolios and Composite Securities

196
Finance Theory

Consider the two-risky asset case of the previous example, but now allow
E( Z 1 ) = Z 1 and E[ Z 2 ] = Z 2 not to be equal. Further, assume

E{( Z 1 - Z 1 )2 } = σ 12 and E{( Z 2 - Z 2 )2 } = σ 22 are not equal, and by convention, assume that

σ 22 > σ 12 . Form a composite security from a combination of security #1 and security #2. Denote its
random variable return by Z and its expected return by Z and variance by σ . Let w = fraction
2

of each dollar of the composite security invested in security #2; 1 - w = fraction invested in #1. Then

Z = E(w Z 2 + (1 - w) Z 1 ) = w Z 2 + (1 - w) Z 1 = Z 1 + w( Z 2 - Z 1 ). If Z 2 > Z 1 , then

Figure X.3.a

If Z 2 < Z 1 , then

197
 Robert C. Merton

Figure X.3.b

The variance of Z,σ , is given by

2

σ = w σ 2 + 2w(1 - w)ρ 12σ 1σ 2 + (1- w ) σ 1

2 2 2 2 2

= [ σ 22 + σ 12 - 2 ρ 12σ 1σ 2 ] w2 + 2[ ρ 12σ 1σ 2 - σ 12 ]w + σ 12

(i.e., Z ,σ ) as w
2
The rest of this section is devoted to finding the characteristics of Z
changes and for different assumptions about Z 1 and Z 2 .

Case 1: Suppose that Z1 and Z 2 are perfectly (positively or negatively) correlated, i.e.,
ρ 12 = + 1, then

2 2
σ 2 = w 2σ 22 + 2w(1- w)σ 1σ 2 + (1- w ) σ 12 = [wσ 2 +(1- w)σ 1 ]

or

σ = | wσ 2 +(1- w)σ 1 |

i.e., σ is in a linear relationship to σ 1 and σ 2 ; and linear in

w for ρ 12 = + 1.

198
Finance Theory

Figure X.4a

The only other case where σ is in a linear relationship to σ 2 and σ 1 is when σ 1 = 0, i.e.,

security #1 is riskless. Then σ = w σ 2 and σ = | w | σ 2

2 2 2

Figure X4.b

Proposition: σ 22 + σ 12 - 2 ρ 12σ 1σ 2 ≥ 0 and = 0 only if ρ 12 = 1.

199
 Robert C. Merton

Proof: 1. clearly, if ρ 12 ≥ 0, then σ 22 + σ 12 - 2 ρ 12σ 1σ 2 > 0

suppose for ρ 12 > 0, it were possible that σ 2 + σ 1 - 2 ρ 12 σ 1σ 2 < 0.

2 2
2.

Then, σ 12 + σ 22 < 2 ρ 12 σ 1 σ 2 . But ρ 12 ≤ 1, so it must be that σ 12 + σ 22 < 2 σ 1 σ 2 or

σ 12 + σ 22 - 2 σ 1σ 2 < 0. But that means that ( σ 1 - σ 2 )2 < 0 which cannot be. Hence,

σ 12 + σ 22 - 2 ρ 12σ 1σ 2 ≥ 0.

Suppose σ 1 + σ 2 - 2 ρ 12 σ 1σ 2 = 0. Then
2 2
3.

σ 12 + σ 22 - 2 σ 1 σ 2 = 2( ρ 12 - 1)σ 1 σ 2 , or ( σ 1 - σ 2 ) = 2( ρ 12 - 1) σ 1 σ 2 . But ρ 12 ≤ 1 and

2

( σ 1 - σ 2 )2 ≥ 0. So the only way for equality to hold is if ρ 12 = 1 and σ 1 = σ 2. .

Case 2: Z 1 and Z 2 are not perfectly correlated, i.e., ρ 12 < 1.

σ 2 = [ σ 22 + σ 12 - 2 ρ12 σ 1 σ 2 ] w2 - 2[ σ 12 - ρ12 σ 1 σ 2 ]w + σ 12
σ = [ ] w2 - 2[ ]w + σ 12
and

1st derivative of σ with respect to w

dσ2
(X.7a) = 2 {w[ σ 12 + σ 22 - 2 ρ 12 σ 1 σ 2 ] - [ σ 12 - ρ 12 σ 1 σ 2 ]}
dw
dσ w[ σ 12 + σ 22 - 2 ρ 12 σ 1 σ 2 ] - [ σ 12 - ρ 12 σ 1 σ 2 ]
(X.7b) =
dw σ

2nd derivative of σ with respect to w

200
Finance Theory

d 2σ 2 = 2[ 2 + 2 - 2
(X.8a) σ 1 σ 2 ρ 12σ 1σ 2 ] > 0, from the Proposit ion
dw2

d σ σ 1σ 2
2 2 2
(X.8b) = (1 - ρ 12 ) > 0 for ρ 12 < 1
dw
2
σ2

From the formulae for σ and σ , σ as a function of w is a parabola and σ as a function

2 2

of w is a hyperbola. From (X.8a) and (X.8b), both σ and σ are strictly convex functions
2

of w.

The minimum variance composite security with "mix" (wmin,1-wmin)

Because σ
2
is a (strictly) convex parabola, there exists a unique minimum value corresponding
to proportion, wmin . Of course, this value of w minimizes σ as well. The minimum point

st dσ 2
will occur where the 1 derivative is zero, i.e., | = 0. From (X.7a) or (X.7b) we
dw w=wmin
have that

( σ 12 - ρ 12 σ 1 σ 2 )
(X.9) wmin =
[ σ 12 + σ 22 - 2 ρ 12 σ 1 σ 2 ]

dσ2 dσ
Thus, for w < wmin , < 0 (and < 0) and for
dw dw
dσ2 dσ
w > wmin , > 0 ( and > 0). Can wmin > 1 ? Suppose so, then from (X.9)
dw dw
σ 12 - ρ 12 σ 1 σ 2
wmin = > 1 or ρ 12 > σ 2 . But, by convention,
[ σ 1 + σ 2 - 2 ρ 12 σ 1 σ 2 ]
2 2
σ1

σ 22 > σ 12 or σ 2 > σ 1 . So, ρ 12 > 1 which is impossible. Hence, wmin < 1. Can wm < 0? From
σ1 σ1 >
(X.9), this would imply that ρ 12 > which is possible since σ 2 > σ 1 . Thus, if ρ 12 ,
σ2 σ2

201
 Robert C. Merton

σ1>
then 0 < w min < 1 (Note : if ρ 12 ≤ 0, then ρ 12 ). So, if σ 2 is not too much larger than σ 1
σ2
or if Z1 and Z 2 are not too highly (positively) correlated, then 0 < wmin < 1. Given (X.7b),
(X.8b), and (X.9), we can graph σ as a function of w as:

Figure X.5

Note in Figure X.5: because σ is a convex function of w, the curve will always lie below the
straight line (w = 0,σ 1 ) and (w = 1,σ 2 ). For our purposes, it will be much more useful to work

with the relationship between σ and Z rather than with σ and w , (i.e., combine graphs
Figures (X.3a) or (X.3b) with (X.5)). Because Z = Z 1 + w( Z 2 - Z 1 ), we have by the "chain rule"
(valid for Z 2 ≠ Z 1 ) that

dσ dσ
dσ dσ dw
(X.10) = = dw = dw
d Z dw d Z d Z ( Z 2 - Z 1 )
dw

202
Finance Theory

and
d σ
2

d σ
2 2
= dw .
(X.11)
d Z 2 ( Z 2 - Z 1 )2

dσ dσ
From (X.10), if Z 2 > (<) Z 1 , then will have the same (opposite) sign as . From
dZ dw

d σ d σ
2 2
(X.11), independent of the relative sizes of Z 2 and Z 1 , 2 will have the same sign as 2
.
dZ dw

d σ d 2σ
2
But, from (X.8b) > 0. So, > 0 and σ is a convex function of Z . Graphically, the
dw
2
dZ2

two cases are:

Figure X.6a

203
 Robert C. Merton

Figure X.6b

By convention, graphs such as Figures (X.6), are plotted with the expected return on the ordinate
and standard deviation on the abscissa:

Figure X.7

σ min occurs where

Note: dσ dZ
= 0 or where =∞
dZ dσ

The curve in Figure (X.7) traces out all the feasible expected return-standard deviation (or mean-
variance) combinations possible from the two risky assets. And to each ( Z ,σ ) point on that
curve, there corresponds a unique portfolio of these assets described by (w,1 - w).

204
Finance Theory

General Composite Securities

The preceding analysis examined the simplest composite security constructed from two
securities. We now analyze composite securities constructed from many assets.

First, consider composite security #I constructed from securities with return Z 1 and Z 2 , where

Z 1 has mean, variance, and covariances ( Z 1 ,σ 1 ,σ 12 ,σ 13 ,σ 14 ) and Z 2 has mean, variance,

2

and covariances ( Z 2 ,σ 2 ,σ 21 ,σ 23 ,σ 24 ). If w is the fraction of composite security #I invested

2 I

in security 1 and (1 - wI ) is the fraction invested in security #2, then the return on #I,
I I
Z I , is Z I = w Z 1 + (1 - w ) Z 2 and Z I has an expected return ZI and variance σ 2I as
constructed in the previous section.

Second, consider the composite security #II constructed from other securities with returns
Z 3 and Z 4 , where Z 3 has mean, variance, and covariances ( Z 3 ,σ 3 ,σ 31 ,σ 32 ,σ 34 ) and Z 4
2

has mean, variance, and covariances ( Z 4 ,σ 4 ,σ 41 ,σ 42 ,σ 43 ). If w is the proportion of the

2 II

II
composite security #II invested in Z 3 and (1 - w ) is the proportion invested in Z 4 , then the

return of #II, Z II = w Z 3 + (1 - w ) Z 4 , and Z II has expected return Z II and variance σ II as

II II 2

constructed in the previous section.

Third, one can compute the covariance between composite securities #I and #II,
Cov( Z 1, Z II ) ≡ σ I,II , from knowledge of the variances and covariances of Z 1 , Z 2 , Z 3 , and Z 4 .

Fourth, form a composite security with return Z, constructed from (composite) securities I and
II with returns Z I and Z II , where ZI has expected return, variance, and covariance
a( Z I ,σ 2I ,σ I,II ). If w is the fraction of the composite security invested in security #I and

205
 Robert C. Merton

(1 - w) is the fraction invested in security #II, then Z = wZ I + (1 - w) Z II , and the mean and

variance as a function of w can be computed as was done in the previous section. Further,
if σ I > 0 and σ II > 0 and Z I and Z II are not perfectly correlated, then Figures (X.5) -(X.7)
2 2

will describe the mean-variance "tradeoff." Otherwise, Figure (X.4) will be the description.
Note:

Z = w[ w I Z 1 + (1- w I ) Z 2] + (1- w)[ w II Z 3 + (1- w II ) Z 4]

= ( ww I ) Z 1 + ( w(1- w I )) Z 2 + ((1- w) w II ) Z 3 + [(1- w)(1- w II )]Z 4
= µ 1Z 1 + µ 2 Z 2 + µ 3Z 3 + µ 4 Z 4

So, we see that composite securities containing many securities can be constructed by combining
securities to form composite securities and combining these composite securities to form (more
complicated) composite securities, etc. Hence, we can generate any portfolio by this process and
each portfolio will have an expected return and variance. Further, the graph of the mean-
variance "tradeoff" will be like either Figures (X.5) - (X.7) or Figure (X.4) as in the two-security
case.

Portfolios and Efficient Portfolios

Suppose that there are n securities with (random) variable returns Z i with expected returns
E( Z i ) = Z i ; variances of returns σ i2 ; covariances of returns σ ij , for i, j = 1,2,..., n . Further,

suppose that all n securities have uncertain returns (i.e., σ i > 0 for all securities). By mixing
2

these securities together to form portfolios, one can create "new" (composite) securities which
will also have expected returns, variances, and covariances with the other (both "basic" and
"composite") securities. Hence, one can create an infinite number of securities from the original
n. Is there a way to reduce the number of securities (or portfolios) that one need consider as
possibilities for selected portfolios? We know that, asked to choose a portfolio from a group of
portfolios all with the same expected return, risk-averse mean-variance maximizers will choose

206
Finance Theory

the portfolio with the smallest variance. Suppose that we classify or subdivide all the possible
portfolios into groups where each portfolio within a given group has the same expected return;
then determine, for each group, which member has the smallest variance. The collection of
"winner" portfolios from each group is called the Frontier portfolio set. A portfolio is a member
of the Frontier portfolio set if and only if among all portfolios possible with the same expected
return, it has the smallest variance. Clearly, it is a necessary condition that a portfolio be a
Frontier portfolio if it is ever going to be chosen by a risk-averter (as an optimal portfolio). We
can reduce the possibilities even more: given a choice between two portfolios with the same
variance, a risk-averter will prefer the one with the larger expected return. So, among Frontier
portfolios, compare all portfolios with the same variance and select the one with the largest
expected return. This final collection of portfolios is called the Efficient Portfolio Set. A
portfolio is a member of the Efficient portfolio set if and only if there does not exist another
portfolio which has a variance smaller or equal to its variance and which has an expected return
greater than or equal to its expected return. Clearly, any portfolio selected by a risk-averter (as an
optimal portfolio) must be an efficient portfolio.

207
 Robert C. Merton

Figure X.8

In Figure X.8, the cross-hatched area represents feasible (possible) portfolios; the boundary line
(which is a parabola) is the Frontier portfolio set; the heavy-lined part of the boundary is the
Efficient portfolio set. The point ( Z min ,σ min ) is called the minimum-variance portfolio, and is
2

a part of the Efficient portfolio set. As noted, it is common practice to plot the portfolio sets in
Expected Return-Standard deviation space where it is a hyperbola.

208
Finance Theory

Figure X.9

We now present an analytical derivation of the Frontier and Efficient Portfolio sets to show that
the qualitative results presented in Figures X.8 and X.9 are correct and to demonstrate that in
practice, given the expected returns, variances, and covariances of the primary securities, the
efficient frontier can be computed.
Let Z be the random variable return on any portfolio (constructed from the n
"primary" securities) which has expected return m , i.e.,
n

(X.12) Z = ∑ wi Z i
i=1

where the portfolio weights are restricted to satisfy

209
 Robert C. Merton

n
1 = ∑w
i=1
i

(X.12a)

and
n n

(X.12b) Z = E(Z) = ∑ wi E( Z i ) = ∑ wi Z i = m.
i=1 i=1

Obviously, all possible combinations of w1 , w2 ,..., wn which satisfy the constraints (X.12a) and
(X.12b) represent all the possible portfolios with expected return m. To find the Frontier
portfolio set, we must determine the particular combination ( w1 , w2 ,..., wn ) which satisfies
constraints (X.12a) and (X.12b) and minimizes the variance. Formally, this is a constrained
σ2
minimization problem which can be solved by using Lagrange multipliers, i.e., minimize
2
subject to (X.12a) and (X.12b), or

1 n n n n

(X.13) Min{ ∑ ∑ wi w j ij 1 ∑
2 i=1 j=1
σ + λ [1 -
i=1
wi ] + λ 2 [m - ∑
i=1
wi Z i ]}

n n

where λ1 and λ2 are the multipliers and remember that σ = ∑ ∑ wi w j σ ij . To determine a

2

i=1 j=1

critical point, we differentiate (X.13) with respect to w1 , w2 ,..., wn , λ 1 , λ 2 and set each partial
derivative equal to zero, to obtain the optimality conditions
n
0 = ∑ w j σ ij - λ 1 - λ 2 Z i for i = 1,2,..., n
j=1
n

(X.14) m = ∑ wi Z i
i=1
n
1 = ∑ wi
i=1

These are (n + 2) linear equations to be solved for the (n + 2) unknowns

w1 , w2 ,..., wn , λ 1 , and λ 2 . Let [ vij ] be the elements of the inverse of the variance

210
Finance Theory

-covariance matrix of returns [ σ ij ] . Then, if we call:

n n n n n n
A ≡ ∑ ∑ vij Z j ; B ≡ ∑ ∑ vij Z i Z j ; C ≡ ∑ ∑ vij ; D ≡ BC - A , 2
the solutions are:
i=1 j=1 i=1 j=1 i=1 j=1

n n
m ∑ (CZ j - A) + ∑v ij(B - AZ j )
v ij
j=1 j=1
wi =
(X.15) D
Cm- A B- Am
λ2 = λ1 =
D D
From (X.15), we can compute the portfolio variance, σ , to be
2

2
Cm - 2Am + B
(X.16) σ =
2
Note: the variance of the Frontier portfolio set is a
D
parabola as a function of the expected return, m.

One can solve for the expected return as a function of the standard deviation [using (X.16)] to be:

A 1
(X.17a) m= + D(Cσ 2 - 1) Frontier
C C
and the Efficient (part of) Frontier to be

A 1
(X.17b) m= + D(C σ 2 - 1) Efficient Set
C C
More on the Role of Financial Instruments and Intermediaries: A Mutual Fund Theorem

The previous analysis assumed that all the securities available were risky (i.e., σ i > 0 ). What
2

st
happens if a (n + 1 ) riskless security becomes available with (certain) return R? Before
answering that question, we digress:

Digression: Suppose that you already have a composite security or portfolio (containing only
risky assets) with (random) variable return Z P ; expected return E( Z P ) = Z P ; variance of

return E[( Z P - Z P )2 ] = σ 2P . Suppose that there is now available a riskless security with

211
 Robert C. Merton

(certain) return R, and you want to construct a new portfolio by combining the "old" portfolio
with the riskless security. Let w = fraction of your wealth invested in the "old" (risky) portfolio
and (1 - w) = fraction of your wealth invested in the riskless asset. If Z is the (random)
variable return on the new portfolio, then

Z = w( Z P ) + (1 - w)R = w( Z P - R) + R

and the expected return is E(Z) = Z = w( Z P - R) + R.

Note: Z - Z = w( Z P - R) + R - w( Z P - R) - R = w( Z P - Z P ) . The variance of the new portfolio is

2 2 2
σ 2 = E[(Z - Z ) ] = E[ w2 ( Z P - Z P ) ] = w2 E[( Z P - Z P ) ] = w2 σ 2P . The standard deviation of

the new portfolio, σ , is σ 2 or

σ =| w |σ P
Note: the standard deviation is linear in the "mix" w, as was shown earlier.

Figures X.10 illustrate how the variance, standard deviation, and expected return vary as one
alters the mix.

212
Finance Theory

Figure X.10.a Figure X.10.b

Figure X.10.c

We can also solve for the expected return as a function of the standard deviation: since
σ = | w | σ P , if w ≥ 0 , then,

( - R)
Z = w( Z P - R) + R = Z P σ + R . Graphically,
σP

213
 Robert C. Merton

Figure X.11

The important point to remember from this analysis is that various combinations of a risky
security with a riskless security plot as straight lines in the Expected Return - Standard Deviation
plane (Figure X.11).
- End of Digression -

Return now to the question posed before the digression: What is the effect on the efficient
portfolio frontier of adding a riskless security? Using the result displayed in Figure X.11, we can
determine the answer geometrically (as is done in Figure X.12) by combining Figure X.11 with
the "old" frontier for (risky) assets as displayed in Figure X.9.

214
Finance Theory

Figure X.12

The curve DBE is the "old" efficient frontier when only risky assets were available. In
particular, the point B is a portfolio which contains only risky assets because it lies on DBE.
Think of this specific portfolio as the "old" risky portfolio analyzed in the digression (i.e., take

Z P = Z and σ P = σ ). In that case, line ABC in Figure X.11 corresponds exactly to the line in
* *

Figure X.10, and it represents various (possible) positive mixes of the "old" portfolio with the
riskless security. Therefore, every point on line ABC is now a feasible portfolio with the
introduction of the (additional) riskless security. Note that every point on ABC is (strictly,
except for point B) above points on DBE, and hence, the new efficient portfolio frontier is the
straight line ABC. Thus, every portfolio in the efficient portfolio set can be interpreted as a
"mix" of two portfolios: namely, a portfolio containing only risky securities in the proportions
described by point B and a (trivial) portfolio containing only the riskless security (point A).
*
Because of the importance of the particular portfolio ( Z ,σ ) represented by point B, the
*

* * *
specific weights of the holdings of basic securities in that portfolio, ( w1 , w2 ,..., wn ) are called

215
 Robert C. Merton

the optimal combination of risky assets. Further we can determine explicitly what these optimal
proportions are from the (formal) analysis previously done. The proportions are (using the
notation of that analysis)
n

∑ v ( Z - R)
ij j

(X.18) wi =
j=1
, i = 1,2,...,n
(A- RC)
We now summarize: (I) we know that risk-averters in selecting an optimal portfolio from
among the (n + 1) individual securities will always choose a portfolio which lies along the
Efficient Portfolio Frontier (line ABC in Figure X.11), which is a straight line with slope
( Z * - R)
and intercept R ; (II) we know that every efficient portfolio can be constructed by
σ
*

mixing two particular portfolios (or "mutual funds"), and that one "fund" holds just the riskless
asset and the other holds only risky assets in the proportions described in (X.18); (III) the
proportions described in (X.12) depend only on the expected returns, variances, and covariances
of the "primary" securities and require no other information to compute.

*
Note: the proportions, wi , in (X.18) do not depend on the individual investors' utility functions
or on how much wealth they have.

A "Mutual Fund" or "Separation" Theorem

Every risk-averse, mean-variance utility maximizer would be indifferent between

selecting his optimal portfolio from among the original (n + 1) securities or from just the two
mutual funds described in (II), provided that the investor agrees with the estimates of ( Z i ,σ ij )
used to form the (optimal) risky mutual fund.
Proof: follows from (I) and (II) and the definition of the efficient portfolio set.

216
Finance Theory

Remember the first such separation theorem was deduced in an earlier set of lectures where it
was shown that the individual investors could hire a "technocrat" to make all production
decisions, and provided that he followed the "right" rule (i.e., maximize market value), they
would be indifferent between his handling production or each of them doing it individually.
Here, we find that all the individual risk-averse investors can hire a "technocrat" portfolio
manager and give him the rule to hold proportions w*i in his fund, and the only decision that the
individual investor need make is what proportion of his wealth to hold in the riskless asset.
(Essentially, the problem solved in the digression). It is a true separation or decentralization
because the portfolio manager need only "worry" about determining the expected returns,
variances, and covariances of the individual securities and need not know what the investors'
preferences or wealth levels are to do his job; and the investors do not need to know the
individual expected returns, variances, covariances, etc. of the securities, but only the aggregate
( Z * , σ * ) to make their decisions.

How the Investor selects the optimal "mix" between the two funds: A Graphical Solution. If the
investor makes his decisions solely on the basis of the mean and variance of his portfolio and if
he is risk-averse, then one can solve for indifference curves (lines of constant utility level)
showing the tradeoff between expected return and standard deviation (or variance), and these
curves will have a shape as displayed in Figure X.12.

217
 Robert C. Merton

Figure X.13

His individual optimal portfolio will be the point where one of his indifference curves is tangent
to the Efficient Frontier, and ( Z optimal ,σ optimal ) are the expected return and standard deviation of

the return on his optimal portfolio. From the lower half of the graph, we see that implies putting
woptimal percent of his wealth in the "risky" fund and the rest in the riskless asset. Note: He
*
only required knowledge of ( Z ,σ , R) to choose his optimal portfolio.
*

218
 Domain of Investment Management: Stages of Production Process

Passive
Well-Diversified
Efficient
Portfolio
“Efficient Exposures” Super
Efficient
Portfolio of
Risky Assets
Active
Asset-Class Client
(Optimal Alter Shape of Structured
Allocation Optimal Households
Combination of Payoffs on Efficient Form
Macro Sector Portfolio of Entrepreneurs
Risky Assets) Underlying of Payouts to
Market Timing Assets Endowment
Optimal Portfolio Client
Corporation

Superior Performing
Riskless (Derivative
Micro Aggregate
Asset Securities with
Excess-Return
Portfolio Non-Linear
Portfolio
“Alpha Engines” Payoffs)

• Risk Modulation • Tax efficient

• Components of •Risk Modulation
through Insurance • Regulatory efficient
Best Performing through Hedging or
or non-linear • Liquidity allocation
Risky Assets Leveraging
leverage
Only Portfolio:
• Pre-programmed
•Diversification •Market Timing dynamic trading
Risk Modulation Active Management
• “Building Block”
State-Contingent
Securities to create
specialized payout
patterns

219
 Robert C. Merton

Passive Management: Efficient Exposures to Various Asset Classes

Macro Asset Classes

Small-Cap Mid-Cap Large-Cap Fixed-Income Real Estate Other

Domestic Domestic Domestic
Equities Equities Equities

Weighted to match a benchmark

Passive
Well-Diversified
Efficient
Portfolio

Implementing Diversification as one of the Three Risk Management Tools

Indexing of portfolios

220
Finance Theory

Active Management: Enhancing Portfolio Performance

Asset-Class Allocation: Macro-Sector Market Timing

“Long-Short” combinations to change fractional allocations from Benchmark Weights

ASSET CLASS BENCHMARK WEIGHT LONG (SHORT) REVISED WEIGHT

INCREMENTAL
Small-Cap Equity 5% +5% 10%
Mid-Cap Equity 10% 0% 10%
Large-Cap Equity 30% (10%) 20%

Emerging Market 15% (5%) 10%

Equity
Domestic Fixed- 30% 5% 35%
Income
Real Estate 10% 5% 15%

100% 0% 100%

Micro “Excess Return” Portfolio: Security Selection: “Alpha Engines”

Engine #1 Engine #2 Engine #3 Engine #4 Engine #5 Engine #N
U.S. Risk Technical Fundamental Foreign Currency Private Equity Mortgage-back
Arbitrage Analysis of Analysis of Forecast Fund Fund Security
Hedge Fund Equities Fund Equities Fund Relative Value Fund

Optimal Weighting
•Security Analysis Super-Performing
Micro Aggregate
•Technical Analysis Excess-Return
•Proprietary Derivative-Security Pricing Models Portfolio

221
 Robert C. Merton

Creating the Optimal Portfolio of Assets:

Mix of Optimal Combination of Risky Assets (“OCRA”) and the Riskless Asset

Expected Reward

Risk/Expected Reward “Menu”

OCRA Reward

Optimal Portfolio
Reward
Riskless Reward

0 Risk
Optimum
Optimal Risk OCRA Risk
Percentage of Percent
Optimal Portfolio
Invested in OCRA
100%

• Hedge or Leverage OCRA to obtain Optimal Portfolio

• Implement Macro Market timing of Risky versus Riskless Asset Performance

222
Finance Theory

Transform Shape of Payoffs from Investing in the Optimal Portfolio: Derivatives

Value of Investor
Insured Portfolio, $
“Uninsured Equity” Payoff
• Insurance and non-linear leverage
• Transform Payoff Pattern to fit precise preferences: custom design

$95,000 Minimum
Guarantee Floor
“Insured Equity” Payoff

0
Value of Optimal Portfolio, $
0
$100,000

Value of Investor
Custom Pattern
Portfolio , $
$190,000 “Ceiling” Maximum Payout

$95,000 Minimum
Guarantee Floor

0
Value of Optimal Portfolio, $
0
$95,000 $190,000

223
 Robert C. Merton

Structured Holdings to Create Most Efficient Form of Payouts to Client

• Tax efficient (income, wealth, estate/inheritance)

• Regulatory efficient
• Liquidity efficient

Tools

Derivatives: Futures, Forwards, Swap Contracts

Special Purpose Vehicle (SPV): Custom-created targeted-purpose security
Asset Substitution:
Municipal (tax-exempt) bonds for taxable bonds
liquid “on-the-run” US Treasury Bonds for “off-the-run” less-liquid US Treasury or Agency bonds
Location of Entity: (e.g., Bermuda for insurance)
Location of Assets and Liabilities: on or off-balance sheet; investment versus trading account; taxable or non-taxable
part of one’s accounts.

224
XI. IMPLICATIONS OF PORTFOLIO THEORY FOR THE OPERATION OF THE
CAPITAL MARKETS: THE CAPITAL ASSET PRICING MODEL

We have shown that for all risk-averse, mean-variance utility-maximizers who agree on
the expected returns, variances, and covariances of the individual basic securities, the optimal
portfolio chosen can be represented as a mix of two securities (portfolios): one security is the
riskless security with return R and the other is a particular combination of risky assets. We now
consider the implications of these results for equilibrium expected returns and asset prices.
Suppose that everyone in the market agreed on expectations. Then, if the market is in
equilibrium (i.e., the prices of securities are such that when investors are holding their optimal
portfolios, the aggregate supply of each security is equal to the aggregate amount of that security
demanded), what must be the composition of the "risky" portfolio represented by point B in
*
X.11 in Section X (i.e., the optimal combination of risky assets with mean Z and variance

σ *2 )? In Section X, it was shown that all investors would be indifferent between selecting an optimal
portfolio from the n risky assets and the riskless asset and from just two assets: the "risky"
mutual fund composed of the optimal combination of risky assets and the riskless asset. Hence,
for expositional purposes, assume that the investors just invest in the risky fund and the riskless
asset and that the fund then invests the money in the primary risky securities according to
formula (X.18). Let there be K investors and consider investor #k, k = 1,2,...,K. Let

k th
w* = fraction of the k investor's wealth invested in the risky fund in his
optimal portfolio;
k th
Wo = amount of initial wealth of the k investor;

d k ≡ w*kW ok = number of dollars invested by the k investor in the fund =

th

demand for the fund.

Define: M= equilibrium market value of all risky assets ("the market")

n

= ∑N
i=1
i Pi where N i = number of shares of security i outstanding

225
 Robert C. Merton

and Pi = equilibrium price per share of firm i

Define: VR = equilibrium market value of the aggregate supply of riskless asset (which
may be zero).

In equilibrium, aggregate wealth W must satisfy

K n
(XI.1) W ≡ ∑ W ok = M +V = ∑N
R i Pi + V R
k =1 i=1

In equilibrium, aggregate demand = aggregate supply, i.e.,

K K K
(XI.2) ∑ d k = ∑ w*k W o = M;
k =1 k =1
k
∑ (1 - w
k =1
k
* )W ok = V R .

So, M is the total number of dollars invested in the fund. How much is (implicitly) invested in
risky primary asset i ? From (X.18) the total dollars of investment demanded in security i is

Di ≡ wi M, i = 1,2,..., n.
*
(XI.3)

But, in equilibrium, the supply of asset i must equal the demand, i.e.,

(XI.4) Di = N i Pi , i = 1,2,..., n .

From (XI.3) and (XI.4),

N i Pi N
(XI.5)
*
wi = = n i Pi , i = 1,2,..., n.
M
∑ N i Pi
i=1

226
Finance Theory

Thus, in equilibrium, the prices must be such that the fraction of the optimal-combination-of-
th
risky-assets portfolio allocated to security i must equal the ratio of the market value of the i
security to the market value of all risky assets. A portfolio which holds assets in proportion to
their market value is called a market portfolio. (XI.5) states that in equilibrium, the optimal
combination of risky assets must be a market portfolio. Since each investor's optimal portfolio is
a combination of the optimal combination of risky assets and the riskless asset, we have that in
equilibrium, each investor holds a combination of the market portfolio and the riskless asset.
Further, since we have that every efficient portfolio (except just holding the riskless asset alone)
is perfectly positively correlated, all investors' portfolios are perfectly correlated. Further, since
the relative holdings of risky assets by each investor are the same as in the market portfolio and
since prices cannot be negative, we have that in equilibrium, no investor will optimally short-sell
any risky asset. Can we say more? Let Z M be the random variable return per dollar invested in the
market portfolio; then E( Z M ) ≡ Z M is the expected return on the market and

E{( Z M - Z M )2 } ≡ σ 2M = the variance of the return on the market. In equilibrium,

n n n

Z M = Z = ∑ wi Z i ; σ M = σ * = ∑ ∑ wi w j σ ij where wi is as defined in (X.18).

* * 2 2 * * *

i=1 i=1 j=1

The following derivation is designed to avoid using any mathematics beyond the
elementary calculus, and therefore, is somewhat tedious. A direct analytical proof can be found
in Merton, "Analytical Derivation... Portfolio Frontier", p. 1868-1871.

Question: In equilibrium, can we deduce the relationship among expected rates of return on
securities and develop a systematic, quantitative measure of the "risk" of a security?

227
 Robert C. Merton

Figure XI.1

Consider a Portfolio of Three Securities

Let w1 = % invested in security i (any security not on efficient frontier)

w2 = % invested in the market portfolio (optimal combination of risky assets)
1 - w1 - w2 = % invested in the riskless asset

(security i has expected return Z i and standard deviation σ i i.e., point A)

The return on the portfolio, Z , is

(XI.6) Z = w1 Z i + w2 Z M + (1 - w1 - w2 )R = w1 ( Z i - R) + w2 ( Z M - R) + R

The expected return is

(XI.7) E(Z) = Z = w1 ( Z i - R) + w2 ( Z M - R) + R

228
Finance Theory

and the variance is

(XI.8) Var(Z) = σ 2 = E{(Z - Z )2 } = w12 σ i2 + w22 σ 2M + 2 w1 w2 σ iM

where σ iM = the covariance between the return on the i security and the market portfolio.
th

Clearly, from the definition of efficient portfolio, the only times that this three-asset portfolio is
efficient is when w1 = 0. I.e., no investor would hold this portfolio as an optimal portfolio unless
w1 = 0. Now, only consider mixes of the three securities which lead to an expected return on the
portfolio = m(= Z i ). (In Figure XI.1, this is represented by the dotted line through AC .) How do we
find the minimum-variance portfolio constructed from these three securities with expected return
m ? Set m = E(Z) = Z = w1 (m - R) + w2 ( Z M - R) + R) (using Z i = m). Then

m- R -R (m - R)
(XI.9) w2 = - w1 Z i = (1 - w1 ) .
Z -R ZM - R ( ZM - R

Substitute for w2 from (X1.9) into the expression for the variance (XI.8) to get

⎡ (m - R )2 2(m - R)( Z i - R) w1 ( Z i - R) ⎤ 2
2

σ w1 σ i ⎢
=22 2
+ 2
- w1 + ⎥σ M
⎢⎣ ( Z M - R ) ( Z M - R )2 ( Z M - R )2 ⎥⎦
(XI.10)
⎡ m- R ( - R) ⎤
+ 2 w1 ⎢ - w1 Z i ⎥ σ iM
⎣ZM - R ( Z M - R) ⎦

or, by rearranging terms,

⎡ 2 ( Z i - R )2 2 2( Z i - R) ⎤ 2
σ = ⎢σ i +
2
σ M - σ iM ⎥ w1
⎣⎢ ( Z M - R )2 ( Z M - R) ⎦⎥
(XI.10')
⎡ ( - R) 2 ⎤ (m - R) σ 2M (m - R )
2
+ 2 ⎢σ iM - Z i σM ⎥• w1 +
⎣ ( Z M - R) ⎦ ( Z M - R) ( Z M - R )2

Now, to find the minimum variance portfolio, we differentiate σ

2
in (XI.10') with respect to w1

and the minimizing w1 will be where

229
 Robert C. Merton

dσ2
= 0.
d w1

*
Call w1 the w1 which minimizes (XI.10').
Differentiating (XI.10), we have that

dσ2 ⎡ 2 ( Z i - R )2 2 2( Z i - R) ⎤
= 2 w ⎢σ i +
*
1 σM - σ iM ⎥
dw1 ⎢⎣ ( Z M - R )2 ( Z M - R) ⎥⎦
⎡ ( - R) 2 ⎤ ⎛ m - R ⎞
(XI.11) + 2 ⎢ σ iM - Z i σ M ⎥ ⎜⎜ ⎟⎟
⎣ ( ZM - R) ⎦ ⎝ ZM - R ⎠
= 0 at w1 = w*1 .

But, we know that the variance-minimizing portfolio will be on the efficient frontier (point C in
Figure XI.1) where w1 = 0. Therefore,

*
(XI.12) w1 = 0 .

But, from this condition (XI.12) and (XI.11), we have that either (a) m = Z i = R; or (b)

⎡ ( Z i - R) 2 ⎤
⎢σ iM - σ M ⎥ ≡ 0.
⎣ ( Z M - R) ⎦

Since we chose security i arbitrarily, unless the expected return on all securities = R , it must
be that condition (b) holds. So that in equilibrium,

⎛ σ iM ⎞
(XI.13) Z i - R = ⎜⎜ 2 ⎟⎟ ( Z M - R)
⎝σM ⎠

for any security. (XI.13) is more commonly written as

(XI.14) Z i - R = β i ( Z M - R)

230
Finance Theory

where (" beta" ) βi ≡ σiM

2
.
σM

(XI.13) and (XI.14) is the fundamental equation relating the equilibrium expected returns on any
security with any other. This equation is called the Security Market Line.
Under conditions of homogeneous expectations and equilibrium, the efficient portfolio
frontier is called the Capital Market Line.

Figure XI.2

The equation of that line can be written as

Z = R + reσ ,
(XI.15)

and, in equilibrium, the market portfolio is an efficient portfolio, and hence, must be on the line.
I.e., Z M = R + r e σ M , or

ZM - R ,
(XI.16) re ≡
σM

231
 Robert C. Merton

where r e is called the price of risk-reduction for efficient portfolios. It will be important for
later analysis to remember that: even if there is not homogeneous expectations; even if the
market is not in equilibrium; even if people are not mean-variance maximizers, one can still form
a market portfolio; and by computing its mean and standard deviation, the Capital Market Line in
Figure XI.2 can be formed. This line represents all portfolio combinations of the market
portfolio with the riskless asset (where the market portfolio is never sold short). The conditions
that the market is in equilibrium and that there is agreement, imply that the Capital Market Line
is the locus of efficient portfolios and that there are no feasible portfolios with mean-variance
combinations above that line.
While the Capital Market Line describes the equilibrium expected return relationship
among efficient portfolios, the Security Market Line (XI.13) or (XI.14) describes the equilibrium
expected return relationships among all individual securities or portfolios (efficient or not). We
can rewrite (XI.13) or (XI.14) as

(XI.17) Z i - R = r s σ iM i = 1,2,..., n

⎛ Zi - R ⎞
where r s ≡ ⎜⎜ 2 ⎟⎟ is called the price of risk-reduction for securities.
⎝ σM ⎠

232
Finance Theory

Figure XI.3

Note: σ kM can be negative in which case the equilibrium expected return on that security, Z k , will
be less than R.

The Risk of a Security

If risk is defined as that measure such that as it increases, a risk-averse investor would
have to be compensated by a larger expected return in order that he would continue to hold it in
his optimal portfolio, then, from Figure XI.3 the measure of a security's (relative) risk is its
covariance with the market. An equivalent measure is from (XI.14), the "beta" of the security.

Note: From Figure XI.3, only the risk of efficient portfolios can be measured by its standard
deviation or variance. By definition, σ iM ≡ ρ iM σ iσ M where ρ iM ≡ correlation coefficient
th
between the return on the i security and the market portfolio. So, we can rewrite (XI.17) in
terms of r e [in (XI.16)] as

233
 Robert C. Merton

(XI.18) Z i - R = r e σ i ρ iM .

For a fixed risk premium, Z i - R, and fixed price of risk-reduction, r e , what value for ρ iM allows
σ i to be as small as possible? Clearly, the largest value of ρ iM , namely, ρ iM = 1. Note: when
ρ iM = 1, (XI.18) and (XI.15) are the same. So again, we see that all efficient portfolios (with σ > 0 )
are perfectly correlated with the market portfolio.
Figure XI.4

Note: For ρ iM = 0, Z i = R, independent of σ i .

The intuition behind the Security Market Line can be developed in a variety of ways.
One way is by using a marginal analysis to study the effect of a small change in portfolio
composition.

234
Finance Theory

If the market portfolio combined with the riskless asset is an efficient portfolio, then one
cannot both increase the expected return and reduce the variance by combining this portfolio with
another asset.
Consider an investor who has selected a portfolio with return given by
*
Z * = w* Z M + (1- w* )R . The expected return on this portfolio is
*
Z = w ( Z M - R) + R and its
2
variance is Var( Z ) = [ w ] σ M . Consider the effect of a small change in this portfolio achieved by
* * 2

increasing the fraction held in asset i by δ and decreasing the holding in the riskless asset by δ . The
return on this portfolio can be written as

Z = w* Z M + (1 - w* )R + δ Z i - δR = Z* + δ ( Z i - R) ,

and it follows that

(XI.19a) Z = Z * + δ ( Z i - R)

and

(XI.19b) Var(Z) = Var( Z * ) + 2 w* δ σ iM + δ 2 σ i2 .

The effect of this small change on the mean and variance can be determined by differentiating
(XI.19) with respect to δ and evaluating the derivative at δ = 0. That is,

(XI.20a) [d Z /dδ ] δ =0 = Z i - R

and

(XI.20b) [d Var(Z)/dδ ] δ =0 = 2 w* σ iM .

Case (i): Suppose that σ iM < 0. If Z i - R ≥ 0, then by moving δ from δ = 0 to δ > 0 , one
could reduce the variance of the portfolio and not reduce its expected return. But, this would contradict
*
the efficiency of Z and therefore, the efficiency of the market portfolio. Hence, if the market portfolio
is efficient and σ iM < 0 , then Z i - R < 0.

235
 Robert C. Merton

Case (ii): Suppose that σ iM > 0 . If Z i - R ≤ 0, then by moving δ from δ = 0 to δ < 0 , one
could reduce the variance of the portfolio and not reduce its expected return. Again, such a possibility
would contradict the efficiency of the market. So, if the market portfolio is efficient and
σ iM > 0, then Z i - R > 0.

Case (iii): Suppose that σ iM = 0. If Z i > R, then by moving δ from δ = 0 to δ > 0, one could
increase the expected return on the portfolio and not increase its variance at the margin. If Z i < R, then

by moving δ from δ = 0 to δ < 0, one could increase the expected return and not increase its
variance. Because either possibility would violate the efficiency of the market, it follows that

Z i = R if σ iM = 0.
Perhaps because equations (XI.17) and (XI.18) lack some intuitive appeal, expression
(XI.14) for the Security Market Line has generally been the preferred form in popular use.
"Beta" in (XI.14) is frequently called the (relative) "volatility coefficient of the security." To see
why, we proceed as follows:
Define the random variables ε k ≡ Z k - Z k and ε M ≡ Z M - Z M . Then, by construction
E( ε k ) = E( ε M ) = 0 and Var( ε k ) = σ 2k and Var( E M ) ≡ σ 2M . ε k and ε M are the unanticipated

parts of the returns on asset k and the market portfolio, respectively.

From the definition of ε k and ε M and from (XI.13), we can write the return on asset k as

σ kM (
Zk = R+ Z M - R) + ε k
σ 2M
(XI.21) = R + σ kM ( Z M - ε M - R) + ε k
σ 2M
= ak + β k Z M + uk

236
Finance Theory

where a k ≡ (1- β k )R; β k ≡ σ kM / σ 2M and u k ≡ ε k - β k ε M . By construction, the random variable

u k has the property that E( u k ) = 0 and Var( u k ) = σ 2k - β k σ M = (1- ρ kM )σ 2k . Further we have

2 2 2

that

Cov( Z M ,u k ) = Cov( ε M ,ε k ) - β k Cov( ε M ,ε M )

(XI.22) = σ kM - β k σ 2M
=0

since β k = σ kM / σ M . That is, u k is uncorrelated with the return on the market. For the reader
2

familiar with basic single variable regression theory, βk is equal to the (theoretical) regression

coefficient from regressing Z k on Z M and a constant. If (XI.21) were viewed formally as a regression

equation, u k would be called the residual and represent that part of the return on asset k , Z k , which

is not "explained" by the return on the market, Z M . As is well-known, the residual in a least-squares
regression is always uncorrelated with the independent variable: Hence, (XI.22) simply reaffirms that
result.
Viewing (XI.21) as a regression equation is probably the reason that β i is thought of as a
relative volatility measure. That is, (neglecting the "residual," u k ), if the market goes up 10%, and if
β k = 2, then the return on security k would be 20% (plus a k ); and if the market goes down 10%,
then the return on security k would be –20% (plus a k ). Thus, securities (or portfolios) with large

"betas" ( β k ) are called "volatile" or "aggressive" securities and in a similar fashion, securities

with small "betas" are called "defensive" securities. From (XI.14), we can draw the Security
Market Line in terms of beta:

237
 Robert C. Merton

Figure XI.5

Note: All securities with β k = 1 have expected returns = Z M ; securities with β k = 0 have
expected returns = R.
The reader is warned that, in general, the regression interpretation of βk is only a

heuristic. (XI.21) was derived simply by construction with no assumptions about the joint distribution of
Z k and Z M . Although, by construction, uk is always uncorrelated with ZM , the stronger

condition that E( u k | Z M ) = 0 is required for (XI.21) to be a valid regression equation. Moreover, even

if this condition were satisfied, one cannot attribute strict causality between Z M and Z k . Nonetheless,
this interpretation does provide some intuition for what beta is.

Systematic (or "Market") Risk and Unsystematic (or "Pure" or "Unnecessary") Risk

Equation (XI.21) holds for all securities or portfolios in equilibrium. Further, if the
portfolio k is efficient, then u k ≡ 0 . Hence, if an investor holds an efficient portfolio, then he will
not be subjected to the (additional) uncertainty of return caused by u k . Since all investors can satisfy
their portfolio demands (in equilibrium) by holding efficient portfolios, any investor who holds a

(inefficient) portfolio where u k ≡ 0 is exposing himself to an (unnecessary) additional risk. Hence,

238
Finance Theory

u k is called the unsystematic or unnecessary risk of security or portfolio k. Note that even if the
investor holds an efficient portfolio ( u k ≡ 0) , the return on the portfolio (for beta ≠ 0 ) is uncertain
because the return on the market is uncertain. This is an irreducible or "necessary" risk that he must take
to get the expected return Z k . Thus β k Z M , or really β k ( Z M - Z M ) = β k ε M , is called the
systematic risk of security k and because it is proportional to the market return, it is often called
the market risk of security k.
The uncertain part of a security's return, ε k , can always be written as the sum of systematic
and unsystematic risk: namely, as

(XI.23) ε k = β k ε M + uk .
From (XI.22) and (XI.23), the total variance of the return can be written as

(XI.24) σ 2k = β 2k σ 2M + Var (u k )
Variance of Variance of
Systematic Part Unsystematic Part

Note: From (XI.14), the equilibrium, (expected) reward or risk-presmium or excess return,
Z k - R, is proportional to β k σ M , the standard deviation of the systematic part of the total risk, and
not σ k , the standard deviation of total risk.
Hence, investors only get extra (expected) return from bearing larger systematic risk, or
alternatively, the market does not reward investors for choosing inefficient portfolios and
exposing themselves to more risk than is necessary.

Implications of the Capital Asset Pricing Model for Portfolio Selection

The analysis of the previous section provides a very simple portfolio selection strategy
(independently of whether the CAPM holds in the "real" world): Namely, (1) diversify your
holdings as much as possible (i.e., hold each security available in proportion to its value relative

239
 Robert C. Merton

to the market); (2) Borrow (or lend) to lever (or "cool down") this portfolio until one achieves the
"right" expected return-standard deviation tradeoff. This selection rule is called a naive or
passive rule because it requires little analysis (only an estimate of the market expected return and
variance) and nothing about individual securities.
Since it is always a feasible portfolio policy, one can use this rule to provide a benchmark
for comparison of overall portfolio performance of active portfolio selection rules as will be
shown in Section XIII. Clearly, such active portfolio management should, as a minimum,
provide at least as good performance (after deducting costs) as the passive policy.
Except for certain bookkeeping and purchasing economy of scale, the naive strategy
eliminates the need for a portfolio manager all together.

240
XII. RISK-SPREADING VIA FINANCIAL INTERMEDIATION: LIFE INSURANCE

As discussed briefly at the end of Section V, financial assets can be traded directly in the
capital markets or indirectly through financial intermediaries. In general, "standardized"
securities are traded in markets (e.g., government bonds, wheat futures, shares of IBM) while
"custom" contracts (e.g., individual mortgage, personal loan, or insurance) are handled through
financial intermediaries. In this section, the classical case of pure life insurance is examined to
show how efficient risksharing can be achieved using a combination of financial intermediation
and the capital market.

The Life Insurance Company

Suppose that there are N people in the economy each with wealth (per capita) W.
Hence, national wealth ≡ W = NW. Suppose further that each person purchases a one-year term
life insurance policy which pays $c in the event of death and we define q ≡ c/W to be the

amount of insurance coverage purchased by each person as a fraction of his wealth. Let yi be a
th
random variable describing the death of the i person where yi = 0 if person i survives the year

and yi = 1 if person i dies during the year. Assume that the mortality tables are such that

E( y i ) = ρ , the same for all people, i = 1,2,..., N and Var( y i ) = V 2 which is also the same for all
N
people. Hence, ρ is the expected number of deaths per person (0 < ρ < 1) . Let Y N = ∑y
i=1
i be the

random variable for deaths of all people and it is equal to the number of deaths in the economy. If the
death of one person is independent of another (a crucial but reasonable assumption), then

E[ Y N ] = Nρ ; Var( Y N ) = NV 2

241
 Robert C. Merton

If a single competitive insurance company writes all the policies, then the analysis will determine
the:

• Premium per policy charged, P N

• The amount of equity capital required by the company to do business, K N .
• The required (expected) return on the equity by investors in the insurance
company.

Premiums are received at the beginning of the year in the amount, NP N . Benefits are paid at the
end of the year in the (random variable) amount C N ≡ cY N . Hence,

E[ C N ] = C N = Nc ρ
Var( C N ) = σ 2N = Nc 2 V 2

Suppose that investors are mean-variance maximizers and that the conditions for the Capital
Asset Pricing Model (Section XI) hold. If R = 1 + rate of interest, then the return per dollar
R( NP N + K N ) - C N
invested in equity of the insurance company is Z N ≡ , and
KN

R[( NP N ) + K N ] - Ncρ
E[ Z N ] = Z N =
KN

Var [ Z N ] = σ 2Z N =
σ = Nc 2 V 2
2
N
2 2
K N KN
In equilibrium, supply must equal demand, and so, for the equity of the insurance company to be
held, we have that Z N must satisfy the basic equilibrium condition for the CAPM:

KN
supply = K N = Demand for asset; or = w*Z N = fraction of the market portfolio
NW

*
where W Z N is also the fraction in optimal combination of risky assets given by:

242
Finance Theory

n
∑v Z j( Z j - R)
N
j=1
w*Z N =
(A-RC)

Suppose (as is reasonable) that Z N is independent of the returns on all other assets

(i.e., Cov( Z N , Z j ) = 0 for Z N ≠ Z j ), then

n
KN Nc 2V 2 Nc 2V 2
Cov( Z N , Z M ) = ∑ w*j Cov( Z N ,Z j ) = w*Z N Var ( Z N )= ( )( )= .
j=1 NW K 2N NW K N

From Section XI, we have from the Security Market Line that

⎛ c2 V 2 ⎞
ZN - R = r S Cov( Z N , Z M ) = r S ⎜⎜ ⎟⎟
⎝ WK N ⎠
Substituting for Z N , we have that

R[ NP N + K N ] - Nc ρ ⎛ c 2V 2 ⎞
- R= rS ⎜ ⎟ or
KN ⎝ WK N ⎠
⎛ c 2V 2 ⎞
RNP N - Nc ρ = r S ⎜ ⎟ or substituting for c = qW
⎝ W ⎠
⎛ q 2WV 2 ⎞
RP N - q ρW = r S ⎜⎜ ⎟⎟
⎝ N ⎠

The (equilibrium) premium per policy can be written as

qWρ 1 2 2 cρ 1 2
PN = + r S q WV = r S cq V and therefore ,
R NR R NR
2
r S cq V
ZN = R+ , the (equilibrium) expected return on equity in the insurance company.
KN

243
 Robert C. Merton

Note: In this formal analysis, we have not taken into account the limited liability feature of the
equity of the insurance company which leads us to the last question to be answered:
What is the appropriate value for K N ?

To answer this question, one must go back and ask what service is the financial intermediary to
provide to the customer? What does he want? The customer wants a certain payment, c, in the
event of death. Now, if the total number of deaths is such that the (ex-post) benefits required to
be paid, c N , is larger than the company's total assets, R( NP N + K N ), then by limited liability on
R
equity, the customer will not receive the full benefits promised, but only ( NP N + K N ) < c.
yN
Obviously, the larger is K N the less likely is default. If K N is "too small", then the probability of
default is higher, and the customer in purchasing the policy does not get the simple security he wanted
which pays $c for sure in the event of his death, but rather has the more complicated security
which pays $c in the event of his death, conditional on the company being solvent, pays the
R
(r.v.) amount ( P N N + K N ) in the event of death, conditional on the company not being solvent.
yN

Clearly, to assess the probabilities about possible payoffs, the customer would have to know the amount
of capital the firm has; the nature and quantity of policies written for other customers; the probability of
these customers dying; etc. In short, nearly everything about the company that the management knows,
the customer would have to know and analyze. Essentially, the customer takes a (partial) equity position
in the company. Since one purpose of the financial intermediary is to limit the amount of information
required by customers to make a decision and because the service wanted is basically life insurance, the
equity capital should be large enough to (virtually) eliminate the chance of default. In doing so, the
separation between customer and equityholder (or general liability or debtholder) is made as large as
possible. Define: reserves as the amount of assets required to be held by the insurance company
to ensure that payment will be made to customers with some probability. Let
R N = reserves requires = R (NPN + K N ) So, given the premium, there is a one-to-one

correspondence between reserves and capital (equity). Clearly, the amount of reserves required to ensure
with absolute certainty that all customers will be paid in every state of the world would come by

244
Finance Theory

requiring that assets be large enough to payoff everyone in the event that everyone dies. I.e.,
max
C N = Nc or R max
N = maximum required reserves = Nc = R( NP N + K max
N ) or

1 cq V 2
K
max
N = Nc( - ρ ) - rS . However, for Nc large and ρ reasonably small, the amount of
R R
capital required to meet the maximum reserves could be prohibitively large. Further, if ρ is small,
there is a very small chance that everyone will die (especially since the events of death are independent)
and one would expect that for large N, there would be some diversifying effects. Hence, it may not
be necessary for the company to hold the maximum reserves while still performing the essential
service required.
Suppose that instead it was required that reserves be such that the probability of default is
less than some assigned level, p , i.e., Prob{C N > R N } ≤ p , and define the associated required
* *

capital as K N ( p ). Note: K N (0) = K N .

* max
Prob{C N > R N } =

n R ( *)
KN p rs cqV
Prob{(∑ yi)c > R K N(p ) + Ncρ + rscqV } = Prob{X N > ( )
* 2
+
1 V c N N
N
∑ yi-Nρ
where X N ≡
1
and E(X ) = 0; Var(X ) = 1 .
N N For large N, X N will be
V N
distributed approximately standard normal (Gaussian). Hence,
* *
R K N ( p ) r S qV R K ( p ) r S qV
Prob{ X N > ( ) + } ≈ 1 - Φ [( ) N + ] = p* where Φ[ ] is the
Vc N N Vc N N
cumulative density function for the normal distribution. For this distribution, there is a one-to-one
*
correspondence between p and the number of standard deviations to the right of the mean. I.e., let
µ = number of standard deviations, then

245
 Robert C. Merton
* *
µ(p ) p

0 .5000
1.0 .1600
2.0 .0230
2.33 .0100
3.10 .0010
3.70 .0001 (1 chance in 10,000)
4.00 .00004 (1 chance in 25,000)

*
R K N ( p ) r S qV Vc
Hence, µ ( p ) = ( ) )µ ( p* )N - r SqV for large N. Thus,
* *
+ or K N ( p ) = (
Vc N N R

for a given p (or µ ( p )) , we have an expression for the required equity capital for large N.
* *

1 *
Asymptotic Results for large N(N → ∞) ( > p > 0)
2

Vc
*
KN( p ) ≈ ( )µ( p* ) N
R

cρ R
2

limit ( P N ) =
N →∞ R
; lim ( Z N ) = R;
N →∞
limit ( σ 2Z N ) =
N →∞ µ ( p* )
2
> 0.

* *
Required Equity K (p ) KN( p )
limit ( ) = limit ( N ) = 0; lim ( ) = 0;
N →∞ Total Premiums N →∞ N PN N →∞ NW
*
Required Equity KN( p )
limit
N →∞ Maximum Equity limit
=
N →∞
( max
KN
)= 0

1
Suppose Suppose ρ = .0025; c = $30,000 p = .00004 =
*

25,000

2
V = .0025 R = 1 [ µ ( p* ) = 4]

246
Finance Theory

K N (.00004)
N NPN K max
N = K N (0) K N (.00004)
NPN
10,000
4 $750,000 $299,250,000 $600,000 .8000
(1x10 )
90,000 $2,693,250,000
4 $6,750,000 $1,800,000 .2667
(9x10 ) ($2.7 billion)
1,000,000
6 $30,000,000,000
(1x10 ) $75,000,000 $6,000,000 .0800
($30 billion)
one million
9,000,000
6 $675,000,000 $270 billion $18,000,000 .0267
(9x10 )
8
1x10 $60,000,000
$7,500,000,000 $3 trillion
one hundred 12 (60 million) .0080
($7.5 billion) (3x10 ) 7
million (6x10 )

Thus, we observe a characteristic property of (many) financial intermediaries: namely, that net
worth is a small fraction of total assets (in the example, less than 1%); further total (potential)
liabilities are many orders of magnitude larger than total assets or reserves. Of course, sales and
other operating expenses would have to be added to the premium and other assets (buildings,
etc.) have been excluded.
The benefits of the financial intermediary in this case are obvious: if each insurance
policy were written by one person for one other person (and if r S ≈ 2 and q ≈ 1), then the
minimum premium for a $30,000 policy would be $225 versus $75 charged by the company.
max
Further, the reserves required would be $30,000 (or K 1 ) per policy versus $.60 per policy for
the intermediary! (50,000 times as much!)
Note despite the tremendous diversifying power of many policies, if there were no equity
capital market to raise the funds, it is doubtful if such an organization could occur without
substantially higher premiums. Suppose one (wealthy) individual provided all the capital ($60
million): (at the derived rates with R = 1) the expected rate of return is zero and there is a .16
probability of one standard deviation to the right which translates into a $15 million loss! Few

247
 Robert C. Merton

risk-averse utility maximizers would accept such an investment. But, by diversifying the risk by
issuing equity in the capital market and if individuals hold well-diversified equity portfolios (as
they should), then the loss would be around 15¢ per investor which is trivial for an investor with
initial wealth of $30,000. Thus, through the combined use of the capital market and the financial
intermediary, the individual investor can get the service or asset he wants (virtually no-default
life insurance) to eliminate a substantial non-systematic risk, at minimum cost.

248
XIII. OPTIMAL USE OF SECURITY ANALYSIS AND INVESTMENT
MANAGEMENT

In Section XI, we used portfolio analysis to derive the Capital Asset Pricing Model which
provides a relationship between expected return and risk in equilibrium. In an environment
where there is no significant differential information among investors (i.e., if distributional
beliefs about security returns are homogeneous), it was further shown that all efficient portfolios
can be represented as a simple combination of the market portfolio and the riskless asset. This
analysis suggested a naive or passive portfolio strategy (namely, hold the market mixed with the
riskless asset) which does not require the investor to undertake security analysis of individual
firms. In an environment where some investors may have differential information, this strategy
is still appropriate for those investors who do not have such information available to them. That
is, it is appropriate for those investors with information sets that do not reveal mispriced
securities. This strategy provides the best “protection” to such investors from those investors
who do have significant differential information (the “information traders”).
This passive strategy does require some forecasting of the “macro” type: namely, an
estimate of ZM and σM . However, this information is only required so that the investor can pick the
right efficient portfolio for his specific preference function. That is, no matter what combination of the
market and riskless asset he selects, the investor will have chosen an efficient portfolio. If his estimates
of ZM and σM are in error, then he will select the wrong efficient portfolio for his specific tastes.

Nonetheless, he will receive the highest expected return available (based upon his information set) for
whatever level of risk he in fact did bear. One simple method for estimating ZM and σM . (R is, of

course, observable) is to use historical data to estimate ZM - R and σ M or ( ZM - R)/ σM .

The passive strategy presumes that market prices for securities reflect the information that
the investor has, and therefore, relative to the investor's information set, security prices will be
such that expected returns will satisfy the Security Market Line. Hence, the strategy's success
depends upon at least some entities undertaking individual security analysis and acting on this
information to ensure that market prices reflect information available to the investor. Who these
“informed” investors are as well as how successful they are, in an empirical question which will
be addressed in Section XVII. In this section, we develop the procedures for optimal use of such
249
 Robert C. Merton

differential information if one were to have it. While the emphasis of the analysis is an optimal
utilization of security analysts who do only “micro” or individual security forecasts, the final part
of the section combines both micro and macro forecasts.
We begin by using the Capital Asset Pricing Model to develop an operationally useful
definition of “under” and “over-priced” securities. The reader should note that all distributional
estimates are computed from a particular entity's information set.

Figure XIII.1

An undervalued security has an expected return greater than that predicted by the SML (e.g.,
security i in Figure XIII.1). An overvalued security has an expected return less than that
predicted by the SML. Write the expected return on security k as

250
Finance Theory

(XIII.1) Zk = R + βk ( ZM - R) + αk

if α k = 0, then security k is "fairly priced"

α k > 0, then security k is "under priced"

α k < 0, then security k is "over priced"

A portfolio with a consistent positive “alpha” (α) shows evidence of ability to forecast a
security (or securities) better than the “market.”

Figure XIII.2

ZP - R > ZM - R .
A superior-performing portfolio (“super efficient”) has An inferior-performing
σR σM

ZP - R < ZM - R .
portfolio (“sub efficient”) has
σP σM

On the Relationship Between Superior Stock Selection & Super-Efficiency

251
 Robert C. Merton

Consider a portfolio constructed by a manager with superior stock selection skills that has a
ZP - R = βP ( ZM - R) + α P , α P > 0
positive alpha. I.e., 2
σP = βP σM + Var[U P]
2 2

Question 1: Is it possible to have a portfolio with a positive alpha that is a subefficient portfolio?
Yes.

Figure XIII.3 Figure XIII.4

Question 2: Is it always possible to construct a super-efficient portfolio if one has available a

portfolio with a positive alpha? Yes.
An analytical demonstration is as follows:
Form a portfolio of three securities:

252
Finance Theory

Let w1 = fraction invested in portfolio P

w 2 = fraction invested in market portfolio
w 3 = fraction invested in riskless security = 1 - w1 - w 2
Z ≡ w1 ( Z p-R) + w 2( ZM -R) + R
Z = w1(ZP -R) + w 2( ZM -R) + R = w1α P + [ w1β P + w 2][ Z M - R] + R
Var(Z) = w12σ 2P + w 22σ 2M +2w1w 2σ PM = w12σ 2P + w 22σ 2M + 2w1w 2β P σ 2M
= w12Var[ U P ] + [w12β 2P + 2w1w 2β P + w 22]σ 2M
= w12Var[U P ] + [ w1β P + w 2]2σ 2M

Find the minimum-variance portfolio with expected return equal to ZM .

Z = ZM = w1 αP + [ w1βP + w 2][ZM - R] + R
w 2 = 1 - w1[βP ( ZM - R) + αP ]/(ZM - R)
σ2M
Var(Z) = w12 Var[ U P ] = [ ZM - R - w1 α P ]2
( ZM - R ) 2

dVar(Z) ⎡ σ2M αP ⎤
= 2 ⎢ w1* Var[UP ] - [ ZM - R - w1* αP ] 2
⎥ = 0 at w1 = w1*
dw1 ⎣⎢ ( ZM - R ) ⎦⎥
αP
w1* = Note : αP > 0 implies that w1* > 0
U α2P
( ZM - R)[Var[ 2P ] +
σM ( ZM - R)

( Z-R) ( - R)
Exercise: Show that > ZM
Var(Z) σM

when Z = w1*( Z P -R) + w *2( Z M -R) + R= ZM

σ 2M
and Var(Z) = ( w1*) 2Var[ U P ] + [ ZM -R-w1*α p ]2
( ZM -R) 2

Hint: just show that Var(Z) < σ2M

253
 Robert C. Merton

If a single entity has micro forecasts for the means, variances, and covariances of all
available securities, then the optimal utilization of such information is to form the risky portfolio
using formula (X.18) (the “optimal combination of risky assets”). This portfolio can then be
mixed with the riskless asset to produce an efficient frontier as described in Section X. However,
operationally, this may not be feasible because:

(i) the large number of available securities make it unlikely that any single unit would be
able to make estimates for all available securities;
(ii) a particular unit may have superior forecasting capability only with respect to a subset
of available securities;
(iii) a control mechanism should be employed to keep the portfolio manager from making
decisions based on forecasts which are inaccurate; and further, to reward analysts for
doing a “good job” at what they were hired to do, it is necessary to develop
performance measures for each of the roles leading to the “best” portfolio. E.g., a
portfolio could have “bad performance” even though the analyst is doing his job
because of poor portfolio management or vice versa.

254
Finance Theory

Figure XIII.5 ACHIEVING SUPER EFFICIENCY FROM SUPERIOR STOCK SELECTION

255
 Robert C. Merton

Suppose that we are doing security analysis on the shares of m companies whose returns are
represented by Zk , k = 1,2,..., m. We can always write the returns as is done in Section XI, as

(XIII.2) Zk = (1 - βk )R + βk ZM + U k , k = 1,2,..., m

σkM
where ZM is the return on the market; βk ≡ 2 ; U k is a random variable such that
σM
Cov( U k , ZM ) = 0. The capital asset pricing model predicts that if security k is “fairly” priced,

E( U k ) = 0. However, the purpose of security analysis is to find securities that are either under-

or over-valued. I.e., where E( U k ) ≠ 0. Define: U k ≡ α k + εk where ε k is a random variable such

that E(εk ) = 0;
Cov(ε k , ZM ) = 0, k = 1,...,m; Cov(ε k,ε j) ≡ σ’kj,k = 1,...,m; j=1,...,m. Rewrite (XIII.2) as

(XIII.3) Zk = (1 - βk )R + βk ZM + α k + ε k , k = 1,..., m

Suppose that we do micro forecasts on the m stocks, but no forecasts on the market. Since from
(XIII.3), one can think of the return on Zk as coming from two sources. (1) movements in the
market, and (2) movements individual to the stock which are independent of the market, this type
of forecasting implies estimates of αk and εk by the analysts, without knowledge of ZM :

I. “Active” Portfolio
Consider the following portfolio constructed from (m+2) securities: the m stocks being
analyzed, the market portfolio, and the riskless asset. Let w ia ≡ fraction of the portfolio invested
stock, i = 1,2,...,m; w aM ≡ fraction of the portfolio invested in the market portfolio;
th
in the i
a
w R = fraction of the portfolio invested in the riskless asset. Further, restrict the portfolio weights

to satisfy

256
Finance Theory

m
(XIII.4a) ∑ w ia + w aM + w aR = 1
i =1

m
(XIII.4b) ∑ w ia βi + w aM = 0
i =1

If Za ≡ return per dollar invested in the active portfolio, then

m
Za = ∑ w i ( Zi - R) + w M ( ZM - R) + R from (XIII.4a)
a a

i =1
m
(XIII.5) = ∑ w ia [βi ( ZM - R) + αi + εi] + w aM ( ZM - R) + R from (XIII.3)
i =1
m m
= R + ∑ w ia αi + ∑ w ia εi from (XIII.4b)
i =1 i =1

Cov( Za , ZM )
Note: since Cov(εi , ZM ) = 0 , from (XIII.5), Cov ( Za , ZM ) = 0 I.e., β a ≡ = 0.
σ 2M
All active portfolios satisfying (XIII.4a) and (XIII.4b) are uncorrelated with the market. By
constructing the active portfolio in this way, the returns on the portfolio, Za , depend only on
the (αi , εi) which we have forecasts on and not on Z M (about which we are assumed to have no
forecasts). If we write Z a in the form of (XIII.3), then

(XIII.6) Za = (1 - βa )R + βa ZM + αa + εa

m m
where: β a = 0; α a ≡ ∑ w ia α i; εa ≡ ∑w ia εi
i=1 i=1

Note that:
m
E( Za ) = α a + R = ∑w ia αi + R = Za;
i=1
(XIII.7) m m
σ a2 ≡ Var( Z m) = ∑ ∑ w ia w aj σi′j
i=1 j=1

257
 Robert C. Merton

Consider the efficient portfolio set constructed from all such active portfolios. (I.e., the set of
portfolios with maximum expected return for a given variance). Mathematically, fix the variance
at σa2 , then

λ 2 m m a a
Max { Za + [ σ a - ∑ ∑ w i w j σ i ′j ]}
{ w 1a ,..., w am } 2 i = 1 j=1
m
λ 2 m m
= Max {R + ∑ w ia α i +
2
[σ a - ∑ ∑ w ia w aj σ i′j]}
{ w ia } i =1 i =1 j=1

where λ is the Lagrange Multiplier. The first-order conditions are

(XIII.8)
∂ m
: 0 = α i - λ ∑ w aj σ i′j i = 1,2,..., m
∂ w ai j =1

∂ m m
: 0 = σ a2 - ∑ ∑ w ia w aj σ i′j
∂λ i =1 j =1

Multiply the first equation by wia and sum i = 1,...,m to get

m m m
0 = ∑ w ai αi - λ ∑ ∑ w ia w aj σi′j
i =1 i =1 j=1

Za -R
(XIII.9) =λ
σ a2

The “efficient” part will have Za ≥ R or from (XIII.9), λ ≥ 0. So, from (XIII.8) and (XIII.9), we
have that

⎛ -R ⎞ m
(XIII.8′) 0 = αi - ⎜⎜ Za 2 ⎟⎟ ∑ w aj σi′j , i = 1,2,..., m
⎝ σa ⎠ j=1

258
Finance Theory

th th
If vi′j = i , j element of the inverse of the variance-covariance matrix [σi′j] , then [in an

analogous fashion to (X.18)],

⎛ σa2 ⎞ m
w i = ⎜⎜ ⎟ ∑ vi′j α j , i = 1,2,..., m
a
(XIII.10) ⎟
⎝ Za - R ⎠ j=1

In the special case where the unsystematic parts of the returns on the securities are uncorrelated
with each other (i.e., Cov(εi , ε j) = 0, i ≠ j), then σi′j = 0, i ≠ j and σi′i = (σi′ ) and (XIII.8') becomes
2

⎛Z −R⎞
0 = α i − ⎜ a 2 ⎟ wiaσ i'2 , i = 1,..., m or
⎜ σ ⎟
⎝ a ⎠

σ a2 αi
(XIII.11) w ia = ( ) i = 1,2,...,m
Za -R (σ i )
'2

From either (XIII.10) or (XIII.11), note that the ratio of “risky assets”
2
m m
αi σ k
w i / w k = (∑ Vi j′ α j)/(∑ V k j′ α j) or
a a
2
j=1 j=1 α k σi

is independent of the point chosen on the frontier. Risky assets is put in “ “ because it refers
only to risky assets 1,2,...,m. From (XIII.4b), this portfolio also contains the market risky asset
(unless βi = 0, i = 1,..., m). However, it is also true that

a m a
wM w
a
= - ∑ ai βi
wk i =1 w k

is the same for all points on the frontier. Thus all “efficient” portfolios constructed from the
active portfolio can be thought of as a combination of a risky-asset only portfolio and the riskless
asset, and therefore they are perfectly correlated. To find the particular “efficient” active

259
 Robert C. Merton

a
portfolio with risky assets only, we set wR = 0 in (XIII.4a) and require therefore that
m
∑ w ia + w aM = 1.
i =1

m
This can be done by requiring that the wia satisfy ∑ w ai (1 - βi) = 1 . (This is possible
i =1

provided that β i ≡ 1 for all i; a technical point is that even if βi ≠ 1 , such an all-risky asset
portfolio may not be efficient although it will be on the frontier). In any event, the important
point is that from (XIII.10) or (XIII.11) the holdings depend only on the forecasted variables
(αi , σij) and not on ZM .

II. Mixing the “Active” Portfolio with the “Passive” (Market) Portfolio to Produce an
Optimal Combination of Risky Assets

Consider an investor presented with the active portfolio ( ZM , σ2M ) and the riskless asset
R. The efficient portfolio set can be generated by maximizing the mean for a given variance. I.e.,
let Zp be the return on a portfolio constructed from Za , ZM , and R; and let δ1 be the fraction of
that portfolio invested in the market; δ2 be the fraction invested in the active portfolio; 1 - δ1 - δ2
be the fraction invested in the riskless asset. Then
2 2 2 2 2
Zp = δ1 ( ZM - R) + δ2 ( Za - R) + R; σp = δ1 σM + δ2 σa because: cov( Za , ZM ) = 0. The problem

becomes
γ
Max {R + δ1 ( Z M - R) + δ 2 ( Z a - R) + [σ 2p - δ12 σ 2M - δ 22 σ a2]}
{δ1, δ2 , w1a ,..., w am} 2

where γ is a Lagrange multiplier. Note: he is not only allowed to pick δ1 and δ 2 , but in

addition, w1a ,..., w am . I.e., he can select which active portfolio he wants. So the investor's choice

260
Finance Theory

is not limited, and he is using the analysts' forecasts of α i , σ’ij as well as an estimate of
2
Z M and σ M .
The first-order conditions are

∂
(XIII.12a) : 0 = Z M - R - γ δ1σ 2M
∂ δ1
∂
(XIII.12b) : 0 = Z a - R - γ δ 2 σ a2
∂ δ2
∂
(XIII.12c) : 0 = σ 2p - δ12 σ 2M - δ 22 σ a2 .
∂γ

and

∂ m
(XIII.13) : 0 = δ 2 { α i - γ δ 2 ∑ w aj σ i ′j } i = 1,..., m
∂ w ia j =1

Since

∂ Za ∂[σ a2] m
= α i and = 2 ∑ w aj σ i ′j .
∂ w ia ∂ w ia j =1

Za -R
From (XIII.12b), γ δ 2 = 2 and substituting into (XIII.13), we have that
σa

⎛Z -R⎞ m a
0 = α i - ⎜⎜ a ⎟
2 ⎟ ∑ w j σ i ′j
(XIII.14) , i = 1,..., m provided δ 2 ≠ 0
⎝ σ a ⎠ j =1

Comparing (XIII.14) with (XIII.8'), they are identical. Thus, the correct combination of the
securities in the active portfolio can be determined by choosing an “efficient” portfolio for the
active portfolio without knowledge of the characteristics of the market portfolio.
From (XIII.12a) and (XIII.12b) we have that

δ 2 ( Z a - R) ⎛⎜ σ M ⎞⎟
2
(XIII.15) = .
δ1 ( Z M - R) ⎜⎝ σ a2 ⎟⎠

261
 Robert C. Merton

Thus, the optimal combination of risky assets will have the active portfolio in the amount

⎡ ⎤ ⎡ ⎤
⎢ 2 ⎥ ⎢ 2 ⎥
⎢ ( Z a - R))/ σ a ⎥ ⎢ ( Z M - R)/ σ M ⎥
⎢ ( Z a - R) ( Z M - R) ⎥ and the market in the amount ⎢ ( Z M - R) ( Z a - R) ⎥
⎢ 2
+ ⎥ ⎢ + ⎥
⎣⎢ σ a σ 2M ⎦⎥ ⎣⎢ σ M
2
σ a2 ⎦⎥

Compare the “new” optimal combination of risky assets (using forecasting) with the “old”
optimal combination (with no forecasting): Note

⎡ ⎤ ⎡ ⎤
⎢ 2 ⎥ ⎢ 2 ⎥
* ⎢ ( Z a - R)/ σ a ⎥ ⎢ ( Z M - R)/ σ M ⎥
Z = R + ⎢ ( - R) ( ( Z a - R) + (
⎢ ( Z a - R) ( Z M - R) ⎥ Z M
- R)
Za Z M - R) ⎥
⎢ 2
+ ⎥ ⎢ + ⎥
⎣⎢ σ a σ 2M ⎦⎥ ⎣⎢ σ a
2
σ 2M ⎦⎥

2 2
⎡ ⎤ ⎡ ⎤
⎢ 2
( Za -R)/ σ a ⎥ ⎢ 2
( ZM -R)/ σ M ⎥
σ *2 = ⎢ ⎥ σ a2 + ⎢ ⎥ σM
2
⎢ a
( Z -R)
+ ZM2 ⎥
( -R) ⎢ a
( Z -R)
+ Z M2 ⎥
( -R)
⎢⎣ σ a2 σM ⎦ ⎥ ⎢
⎣ σa
2
σ M ⎥⎦

2 2
Z - R = ⎛⎜ Z a - R ⎞⎟ + ⎛⎜ Z M - R ⎞⎟ > Z M - R for { Z a ≠ R
*
⎜ ⎟ ⎜ ⎟
σ* ⎝ σa ⎠ ⎝ σM ⎠ σM σa < ∞

So, the “new” combination is super-efficient.

262
Finance Theory

Summary: Organizational Structure for Portfolio Management [here]

Product

Customers
(good;bad) Final Risky Portfolio
Zp ; Zp , σ 2p

Final Portfolio Manager ⎛ Z , σ2 ⎞ Control

⎜ p p⎟
1. Form Final Portfolio by mixing Active ⎝ ⎠ Is Final Portfolio
good;
and Passive Portfolio Super-Efficient?
bad 2. Monitor Performance of Active Portfolio
3. Monitor Performance of Passive (Market) good; bad
Portfolio good; bad

Product Product

Active Portfolio Passive Portfolio

Za ; Za , σa2 Z M ; ZM , σ 2
M

⎛ Z2 , σ ⎞
⎜ a α⎟
Active Portfolio Manager ⎝ ⎠ Control Passive Portfolio Manager
1. Form Active Portfolio 1. Is Za − R − βa . 1. Form a Market Portfolio
2. How good are beta
estimates?
good; (ZM − R ) > 0 ? at minimum cost

3. How good are αi , σ´ij bad 2. Is βa ≈ 0? (good; bad) good;

estimates? good;
bad
bad
Control
1. Is correlation of
βi ⎛ α , σ' ⎞
⎜ i ij ⎟ portfolio with
⎝ ⎠ ⎛ Z , σ2 ⎞ Market = 17
⎜ M M⎟
⎝ ⎠

Beta Analyst Group Analyst αie Micro Analysts Micro Analysts

e
1. Estimate β 1. Correct Analyst bias 1. Estimate α i 1. Estimate
i good; 2
2. Estimate σ´ij Z M, σ M
bad

263
 Robert C. Merton

This rough chart lays out the organizational structure of product and information flows and
responsibility. The dotted line shows decentralized units which could be feasibly separate
organizations. Thus, there are three separate units: (1) the producer of the final product which is
a risky portfolio that should be (at least) efficient and hopefully super-efficient. Its final product
is a risky portfolio (and the information ( Z p , σ 2p)) which is suitable for mixing with the riskless

asset and being held as a final portfolio by individual investors. (2) the producer of an active
portfolio which has a zero-beta and makes optimal use of the micro-analysis done by the security
analysts. Its final product is a risky portfolio (with a zero-beta) whose expected return is greater
than the riskless rate (lies above the Security Market Line). However, it will not, in general, be
suitable for a final portfolio for individual investors. (3) the producer of a passive portfolio
which should be as close as possible to the market portfolio, but constructed at minimum cost.
Estimates of Z M and σ 2M are produced. Its final product is a well-diversified portfolio suitable
for a final portfolio for individual investors who want to follow the naive or passive strategy or
for mixing with active portfolios for those investors willing to try for superior performance
through a blending of “managed” and “unmanaged” portfolios. The various roles and
responsibilities of the “boxed-in” sub-units are now described.

In Unit 1 (Product: Final Risky Asset Portfolio for Individual Investors)

A. The Final Portfolio Manager

1. (a) He receives estimates of Z a , σ a2 , Z M , σ 2M and forms the final portfolio
with return Z p according to the rule (XIII.15)

δ 2 / δ1 = [( Z a - R) σ 2M ]/[( Z M - R) σ a2]

(b) He computes Z p and σ 2p and announces them to his customers. He also

supplies these figures to his Control Management.

2. He receives (complaints/compliments) on the portfolio's performance from
customers (external) and his Control Management (internal).

264
Finance Theory

3. He is responsible for tracing back to the source of the (poor/good) performance.

The three sources are: (a) himself: did he follow the " (δ 2 / δ1)" rule? (b) the

active portfolio: were Z a , σ a2 , β a (= 0) accurate? Was Z a - R - β a ( Z M - R) > 0?

He either complains to or compliments the active portfolio manager. (c) the

passive portfolio: was the passive portfolio highly correlated with the market?
Were the estimates of Z M and σ 2M accurate? He either complains to or
compliments the passive portfolio manager.

B. (Final) Control

1. They monitor the performance of the final portfolio by comparing

Zp - R ZM - R
to , and then, either compliment or complain to the Final Portfolio
σp σM
Manager.

In Unit 2 (Product: Zero-Beta Risky Asset Portfolio representing superior Micro Forecasting)

A. Active Portfolio Manager

1. (a) He receives estimates of β i from the beta analyst and of (α i , σ’ij) from

the group analyst and forms the active portfolio according to the rule
(XIII.10) or (XIII.11): w ia / w ak = ∑ v i ′j α j / ∑ v k ′j α j ; i, k = 1,..., m;
m
∑ w ia (1 - β i) = 1; w aM = 1 - ∑ w ia .
i =1

(b) He computes Z a , σ a2 and announces them to his customer (unit 1), and to
his control management.
2. He receives (complaints/compliments) on the portfolio's performance from the
Final Portfolio Manager (“external”) and his Control Management (internal).

265
 Robert C. Merton

3. He is responsible for tracing back to the source of the (poor/good) performance of

the active portfolio. The three sources are: (a) himself; did he follow the
“(XIII.10) rule?” (b) the beta analyst: if the active portfolio does not have a zero-
beta (β a = 0), is it because the β i estimates supplied were inaccurate? (c) the
group analyst: were the α i and σ ij estimates accurate? After finding the

source(s), he then complains to or compliments the responsible analyst.

B. (Active Portfolio) Control

1. They monitor the performance of the active portfolio by measuring whether the
portfolio shows evidence of superior forecasting capabilities: I.e., does it lie
above the Security Market Line? Is Za - R - β a ( Z M - R) > 0?
2. Does the active portfolio have a zero beta (β a = 0)?
3. They either compliment or complain to the Active Portfolio Manager.

C. Beta Analyst
1. It is his job to provide the estimates of the betas on those stocks being actively
considered by the micro-security analysts.

D. Group Analyst
1. It is his job to get the individual micro-security analyst's estimate (α ie) of the α i

and correct them for bias (historical) of each analyst to get (an unbiased) estimate
αi ; he must also estimate σ’2
i and σ i ′j . These estimates are provided to the

Active Portfolio Manager.

2. He receives (complaints/compliments) on these estimates from the Active
Portfolio manager.
3. He is responsible for rating the individual analyst's performance and deciding on
whether the analyst is worth keeping. A (rough) measure would be the ratio

266
Finance Theory

(α i / σ i2) . The larger (in absolute value) this ratio is on average, the more valuable

the analyst.

E. Micro Security Analysts

1. They are responsible for estimating the mean return on the nonsystematic part of
the return on individual securities. They are not responsible for that part of the
returns which can be explained by macro-market effects.

In Unit 3 (Product: A portfolio which is perfectly correlated with the Market and estimates
of ZM and σ2M showing superior Macro Forecasting capabilities)

A. Passive Portfolio Manager

1. He must form a well-diversified portfolio which is as nearly perfectly correlated
with the market portfolio as possible at minimum cost.
2. He receives estimates Z M and σ 2M from the Macro Analysts and then reports
them to his customer (unit 1) and to his control management.
3. He receives (complaints/compliments) on the portfolio's performance (i.e., how
highly correlated it was with the market) and on the estimates Z M and σ 2M from
the Final Portfolio Manager (“external”) and his Control Management (internal).
4. He is responsible for the portfolio's performance. If the estimates by the Macro-
Analysts are (good/poor), he registers the appropriate compliment or complaint
with them. He measures their performance by comparing a strategy of a variable
ZM - R
position in the market depending on the estimated ratio and a buy-and-
σM

hold strategy.

267
 Robert C. Merton

B. (Passive Portfolio) Control

1. They monitor the performance of the passive portfolio by estimating its
correlation with the market, and then complain to or compliment the Passive
Portfolio Manager.

C. Macro Analysts
1. They are responsible for estimating the expected return and variance of the market
portfolio or equivalently, the slope of the Capital Market Line.

Of course, this is not the only way the organization could be structured. However, it does have
the property that the three major sub-units operate in a decentralized fashion. All measurements
of performance are net of operating costs (e.g., management fees, salaries, transactions costs,
computer costs, etc.). If, for example, the active portfolio “extra” returns do not cover costs,
then, like an unprofitable division of a manufacturing firm, it should be dropped. Many of the
basic techniques used here could be applied to the management of a manufacturing corporation.
Note that each level of decision making is subject to two forms of control: external and internal.
Each decision-maker's performance is judged on that aspect of the operation for which he is
responsible and over which he has the authority to do something about. Note: that throughout
there is a kind of “automatic” control which keeps “poor” performers from having much impact
on the final product. In the special case where σ i ′j = 0 (i ≠ j), the weight of the individual micro

analyst in the active portfolio is proportional to which means that if his error in estimating is
large (i.e., σ i2′ large), then independent of α i , “his” security does not get into the portfolio.

Similarly, if the aggregate error in the active portfolio estimates are large, then
⎛ Za - R ⎞
⎜ ⎟
⎜ σ2 ⎟
⎝ a ⎠

268
Finance Theory

will be small and it will have little weight in the final portfolio. In the limit as
⎛ Za - R ⎞
⎜ ⎟
⎜ σ2 ⎟ → 0 ,
⎝ a ⎠

the final portfolio is the market portfolio as it should be for someone with no forecasting
capability.

269
XIV. THEORY OF VALUE AND CAPITAL BUDGETING UNDER UNCERTAINTY

The valuation formulas and capital budgeting rules developed in Sections VI and VII take
into account the intertemporal characteristics of the firm's cash flows. In this section, the
analysis is extended to explicitly recognize the uncertainty associated with these flows. The
introduction of uncertainty makes the analysis much more complex. Hence, we begin with the
simple case of a one-period firm whose end-of-period output is distributed to its stockholders
through liquidation. In studying this case, it is further assumed that the equity market is such that
the Capital Asset Pricing Model (of Section XI) holds, and therefore, in equilibrium, securities
are priced so as to satisfy the Security Market Line. Having analyzed this case, we then derive
the valuation formulas for a multi-period firm.
It is assumed throughout this section that the firm is all-equity financed. Although, in
principle, the capital budgeting and financing (capital structure) decisions cannot be made
independently, the study of the financing decision in Section IX and, in particular, the derived
Modigliani-Miller Theorem, suggests that the method of financing should generally have little, if
any, impact on the choice of investment projects by the firm. Indeed, a good rule of thumb is to
be suspicious of projects which do not look attractive when evaluated on an all-equity financed
basis but which do appear attractive when presented in conjunction with a "creative" financing
plan. [There are, of course, exceptions to this rule as for example, when the government
provides subsidies to certain private sector projects by using below-market interest rate loans or
guarantees loans.] With this as background, we now turn to the analysis of a one-period firm.
Suppose that firm i has made (or is considering making) an investment of $ Ii in a
project with end-of-period random variable cash flow of Ii ~x i where ~x i is the random variable

average cash flow per dollar of investment. Let x i ≡ E(~x i) and νi2 ≡ Var (~x i) . If Vi is the

current market value of the firm after the investment is made and if the firm has no other
projects, then the return per dollar on the shares of the firm is given by

≡ I i x% i with ≡ E( ) =
I i xi
and Var( ) = I i2ν i2 .
Zi Zi Zi Zi The covariance of the return of the
Vi Vi V i2
firm's equity with the market, σiM , is given by

270
Finance Theory

⎡
I i xi ,% ⎤ I iν iσ M ρ iM
σ iM ≡ Cov[ Z i , Z M ] = Cov ⎢ ZM ⎥=
⎣ Vi ⎦ Vi

where ρiM ≡ the correlation coefficient between ~x i and ZM .

In equilibrium, the equity of firm i will be priced so as to satisfy the Security Market Line:

σiM σ
Zi - R = (
2 ZM
- R) = λe iM
σM σM
(XIV.1)
Ii νi ρiM
= λe ( )
Vi

where λ e ≡ ( Z M - R)/ σ M is the Market Price of Risk and it does not depend upon the decisions
Ii x i Ii νi ρiM
made by firm i . Substituting for Zi into (XIV.1), we have that - R = λe ( ), or
Vi Vi

rearranging terms, that

Ii
(XIV.2) Vi = [ x - λ e νi ρiM ]
R i

(XIV.2) gives the equilibrium market value of the firm after having expended $ Ii in resources
in the project. Under what conditions should the firm make this expenditure and take on the
investment? If the firm operates so as to maximize its market value, then it should take all
projects which increase its market value; be indifferent to projects which leave its value
unchanged; and not take projects which will lower its market value. Thus, it should take the
project if Vi - Ii > 0 ; be indifferent if Vi - Ii = 0 ; and not take it if Vi - Ii < 0 . From (XIV.2), we
have that

Ii
(XIV.3) Vi - Ii = [ x - R - λe νi ρiM ] .
R i

271
 Robert C. Merton

So, for a given Ii , if [ x i - R - λe νi ρiM ] > 0 , take it; if [ x i - R - λe νi ρiM] = 0 , be indifferent; and if
[ x i - R - λe νi ρiM] < 0 , do not take it.

p
Define: The beta of a project or project beta, βi , by

Cov[ x% i , Z M ] ν iσ M ρ iM ν i ρ iM
β ip ≡ = = .
Var( Z M ) σ 2M σM

Note: The beta of the equity of firm i (its "market beta") is given by

σ iM = ( ν i ρ iM )( I i ) = I i p .
βi ≡ βi
σ 2M σM Vi Vi
> >
Hence, βip = βi as Vi - Ii = 0 .
< <

Consider a concept similar to the Security Market Line except use project instead of market
p
betas: I.e., Define the Project Market Line by x - R = β ( ZM - R) where x is the expected cash
p
flow per dollar of investment in the project and β is the project beta.

272
Finance Theory

The graph of the Project Market Line is analogous to the Security Market Line in Section
XI. However, unlike the SML, this graph relates non-market assets or projects returns to market
returns. From (XIV.3), we express the capital budgeting rule as

x i > R + β i ( Z M - R) ⇒ V i > I i ⇒ Take the Project

p
(XIV.4a)

x i = R + β i ( Z M - R) ⇒ V i = I i ⇒ Indifference to the Project

p
(XIV.4b)

x i < R + β i ( Z M - R) ⇒ V i < I i ⇒ Do not take theProject

p
(XIV.4c)

In the graph, project #2 corresponds to (XIV.4a); project #3 corresponds to (XIV.4b); project #1

corresponds to (XIV.4c). That is, the firm should take all projects that lie above the Project
Market Line and reject all those that lie below the line.

273
 Robert C. Merton

In the capital budgeting analysis in Section VII, we defined the cost of capital k and used
it for deriving capital budgeting rules rather than the riskless rate of interest. Although in the
certainty environment of that section, the two must be equal to avoid arbitrage, it was noted there
that the distinction was made in preparation for the analysis of projects whose future cash flows
are uncertain. To connect the results here with the rules of this earlier section, we restate the
capital budgeting rule in terms of the cost of capital.
Define the cost of capital for project #i, k i , by

k i ≡ R + β i ( Z M - R) .
p
XIV.5)

It follows from (XIV.4) that the correct rule for choosing projects is to take all (independent, as
defined in Section VII) projects whose expected return per dollar of investment, xi , exceeds the
associated cost of capital, k i , and to reject all projects whose expected return per dollar is less
p
than its cost of capital; ki is also called the "hurdle rate" for project i . The larger is βi , the
larger is the hurdle rate or the minimum required expected return on the project in order to justify
p
taking the project. In analogous fashion to securities, βi is the appropriate measure of the risk
of project i, and the riskier is the project, the higher is its hurdle rate. As with securities, it is
p
the project's systematic risk (βi ) that matters in making the decision whether to invest or not,

and not the project's total risk ( νi2) .

Two important implications for firm investment behavior (which were not evident from
the certainty analysis of Section VII) follow from the derived capital budgeting rule: First, the
cost of capital to be used for evaluating a project is the one associated with the project and not
the firm evaluating the project. That is, two different firms evaluating the same project (by
"same" we mean that ~x i has the identical distribution from both firm's perspectives) should use
the same cost of capital [given by (XIV.5)]. To see this, note by inspection of (XIV.5) that k i
depends only upon the distribution of ~x i and its joint distribution with the market. It does not
in addition depend upon the joint distribution of ~x i with other projects that the firm may have

274
Finance Theory

(or plan to undertake).

Second, since the correct decision on the project depends only upon its systematic risk
(and not its total risk), unlike a person selecting his optimal portfolio, a firm has no need to
consider (internal) diversification. This important conclusion will be discussed in depth in
Section XV.
Having established the correct capital budgeting rule in a one-period model, we now turn
to the evaluation of the firm and its projects in a multi-period or intertemporal framework.

Theory of Value Under Uncertainty (Multi-period Cash Flows)

Before proceeding to the development of the valuation formulas, we provide a quick

review of conditional expectation. (For further discussion, consult any reasonable book on
probability.)

Digression: Review of Expectation and Joint Probabilities

Let X be a random variable which can take on the values x1 , x 2 , x3 ,... .

Let Y be a random variable which can take on the values y1, y2 , y3 ,... .
Let P{X = x j} = f(x j) be the probability that X = x j , j = 1,2,3,... .

Let P{Y = yk} = g( yk ) be the probability that Y = yk , k = 1,2,3,... .

Let P{X = x j , Y = yk} = p( x j , yk ) be the probability that X = x j and Y = yk j, k = 1,2,... .

{p(x, y)} is called the joint distribution for X and Y and {f(x)} and {g(y)} are called the

marginal distributions for X and Y, respectively.

(XIV.6) f(x j) = ∑ p( x j , yk ) ; g( yk ) = ∑ p( x j , yk ) .
k j

Let P{Y = yk | X = x} be the conditional probability that Y = yk , given that X = x j .

275
 Robert C. Merton

p( x j , yk )
(XIV.7) P{Y = yk | X = x j} = .
f(x j)

Let E(X) = (unconditional) expected value of X = ∑ x j f(x j) .

j

Let E(Y | X = x j) = conditional expected value of Y , given that X = x j .

yk p( x j , yk )
(XIV.8) E(Y | X = x j) = ∑ yk p {Y = yk | X = x j } = ∑
k k f(x j)

E(E(Y | X)) = ∑ E(Y | X = x j)f(x j) = ∑ ∑ yk p(x j , yk )

j j k
(XIV.9)
= ∑ yk (∑ p(x j, yk )) = ∑ yk g( yk ) = E(Y)
k j k

If X and Y are mutually independent, then

(XIV.10) E{XY} = E(X)E(Y) ; p( x j , yk ) = f( x j)g( yk ) .

For purposes of this course, we will be dealing primarily with random variables describing an
outcome as of a given date t . E.g., ~π (t) may be a random variable describing profits for date t.
In general, the distribution for such a random variable, X(t), will depend on outcomes which
occur at an earlier date: denote these random variables by Y(t - 1), Y(t - 2),... . If the value of
X(t) = function of these random variables = F (Y (t -1), Y (t - 2),... ), then the expected value of

X(t) will depend on the point in time at which the expectation is computed. Let " E t " denote

the conditional expectation operator, conditional on knowing all (relevant) information that has
~
occurred up to and including time t . Then, E t {X(t)} = x(t), the particular value that X(t) took on
~
at time t and X(t) is not a random variable relative to time t. If X(t) depends on Y(t - 1),...,
then

276
Finance Theory

~
E0 {X(t)} will include the joint distribution over all {Y(t)}
~
conditional on knowing that Y(0) = y0 . E t -1{X (t)} will be the conditional expectation,

conditional on knowing that Y t -1 = y t -1 , Y t - 2 = y t - 2 ,...,Y 0 = y 0 . From (XIV.9), we have that

~ ~
(XIV.11) E t -2 {E t -1[X (t)]} = E t -2 {X (t)}

or more generally,
~ ~
(XIV.11') E t -k {E t- j[X (t)]} = E t-k [X(t)] for k ≥ j ≥ 0

End of Digression -

Valuation Under Uncertainty: The General Case

The derivation of the valuation formula follows the same format as the certainty analysis
~
in Section VI. If Z(t) is the (random variable) return per dollar from investing in the equity of
the firm between time t and t+1, then, by definition,
~
~ d (t + 1) + ~s (t + 1)
(XIV.12) Z(t) =
s(t)

where tildes ~ denote random variables relative to time t (e.g., s(t) will be known for certain
at time t).
Let k(t) be the equilibrium market required expected rate of return for investing in the
firm between t and t+1. (Again, k(t) may be a random variable relative to dates earlier than t,
but at time t, it is known). Then, in equilibrium, the price per share of the stock at time t must
~
be such that E t {Z(t)} = 1 + k(t) or

1 ~ ~
(XIV.13) s(t) = E t [ s (t + 1) + d (t + 1)] ,
[1 + k(t)]

277
 Robert C. Merton

and for equilibrium, (XIV.13) must hold for each t.

Consider a firm which will remain in business for T periods (from now) and then
liquidates. As in the certainty analysis, to deduce the value of the stock today, we first go
forward in time and then, work backwards to today (time zero).
At time T in the future, the firm will pay its last dividend per share, d(T), and as
discussed in the parallel analysis in Section VI, without loss of generality, we can assume that the
salvage value at that time is zero, and hence, with probability one, the ex-dividend price per share
at time T will be zero (i.e., S(T) = 0).
Consider an investor at time (T–1): If he buys one share of stock, his expected dollar
~
return at time T is ET-1[ d (T)] . For the market to be in equilibrium, we have that S(T–1) must
be such as to satisfy (XIV.13). I.e.,

1 ~
(XIV.14) S(T - 1) = ET-1[ d (T)] .
[1 + k(T - 1)]

Consider when we reach time (T-2). In order for the market to be in equilibrium S(T–2)
must again satisfy (XIV.13). I.e.,

1 ~ ~
(XIV.15) S(T - 2) = ET-2 [ d (T - 1) + S(T - 1)] .
[1 + k(T - 2)]

Substituting for S(T–2) from (XIV.14) into (XIV.15), we have that

~
[ d (T - 1)] 1 1 ~
(XIV.16) S(T - 2) = ET-2 + E T -2 { ~ ET-1[ d (T)]}
[1 + k(T - 2)] [1 + k(T - 2)] [1 + k (T - 1)]

where k(T-1) has a ~ over it because relative to time (T–2) it may be uncertain (i.e., a random
~
1 ~ d (T)
variable). Noting that ET-1[ d (T)] = ET-1{ } because k(T–1) is not a
[1 + k(T - 1)] [1 + k(T - 1)]

random variable relative to time (T-1), we have that

278
Finance Theory

1 % d% (T ) d% (T )
E T -2{ E T -1[ d (T )]} = {
E T -2 E T -1[ ]} = E T -2[ ] using the fundamental
[1 + k% (T -1)] [1 + k% (T -1) [1 + k% (T -1)

relationship on conditional expectations given in (XIV.11) or (XIV.11'). Thus, we can rewrite

(XIV.16) as
~
1 ~ d (T)
S(T - 2) = ET-2 {d (T - 1) + ~ }
[1 + k(T - 2)] [1 + k(T - 1)]
(XIV.17) ~ ~
d (T - 1) d (T)
= ET - 2 { + ~ }.
[1 + k(T - 2)] [1 + k(T - 2)][1 + k (T - 1)]

At time (T-3), we have that for markets to clear that S(T-3) must satisfy (XIV.13) or

1 ~ ~
(XIV.18) S(T - 3) = ET-3 {d (T - 2) + S(T - 2)}
[1 + k(T - 3)]

Substituting from (XVI.17) into (XIV.18); noting that k(T-2) may be a random variable relative
to time (T-3) and using the result that ET-3 • ET-2 = ET-3 , we can rewrite (XIV.18) as
~ ~
1 ~ d (T - 1) d (T)
S(T - 3) = ET-3 { d (T - 2) + ~ + ~ ~ }
[1 + k(T - 3)] [1 + k(T - 2)] [1 + k (T - 2)][1 + k(T - 1)]
~ ~
d (T - 2) d (T - 1)
(XIV.19) = ET-3{ + ~
[1 + k(T - 3)] [1 + k(T - 3)][1 + k (T - 2)]
~
d (T)
+ ~ ~ }
[1 + k(T - 3)][1 + k (T - 2)][1 + k (T - 1)]

Proceeding inductively in this backwards fashion, we arrive at the price per share today (time
zero) which ensures that an investor buying the stock at any time and selling at any other time
will face an ex-ante expectation of a fair return and that the markets will clear. I.e.,

279
 Robert C. Merton

~ ~
d (1) d (2)
S(0) = E0 { + + ... +
[1 + k(0)] [1 + k(0)][1 + k(1)]
~
d (T)
~ ~ }
[1 + k(0)][1 + k (1)]...[1 + k (T - 1)]
T ~
(XIV.20) d (t)
E0 {∑ t }
~
t =1
∏ [1 + k (s - 1)]
s =1
T ~
d (T)
= E0 {∑ ~ t}
t =1 [1 + K (t) ]

~
where K (t) is a random variable defined for notational convenience as

1/t
⎡ t % - 1)] ⎤⎥
K(t) ≡ ⎢∏ [1+k(s
% -1
⎣ s=1 ⎦

Comparing (XIV.20) with the certainty case, in Section VI, there are some obvious similarities.
~
Moreover, if d (t) = d(t) (i.e., future dividends are known with certainty), then by arbitrage
~
d (t) d(t)
k(t) = r(t) ; E0 [ t
]= , and (XIV.20) becomes the same as in VI. As in the
[1 + k(t) ] [1 + k(t) ]t

certainty case, we can write (XIV.20) in its infinite-lived form and for an all-equity financed
firm, we have that V(0) = n(0)s(0) . I.e.,

∞ ~
d (t)
(XIV.20') S(0) = E0 { ∑ ~ t}
t =1 [1 + K (t) ]

and

∞ ~
n(0)d (t)
(XIV.21) V(0) = E0 { ∑ ~ t}
t =1 [1 + K (t) ]

While (XIV.20), (XIV.20'), and (XIV.21) represent a completely general valuation formula, they
are operationally of little use without some further specification of the structure for the
~ ~
probability distributions for both the {d (t)} and the {k (t)} .

280
Finance Theory

The balance of this section will be devoted to specific forms for (XIV.20') and (XIV.21)
~
deduced from special characteristics assumed for the structure of the market (i.e., k (t) ) and the
~
firm-specific characteristic (i.e., d (t) ). It should be remembered that these cases are only
representative, and in any given situation, it may be appropriate to return to the general form
(XIV.20') and (XIV.21).

Cost of Capital:

"The cost of capital" is a term often used in corporate finance, and is usually defined as
the opportunity cost (expressed as a rate of return) to investors of a given risk project. It is
definitely an external (to the firm) rate. While in certainty analysis, it is well-defined (namely,
equal to the {r(t)}), under uncertainty, it is a "fuzzy" notion. Nonetheless, the term is usually
taken to describe the structure of the {k(t)} .

Special Cases of Valuation Under Uncertainty

~
Case A. Suppose that the required expected returns {k (t)} and the dividend stream per
~ %
share {d (t)} are mutually independent. Define d(t) ≡ E 0 [ d(t)] = expected
dividend per share at time t. Then, from (XIV.10), we have that
~
d (t) ~ 1
E0 [ ~ t ] = E0 [ d (t)] • E0 [ ~ t ]=
[1 +K (t) ] [1 + K (t) ]

d (t) 1 1
where ρ(t) is defined by ≡ E0 [ ].
[1 + ρ(t) ]t
[1+ ρ (t) ] t % ] t In this case,
[1+K(t)

(XIV.30') and (XIV.21) can be written as

∞ d(t)
(XIV.22) S(0) = ∑ t
and
t =1 [1 + ρ(t) ]

281
 Robert C. Merton

∞ n(0)d (t)
(XIV.23) V(0) = ∑ t
.
t =1 [1 + ρ(t) ]

~ ~
Warning: ρ(t) ≠ E0 [K (t)] and ρ(t) ≠ E0{ k (t - 1)} .

~ ≡
Case B: Suppose that the {k (t)} are nonstochastic and constant, i.e., k(t) k . Then
t

(XIV.20') and (XIV.21) can be rewritten as

∞ d (t)
(XIV.24) S(0) = ∑ t
and
t =1 [1 + k ]
∞ n(0)d(t)
(XIV.25) V(0) = ∑ .
t =1 [1 + k ]t

In this case, k is the required expected rate of return by investors in the firm, i.e., the cost of
capital. Therefore, the value of the stock is equal to the present discounted value of expected
dividends per share, discounted at the cost of capital. This is very close to the certainty formula
in VI where "expected dividends" replace "dividends received" and the market "expected rate of
return" replaces the market "realized rate of return."

~
Case C. A slight generalization of Case B is when the {k (t)} are nonstochastic, but vary in a
~
deterministic way over time. I.e., k (t) = k(t). Then (XIV.20') and (XIV.21) can be written as

∞ d (t)
(XIV.24') S(0) = ∑ t
and
t =1 [1 + K(t) ]
∞ n(0)d (t)
(XIV.25') V(0) = ∑ t
.
t =1 [1 + K(t) ]

Note: the K(t) are nonstochastic because the k(t) are not. However, K(t) is not the cost of
capital, and in an analogous fashion to the R(t) in the certainty case, the required expected return
is not K(t). However, K(t) is the average expected compound return from investing in the
282
Finance Theory

stock (including reinvesting dividends paid) from time zero to time t. I.e., if at time zero, one
invested W0 dollars in the stock and reinvested all dividends received in the stock, then the
t
~
expected value of the position at time t would be E0 [ W t ] = W0 ∏ [1 + k(s - 1)] = W0 [1 + K(t) ] or
t

s =1

1/ t
E 0[W% t ] ⎡ E [ % ]⎤
= [1 + K (t )]t or ⎢ 0 W t ⎥ -1 = K (t ) .
W0 ⎣ W0 ⎦

~
Case D. Suppose that the {k (t)} are nonstochastic and constant, and the expected dividend per
t
share grows at a constant rate per period g. I.e., d (t) = d(0)[1 + g ] Substituting for d (t) into
(XIV.24), we have that

∞
1+ g t ∞ 1+ g
S (0) = ∑d (0)[ ] = d (0) ∑ y t for y equiv
t =1 1+ k t =1
1+ k
d (0)[1+ g ]
(XIV.26) = , provided y < 1 (i.e., k > g )
k -g
d (1)
= because d (1) = d (0)[1+ g ]
k -g

V (0) = n(0) S (0) because D% (t ) ≡ n% (t -1)d% (t )

(XIV.27) E 0{D% (1)} = D(1) = E 0[n(0)d% (1)]
= n(0) E 0[d% (1)] = n(0)d (1)

In the certainty analysis of Section VI, the cash flow accounting identity was used to
show that the four statements of what determines the value of a firm are equivalent. Fortunately,
the analysis presented in that section carries over almost completely to the uncertainty case.
Under the assumption that the firm is financed entirely by equity, the current market value of the
firm is given by V(0) = n(0)S(0) where n(0) is the number of shares currently outstanding.

283
 Robert C. Merton

Moreover, at each point in time t, V(t) = n(t)S(t). Equation (XIV.21) gives an expression for
V(0), and from (XIV.13), we have that

n(t) % +1)]
% +1) + d(t
V(t) = n(t)S(t) = E t [ S(t
[1+ k(t)]
1
= E t {n(t)S(t % +1)}
% +1) + n(t)d(t
[1+ k(t)]
(XIV.28)
1 % +1) - [ n(t
% +1)+ n(t)d(t % +1)}
= % +1)S(t
E t{ n(t % +1) - n(t)]S(t
[1+ k(t)]
1 % +1)}.
= E t{V% (t +1)+D(t
% +1) -m(t
% +1)S(t
[1+ k(t)]

Moreover, the accounting identity in Section VI, is an identity, and therefore, holds for
each possible outcome. I.e., it states that
~ ~ (t + 1)~
R (t + 1) + m
~ ~
S(t + 1) = O(t + 1) + D(t + 1)
(XIV.29)

Or equivalently, that
~ ~ (t + 1)~
D(t + 1) - m
~ ~
S(t + 1) = R (t + 1) - O(t + 1)
(XIV.30a)
~ ~ (t + 1)~
D(t + 1) - m
~ ~
S(t + 1) = X (t + 1) - I (t + 1)
(XIV.30b)
~ ~ (t + 1)~ ~
(XIV.30c) D(t + 1) - m S(t + 1) = ~
π (t + 1) - i (t + 1)

Substituting from (XIV.30) into (XIV.28), we have that

1 ~ ~ ~
(XIV.31a) V(t) = E t {V (t + 1) + R(t + 1) - O(t + 1)}
[1 + k(t)]
1 ~ ~ ~
(XIV.31b) V(t) = E t {V (t + 1) + X (t + 1) - I (t + 1)}
[1 + k(t)]
1 ~ ~ ~
(XIV.31c) V(t) = E t {V(t + 1) + π (t + 1) - i (t + 1)} .
[1 + k(t)]

284
Finance Theory

We can solve (XIV.31) using the same backward technique used to solve for S(0) starting with
(XIV.13). Namely, we have that

∞ ~ ~
[R (t) -O(t)]
(XIV.32a) V(O) = E0 {∑ ~ t }
t =1 [1 + K (t)]

∞ ~ ~
[X (t) - I (t)]
(XIV.32b) V(O) = E0 {∑ ~ t}
t =1 [1 + K (t)]
~
[~
∞ π (t) - i (t)]
(XIV.32c) V(O) = E0 {∑ ~ t}
t =1 [1 + K (t)]

Coupled with (XIV.21), (XIV.32) and (XIV.21) provide four alternative but equivalent
expressions for the value of the firm under uncertainty.
Using (XIV.21) and (XIV.30), we have the following expressions for the expected change
in the value of the firm from time t to t+1:
~ ~ ~ ~
(XIV.33a) E t [V (t + 1) - V(t)] = E t {∆ V t} = k(t)V(t) + E t [m(t + 1)S(t + 1) -D(t + 1)]
~ ~
(XIV.33b) E t {∆ V t} = k(t)V(t) - E t {R (t + 1) - O(t + 1)}
~ ~
(XIV.33c) E t {∆ V t} = k(t)V(t) - E t {X(t + 1) - I (t + 1)}
~ ~
(XIV.33d) E t {∆ V t} = k(t)V(t) - E t {π (t + 1) - i (t + 1)}

so, from (XIV.33), the expected change in the value of the firm is not equal to the expected
change in shareholders' wealth {i.e., k(t)V(t)}.
As promised, the evaluation of projects in an uncertain environment is considerably more
complex than in the certainty case. While the formulas for value under certainty derived in
Section VI do bear some resemblance to the ones derived here, the valid application of the
former has been shown to be limited to cases of projects with specific distributional
characteristics and specific market structures (e.g., CAPM).
While further development of these techniques are beyond the scope of the course, we
285
 Robert C. Merton

end this section with a brief discussion of the certainty equivalent method of valuation.
~
The certainty equivalent to a particular cash flow X(t) is defined to be that number of
dollars, Xce (t), such that an investor would be indifferent between receiving Xce (t) for certain

at time t or the random variable cash flow X% (t) at time t. Since, by definition, the market

would be willing to exchange X ce(t) dollars for certain for the X% (t), it must be that

∞ α(t)X (t)
(XIV.34) V(0) = ∑
t =1 [1 + r ]t

where α (t) ≡ X ce(t)/ X (t) and X ( t ) ≡ E 0 [ X% (t)] .

While, in general, one might expect α(t) < 1, it need not be as for example in the CAPM
~
if X(t) has a negative beta. Moreover, α(t) need not be a decreasing function of t. That is, it
is not always true that the farther in the future a cash flow will occur, the more uncertainty or risk
it must have.

286
XV. INTRODUCTION TO MERGERS AND ACQUISITIONS: FIRM
DIVERSIFICATION

In the introduction to Section VII, it was noted that firms can acquire assets by either
undertaking internally-generated new projects or by acquiring existing assets of other firms.
Having examined the former there and again in Section XIV, we now turn to the latter.
Under the operational criterion for good management of maximizing current shareholders'
wealth, there are essentially three reasons for considering the acquisition of another company:

1. Synergy: By combining the two companies, the value of the operating assets of the
combined firm will exceed the sum of the values of the operating assets of the two
companies taken separately. Such synergy will occur if there are economies of scale in
marketing, purchasing of materials, plant size, and distribution system. It can also occur
through the elimination of duplicate efforts in management or research and development.
Such economics are most likely to occur with either horizontal or vertical mergers. In
essence, the value goes up because the factors of production are more efficiently
organized in the combined firm.

2. Taxes: The market value of the firm reflects its value to the private sector. Of course,
since the firm pays taxes (or may pay taxes in the future), there is an additional "shadow"
value of the firm to the public sector in the form of the present value of its tax payments.
The sum of the market value and this "shadow" value is the value of the firm to society.
In the case of synergy, the value of the firm to society is increased with a corresponding
increase in both the market and shadow values of the firm. However, if a combination of
two firms can reduce the combined present values of these firms' tax payments taken
separately, then the market value of the combined firm can exceed the sum of the values
of the two firms taken separately even if the value of the combined firm to society is just
equal to the sum of the values to society of the two firms. I.e., this combination does not
increase the total value to society, but it does redistribute the total between the

287
 Robert C. Merton

shareholders of the firms and the public sector. Two examples are: (a) a more-effective
use of a tax-loss carryover; (b) increased debt capacity for the combined firm which may
reduce taxes if there is a "tax-shield" value to the deductibility of interest [see Section IX
for further discussion].

3. The Firm to be Acquired is a "Bargain": If the firm to be acquired has a market value
which is less than its "fair" value, then by acquiring the firm, the management of the
acquiring firm can increase its stockholders' wealth. There are two distinct reasons why a
firm could be selling for less than "fair" value. The first is that relative to the acquiring
firm's information set, the stock market is not efficient in the sense to be discussed in
Section XVII. That is, the management of the acquiring firm believes that it has
information such that if this information were widely-known, the market value of the firm
to be acquired would be higher than its acquisition cost. If this is the principal reason for
the acquisition, then the management's behavior is identical to that of a security analyst
whose job it is to identify mispriced securities. In terms of the CAPM and Section XIII,
the management believes it is purchasing a security with a positive "alpha" (α) . Hence,
all the warnings about being able to "beat the market" given in that section apply equally
well here.
A second reason why a firm could be selling for less than its "fair value" is that the firm
to be acquired is currently being mismanaged. That is, through either incompetence or
malevolence, the current management is not managing the firm's resources so as to maximize the
market value of the firm. Unlike the first reason, this reason is completely consistent with an
efficient capital market. Indeed, as discussed at length in Section III, from society's point of
view, this reason is probably the most important one for permitting mergers and takeovers.

288
Finance Theory

Firm Diversification

Notable by its absence among the three reasons for acquisitions is diversification: That
is, the acquisition of another firm for the sole purpose of reducing the volatility (variance or
"total" riskiness) of the firm's operations. Although "diversification" is a frequently cited reason
for an acquisition, it is often not the "real" reason. More often than not, it will be for one of the
three reasons already given. However, if diversification is the real reason, then the acquisition
route will in general be an inefficient way to achieve it.
The argument for firm diversification is often presented by analogy with an individual
investor where we have seen that diversification is quite important. However, this type of
argument simply illustrates the pitfalls of treating the firm "as if" it were an individual household
with exogenous preferences rather than as an economic organization designed to serve specific
economic functions.
To show why firm diversification is not an important activity for management and if it is
undertaken, why the acquisition route is inefficient, we begin with an explicit analysis of the
value of the firm under the capital asset pricing model. Let there be two firms where each firm
has a single project as described in the beginning of Section XIV. From formula (XIV.2), the
value of firm i (i = 1,2) is given by

Ii
(XV.1) Vi = [ x - λ e νi ρiM ], i = 1,2.
R i
Suppose that firms #1 and #2 merge to form firm #3. In an analogous fashion to firms #1 and #2,
define I3 as the investment in firm #3 and ~x 3 as the random variable end-of-period cash flow of
firm #3 per dollar of investment. If no changes in the investment plans of the firms occur as a result of
the combination, then I3 ≡ I1 + I2 and ~
x3 ≡ δ ~
x1 + (1 - δ) ~
x2 where

δ≡ I 1 . 2 ≡ VAR( ) = 2 2 + (1- δ 2 2 + 2δ (1- δ )Cov( , ).

ν3 x% 3 δ ν 1 ) ν2 x% 1 x% 2
I 1+ I 2

289
 Robert C. Merton

(XV.2)
Cov[~
x 3 , ZM ]
ρ3M ≡ V
ν3 σM
δCov[ x1 , ZM] + (1 - δ)Cov[~
~ x 2 , ZM ]
=
ν3 σ M
δ ν1 σM ρ1M + (1 - δ) ν 2 σM ρ2M
=
ν3 σM
δ ν1 ρ1M + (1 - δ) ν 2 ρ2M
=
ν3

From Section XIV, (XIV.2), the value of firm #3 will satisfy

I3
(XV.3) V3 = [ x - λe ν3 ρ3M]
R 3

Substituting into (XV.3) for ρ3M from (XV.2), we have that

I3
(XV.4) V3 = { x - λe [δ ν1 ρ1M + (1 - δ) ν 2 ρ2M]}
R 3

Noting that x ≡ E[~x 3] = δ x1 + (1 - δ) x 2 , we have that

I3
(XV.5) V 3= { δ [ x1 - λ eν 1 ρ 1M ] + (1- δ )[ x 2 - λ eν 2 ρ 2M ]}
R
But I3δ = I1 and I3(1–δ) = I2 . Hence, from (XV.5), we have that

I1 I
V3 = [ x1 - λ e ν1 ρ1M ] + 2 [ x 2 - λe ν 2 ρ2m]
(XV.6) R R
= V1 + V 2 .

Thus, the value of the combined firm will just equal the sum of the values of the two firms prior
to the merger.
In connection with both mergers and firms possibly undertaking many (independent)
capital budgeting projects, we generalize the above demonstration to a firm with m projects.

290
Finance Theory

Let firm P take on m different projects where physical investment in project i is Ii

m
and the random variable end-of-period cash flow is IP = ∑ Ii , and total firm end-of-period cash
i =1

flow per dollar of physical investment x P , can be written as

m
x P = [∑ Ii x i]/ IP
i =1
(XV.7) m
= ∑ δi x i
i =1

m
where δ i ≡ I i / I P and ∑δ i = 1 .
i=1

It follows from (XV.7) that

m
(XV.8) x P = ∑ δi x i
i =1

and
m m
(XV.9) Var( x P ) ≡ ν 2P = ∑∑δ iδ j Cov( xi , x j ) .
i=1 j=1

It follows also that

Cov( x P , ZM ) = ρPM ν P σ M
(XV.10) m
= ∑ δi ρiM νi σM .
i =1

291
 Robert C. Merton

From (XIV.2), (XV.8), and (XV.10), we have that

IP
VP = [ x - λe ν P ρPM ]
R P
IP m m
= [∑ δi x i - λ e ∑ δi νi ρiM ]
R i =1 i =1
(XV.11)
IP m
= [∑ δi ( x i - λe νi ρiM )]
R i =1
m
= ∑ Vi
i =1

where Vi = Ii ( x i - λe νi ρiM )/R is the "stand-alone" value of project i .

Hence, diversification does nothing to the market values of the firms and hence,
according to the value-maximization criterion, it is not important. The result shown in (XV.6)
and (XV.10) is called value additivity and can be shown to obtain in quite general structures
(provided that there exists a well-functioning capital market).
An intuitive explanation of why the market values are unaffected even though the
combined firm may have a smaller total risk (variance) than the individual firms is as follows: In
order for investors to be willing to pay a higher price for the combined firm than they were
willing to pay for the two firms separately, the act of combining the two firms must provide a
"service" to the investors which they were previously unable to obtain. However, prior to the
combination, any investor could purchase shares of either or both firms in any mix he wants.
And, in particular, in the case of the merger, the investor could purchase the shares of firm #1 to
firm #2 in the ratio V1 / V 2 which is exactly the ratio implicit in the combined firm. Hence, each
investor could achieve for himself (prior to the merger) the same amount of diversification (of the risks
of the firms #1 and #2) as is provided by the combined firm, and therefore, the merger provides no new
diversification opportunities to investors. For that reason, investors would not pay a premium for the
combined firm.
Although it will not be the case for the capital asset pricing model, it is possible that the
combined firm could sell for less than the sum of the values of the two separate firms, i.e., that
firm diversification could "hurt" market value. The reason is that post-consolidation, investors
have fewer choices for portfolio construction than they did pre-consolidation. For example, prior
292
Finance Theory

to the merger, an investor could hold positive amounts of firm #1 and none of firm #2 or vice
versa. Post the merger, the only way that an investor can hold firm #1 is to invest in the
combined firm #3 which means he must also invest in firm #2.
Indeed, he can only invest in firm #1 if he is willing to invest in firm #2 in the relative
proportion V1 / V2 . The reason that this "loss of freedom" does not have a negative effect on the
combined firm's value is the CAPM is that in that model, it is optimal for all investors to hold firm #1 and
firm #2 in the relative proportions V1 / V 2 which is exactly the proportion provided by the combined
firm #3.
Note that this "negative" aspect of firm diversification applies even in a "frictionless"
world of no transactions costs and where the merger takes place on terms where no premium
above market value is paid for the acquired firm by the acquiring firm. In the real world, the
acquiring firm must usually pay a premium above the market value to acquire a firm. The
premium can range from 5 to more than 100 percent with an average somewhere around 20
percent. A natural question to ask is "Why do the owners of the firm to be acquired demand a
premium for their shares?" While there are several possible explanations, one that is consistent
with our previous analyses is as follows: If the acquiring firm's management is behaving
optimally, then the reason for their making a takeover attempt must be one of the three reasons
discussed at the outset of this section. Since anyone of these three reasons will increase the value
of the acquiring firm's shares, the acquired firm's shareholders are demanding compensation for
providing the means for this increase in value. How this potential increase in value is shared
between the acquiring and acquired firms' shareholders cannot be determined in general (as is the
usual case for bilateral bargaining), but almost certainly, the acquired firm's shareholders will
demand some positive share. Of course, the acquired firm's shareholders do not know what the
acquiring firm's management believes the value of the acquired firm is. Hence, it might appear
that no consolidation could be consummated because whatever price is offered, clearly, the
acquiring firm's management believes it is worth more, and therefore, the acquired firm's
shareholders should demand more. However, the fact that the acquiring firm believes it is worth
more does not mean that it is, indeed, worth more. I.e., their beliefs may be wrong. Hence, at a

293
 Robert C. Merton

high enough price above market, the acquired firm's shareholders will take the "sure" premium,
and let the acquiring firm take the risk (and earn the possible reward) that its information is
sufficiently superior to the market's that the acquired firm is still a "bargain."
Whether or not the acquired firm's shareholders or the acquiring firm's shareholders come
out ahead on these takeovers is still an open empirical question. However, it is clear that
acquiring another firm for the sole purpose of diversification is a losing proposition for the
acquiring firm because it must pay a premium for a firm whose acquisition promises no increase
in market value even if it is purchased at market.
While the premium paid over market for the acquiring firm is usually the principal cost of
an acquisition, there are other costs as well which can frequently be substantial. In an
uncontested merger, there are legal costs and management's time which could be spent on other
activities. There are uncertainties created for the acquired firm's management, suppliers, and
customers which could affect the operations of that firm during the negotiations and subsequent
transition. Of course, if the merger is contested, then litigation costs will be substantial.
Even if it is decided that firm diversification is warranted, then achieving this
diversification through acquisition is very costly. If, because of management risk aversion or
debt capacity or supplier concerns, it is decided that the volatility or total risk of the firm should
be reduced, then this can be achieved much more efficiently (i.e., at lower cost) by simply
purchasing a portfolio of equities and fixed-income securities where no premium must be paid
over market and no significant transactions costs must be paid. If diversification is desired to
provide "cash flow" from these operations to fund growth investments in current operations, then
it is almost certainly less costly to issue securities and raise the funds in the capital markets.
Don't pay $12 to $20 to acquire $10 in cash!
If it is costly for your shareholders to diversify their portfolios by direct purchase of
individual firms' shares, then this service can be provided at less cost by mutual funds,
investment companies, and other financial intermediaries. In summary, there are three types of
reasons for a firm to consider the acquisition of another firm:

294
Finance Theory

1) Synergy
2) Taxes
3) The firm to be acquired is a "Bargain"

They all have in common that the acquisition should increase the value of the acquiring firm's
current stockholders' wealth.

The possibility of a takeover of one firm by another is an important "check" which serves to force
managements to pursue policies which are (at least approximately) value-maximizing.

Diversification by the firm is, in general, not an important objective for the management of the
firm. Hence, if pursued, then a minimum of resources should be used to achieve it. Specifically,
the acquisition of another firm is a costly way to achieve diversification.

Warning: "diversification" is frequently given as the reason for acquiring a firm by the
acquiring firm's management. If carefully investigated, (most of the time) the meaning of
"diversification" as used is not the one described here, and the real reasons will be one or more of
the three (proper) reasons for making an acquisition.

295
XVI. THE FINANCING DECISION BY FIRMS: IMPACT OF DIVIDEND POLICY
ON VALUE

In Section IX, the choice of capital structure part of the firm's financing decision was
examined to determine if this choice has a significant effect on the market value of the firm. In a
parallel fashion, we examine here the impact of dividend policy on the market value of the firm.
That is, we address the question, "Does dividend policy ‘matter’?" As with the analysis of the
capital structure issue in Section IX, this question is well posed only if it is qualified to reflect
what are the "givens" of the environment. As in Section IX, we ask this question in the context
of a given or prespecified investment policy. That is, given that the firm has already set its
investment plan in real assets, can alternative choices among dividend policies change the market
value of the firm? In this framework, (as will be shown using the basic cash flow accounting
identity), asking the question in this context is equivalent to asking whether or not it matters that
the firm finances its investments by internally-generated funds or by raising the necessary money
externally in the capital markets (or through financial intermediaries).
Using the notation of Sections VI and XIV, an investment policy or plan corresponds to a
specific set of cash flows over time, {X(t)}, and investments over time {I(t)}. From the valuation
formulas (XIV.20) or (XIV.21), a seemingly obvious answer is that "of course, dividend policy
affects the value of the firm." From the valuation formula (XIV.32b), however, an equally
obvious answer is that "given that the distribution for {X(t)} and {I(t)} is fixed, V(0) cannot
change by changing the payout stream, and hence, dividend policy does not affect the value of
the firm." In fact, neither answer is universally correct. Thus, the second answer is correct
provided that the cost of capital, {k(t)}, does not depend on dividend policy. But, it remains to
be determined under what conditions this lack of dependence will obtain.
Before exploring this issue, we briefly digress to list some factors which appear to
influence dividend policy:

296
Finance Theory

Factors influencing dividend policy:

(1) legal restrictions (6) profit rates

(2) cash position (7) access to capital markets

(tradeability of equity)

(3) need to repay debt (8) control of the firm

(4) restrictions in debt contracts (9) tax position of shareholders

(5) rate of asset expansion (10) corporate tax liabilities

Observed stability of dividend policy with respect to earnings or cash flows.

Modigliani-Miller Theorem on Dividend Policy

First proof that "dividends do not matter"

Assume an environment in which short sales are allowed with full use of the proceeds. Suppose
there are two firms with identical investment policies, i.e.,
~ ~ ~ ~
X1 (t) ≡ X 2 (t) and I 1 (t) ≡ I 2 (t)

Suppose that the dividend policies of the two firms for time t > T are identical, but their
dividend policies differ from t ≤ T . Suppose that their values today are different. By
convention, V 2 (0) > V1 (0) . For simplicity, assume that n1(0) = n2(0) . Where n1 and n2 are the
number of shares issued by the two firms.

Consider the following portfolio strategy:

At time zero, buy λ% of firm #1 and sell short λ% of firm #2. Since V2(0) > V1(0), my total

297
 Robert C. Merton

position is at this point:

(a) cash = λ[V2(0) – V1(0)] > 0

(b) long λn1(0) shares of firm #1

(c) short λn2(0) shares of firm #2

Suppose that the portfolio policy is pursued of always maintaining a long position in firm #1
equal to λ% of its value and a short position in firm #2 equal to λ% of its value.

Let N1(t) = number of shares of firm #1 which you are long at time t.
Let N2(t) = number of shares of firm #2 which you are short at time t.

Then N1(0) = λn1(0) and N2(0) = λn2(0) and

(XVI.1a) N 1(t + 1) = N 1(t ) + λ m1(t + 1), and

(XVI.1b) N 2(t + 1) = N 2(t ) + λ m 2(t + 1)

Where m1 and m2 are the changes in n1 and n2 .

Let C(t) = total cash flow from this portfolio strategy at time t. Then:

C (0) = λ n 2(0) S 2(0) - λ n1(0) S 1(0)

= λ[V 2(0) - V 1(0)] > 0

Where S1 and S2 are the share prices for firms #1 and #2.

Assume that C(0) is invested in riskless-in-terms-of-default, T-period discount bonds with yield
to maturity of R(T). For t > 0 and t < T–1 , we have that

(XVI.2) C (t + 1) = N 1(t )d 1(t + 1) - N 2(t )d 2(t + 1) + λ m 2(t + 1) S 2(t + 1) - λ m1(t + 1) S 1(t + 1)

Where d1 and d 2 are the dividends per share.

298
Finance Theory

From the strategy design described in (XVI.1), we have that:

(XVI.3a) N 1(t )d 1(t + 1) = λ D1(t + 1)

(XVI.3b) N 2(t )d 2(t + 1) = λ D 2(t + 1)
Where D1 and D2 are the total dividends paid by the two firms respectively.

Substituting from (XVI.3) to (XVI.2), we have that:

C (t + 1) = λ D1(t + 1) - λ D 2(t + 1)+λ m 2(t + 1) S 2(t + 1)-λ m1(t + 1) S 1(t + 1)

(XVI.4)
= λ{[ D1(t + 1) - m1(t + 1) S 1(t + 1)]-[ D 2(t + 1) - m 2(t + 1) S 2(t + 1)]}

From the cash flow accounting identity (VI.12), we have that:

(XVI.5a) D1(t + 1) - m1(t + 1) S 1(t + 1) ≡ X 1(t + 1) - I 1(t + 1) ≡ Y 1(t + 1)

(XVI.5b) D 2 (t + 1) - m2 (t + 1) S2 (t + 1) ≡ X 2 (t + 1) - I2 (t + 1) ≡ Y 2 (t + 1) ,

and by hypothesis of a fixed investment policy, Y 1(t + 1) ≡ Y 2(t + 1) for all t. Therefore,
substituting into (XVI.4), we have that:

(XVI.6) C (t + 1) = 0 for 0 < t < T -1

If the positions are liquidated at time T, then we have that:

(XVI.7) C (T ) = λ{[ D1(T ) - m1(T ) S 1(T )] - [ D 2(T ) - m 2(T ) S 2(T )]}

T
+λV1(T) -λV2(T) + C(0)[1+R(T)]
sale of purchase of cash and interest on
shares long shares short maturity of bonds.

By assumption, after date T, the dividend policies of the two firms are identical. So after the
dividend payments at time T, it must be that the two firms have identical market values, i.e.,
V1(t) = V2(t) for t ≥ T . In particular, V1(T) = V2(T) . From this and (XVI.5), we have that:

299
 Robert C. Merton

C (T ) = C (0)[1 + R(T )]T

(XVI.8)
= λ[V 2(0) - V 1(0)][1 + R(T )]T >0 if V 2(0) > V 1(0)
Therefore, by investing no money at any time during the interim, the investor can earn C(T) at
time T . Therefore, to avoid arbitrage, C(T) ≡ 0 or:

(XVI.9) V 2(0) = V 1(0)

Therefore, the values of the two firms must be equal and dividend policy "does not matter."

Second proof that "dividends do not matter":

Assume that: 1. Imputed Rationality: If, in forming expectations, each individual investor
assumes that every other trader in the market (A) is rational in the sense of preferring more
wealth to less, independent of the form an increment in wealth may take, and (B) imputes
rationality to all other investors. (2) Symmetric Market Rationality (SMR): Market as a whole
satisfies SMR, if every trader is both rational in behavior and imputes rationality to the market.
We do not assume that short sales can be made with the full use of the proceeds. Consider two
firms as in the "first proof." Suppose that at time t = T–1 there is an investor who is considering
buying λ% of firm #2 for $λV2(T–1). Suppose instead he bought λ% of firm #1 and did the
following: at time T, he will receive λD1(T) in dividends. Suppose he sells (ex-dividend)
$λ[D2(T) – D1(T)] of his stock for cash if D2(T) ≥ D1(T), or if D1(T) > D2(T) , then he buys
$λ[D1(T) – D2(T)] of the stock of firm #1. At this point, he will then have ${λD1(T) + λ[D2(T)
– D1(T)]} = $λD2(T) in cash and ${λ[V1(T) – m1(T)S1(T)] + λ[D1(T) – D2(T)]} worth of firm
#1's stock. From (XVI.5), we have that D1(T) – m1(T)S1(T) = D2(T) – m2(T)S2(T). So, –
λ[m1(T)S1(T) – D1(T) + D2(T)] = – λm2(T)S2(T). Therefore, our investor would have:

$ λD2(T) , in cash, and

$ λ[V1(T) – m2(T)S2(T)] = $ λ[V2(T) – m2(T)S2(T)] , in stock,

because V1(T) = V2(T) .

300
Finance Theory

But, this is exactly the amount of cash and stock which he would have had if he bought λ% of
firm #2. If V1(T–1) < V2(T–1), then every investor (who prefers more to less) would be better
off to buy firm #1 instead of firm #2. Hence, unless V1(T–1) = V2(T–1), there will be a
dominance of one of the firms over the other. If one firm dominates the other, who would buy
the dominated firm, or who would hold it? Clearly, no one. Hence, V1(T–1) = V2(T–1) .
Suppose at some date τ, V1(τ) = V2(τ) , then, by the same argument (with "τ" replacing "T"),
we have that V1(τ–1) = V2(τ–1). Proceeding inductively, we have that V1(0) = V2(0). Both
proofs neglect transactions costs and personal taxes. We now explore what effect these might
have.

Dividend Policy & Market Imperfections: It appears that reductions in current dividends per
share (for fixed investment policy) may increase stockholders' wealth.
(i) because substantial underwriting costs are incurred in issuing stock, shareholders should
prefer a reduction in dividends to a stock issue.
(ii) because capital gains are taxed at a lower rate than dividends and only at the time of their
realization through sale.

Informational Content of Dividends

Since the practice is that dividend payments are smoothed to conform to managers'
estimates of average earnings, the announcement of an increase in dividend payments implies
that management has raised its estimate of average future earnings. If unanticipated through
other means, such an announcement would be expected to affect the stock price.

Generally, (i) Managers are reluctant to cut the dividend rate for fear that this would be
interpreted as a sign of poor earning prospects.
(ii) Dividends are increased only when management is reasonably confident that

301
 Robert C. Merton

the increase can be maintained.

⎛ Dividends ⎞
(iii) Payout ratios ⎜⎜ ⎟⎟ fluctuate because dividends are more stable than
⎝ Earnings ⎠

earnings. But, a firm's target payout ratio is normally stable over time.
(iv) Target payout ratios vary widely from company to company. A typical ratio
is .50 - .60 .

Example: The Constant-Growth Case: Growth Stocks

Review Section VI, pp. 6-20 and 6-21.
Consider the constant growth examined there:
We have from (VI.18) that:

(1- δ )π (0)
(XVI.10) V (0) =
r - r *δ

where r δ = rate of growth of earnings

*

and δ = fraction of profits allocated to new investment.

From (XIV.27), we have that:

D(1)
(XVI.11) V (0) =
r-g
where g is the rate of growth of dividends per share. From the accounting identity, D(t) –
m(t)S(t) = X(t) – I(t). If I(t) = δX(t) , then D(t) – m(t)S(t) = [1–δ]X(t). Let δr = fraction of
current earnings retained (i.e., D(t) = [1–δr]X(t). Let δe = the amount of external financing
required expressed as a fraction of current earnings. It follows that [1–δr]X(t) – δeX(t) = [1-
δ]X(t) or δe = δ – δr.

X (1) = π (0), so, D(1) = [1 - δ r ] π (0).

From (XVI.10) and (XVI.11), we have that

302
Finance Theory

(1- δ )π (0) [1- δ r ]π (0)

(XVI.12) = V (0) =
r - r δ
* r - g
or

δ r * (1- δ r )
- δe
r
(XVI.12') g=
(1- δ ) 1- δ

303
 Robert C. Merton

Note: Unless δe = 0 (i.e., no external financing), the rate of growth of dividends, g , is less
than the rate of growth of profits, δ r . Further, even if the firm pays out all of its current earnings in
*

δ (r * - r )
dividends, i.e., δr = 0 , dividends and price per share will grow over time, i.e., g = .
(1- δ )

Example: three firms all with π (0) = $100 and identical investment policies:

Firm I II III
π (0) $100 $100 $100

r .10 .10 .10

r* .20 .20 .20

δ .40 .40 .40

δr .40 0 .20

δe 0 .40 .20
V(0) $3,000 $3,000 $3,000
I(1) $40 $40 $40
n(0) 1,000 1,000 1,000
S(0) $3.00 $3.00 $3.00

Firm I: Finances all its investment internally through retained earnings, i.e., δr = δ = .40 and
δe = 0 .
From (XVI.10),

(1- δ )π (0) (.6)(100) 60

V (0) = = = = $3,000
r -δ r * .10-.4(.2) .02
I (1) = δ X (1) = .4($100) = $40.

304
Finance Theory

Since this firm does no external financing, D(1) = dividends = X(1) – I(1) = $100 – 40 = $60, by
D(1) 60
the accounting identity. Dividends per share, d (1) = = = $.06 per share. We
n(0) 1, 000

have that X (t ) = X (t -1) + r * I (t -1) = π (0)[1 + r *δ ]t -1. Hence,

X (2) = X (1)[1 + r *δ ] = 100(1.08) = $108. Therefore, the value of the firm next period will

be:
X (2)(1- δ ) $108(1-.4)
V (1) = = = $3, 240 . Since no new shares are issued, n(1) = n(0) =
r - rδ .10 -.4 x .2
1,000 shares. So, the price per share will be $3.24. The total rate of return to the stockholder will
be:

d(1) + S(1) - S(0) .06 + 3.24 - 3.00

= = 10% = r .
S(0) 3.00

The rate of growth of dividends, g, from (XVI.12'), will be

(1- δ r ) δ r V (1) - V (0)

g = δr* - e = δ r * .08 = 8% = rate of growth of the firm
(1- δ ) 1- δ V (0)
S (1) - S (0)
= rate of growth of price per share .
S (0)

Firm II: Finances all new investment by issuing new shares and pays out all earnings as
dividends. As has been demonstrated previously, since the investment policy is the same for all
three firms, X(1), X(2), V(0), V(1), and S(0) will be the same for all firms, and they depend on
the profitability of current assets and future investment opportunities which are independent of
dividend policy. Hence, V(1) = $3240 for this firm, but at that point, the firm will not belong
completely to the shares outstanding at time zero. Namely, it must issue m(1) new shares at
price S(1) to finance investment I(1). I.e., m(1)S(1) = I(1) = $40, V(1) = n(0)S(1) + m(1)S(1) or
V (1) - m(1) S (1) 3240-40
S (1) = = or S(1) = $3.20 and m(1) = 12.5. The return to the
n(0) 1000

305
 Robert C. Merton

d (1) + S (1) - S (0) .10+3.20-3.00

shareholders is = = 10% = r since D(1) = X(1) = $100
S (0) 3.00
D(1) 100
and d (1) = = = .10 . Note that the larger dividend of Firm II is offset by a smaller capital
n(0) 1000
gain.

The rate of growth of dividends, g, from (XVI.12') is

δ r * (1- δ r ) δ er (.4)(.2)(1-0) .4(.1) S (1) - S (0)

g= - = - = 0.0667 = rate of growth of price per share = .
1- δ 1- δ .6 .6 S (0)

Note: The growth of dividends is smaller than for Firm I.

Firm III: Uses a mix of one-half internal and one-half external financing. Hence, m(1)S(1) =
.5I(1) = $20 and again,

V (1) - m(1) S (1) 3240-20

S (1) = = = $3.22 per share
n(0) 1000
D(1) = X (1) - I (1) + m(1) S (1) = 100 - 40 + 20 = $80 and
D(1) d (1) S (1) - S (0) .08+3.22-3.00
d (1) = = $0.08 and = = 10% = r
n(0) S (0) 3.00
(1- δ r ) δ er (.8)(.8) (.2)(.1)
g = δr* - = - = .07334
(1- δ ) 1- δ .6 .6
S (1) - S (0)
= = rate of growth stock price .
S (0)
d (1)
Note: r = 10% = + g = current dividend yield + growth .
S (0)

On Corporate Earnings and Investor Returns

What is the relationship between total dollar returns to shareholders in a particular period

306
Finance Theory

(i.e., dividends plus capital gains) and total dollar earnings of the firm, X(t)? If G(t) = capital
gains to shareholders between period t – 1 and t, then D(t) + G(t) = (1–δr)X(t) + gV(t–1) ,
because (1–δr)X(t) is the amount of earnings not retained, and g, the rate of growth of
(1- δ ) X (t )
dividends, is equal to the rate of growth of price per share. We have that V (t -1) =
r - r *δ
δ r * (1- δ r )
- δ e . Hence,
r
and from (XVI.12'), that g =
(1- δ ) (1- δ )

δ r *(1- δ r )(1- δ ) δ er (1- δ ) }

D(t ) + G (t ) = X (t ){(1- δ r ) + -
(r - r *δ )(1- δ ) (r - r *δ )(1- δ )
X (t )
= {r (1- δ r ) - δ r * (1- δ r ) + δ r * (1- δ r ) - δ er }
r-r δ *

rX (t ) rX (t )
= {1- δ r - δ e} = [1- δ ] since δ = δ r + δ e .
r - r *δ r - r *δ

So,

D(t ) + G (t ) r (1- δ )
= = 1 for r * = r
X (t ) r - r *δ
> 1 for r * > r for 0 < δ <1 for 0 < δ < 1
< 1 for r * < r

X (1) π (0) r - r *δ D(1)

Note: From (XVI.10), = = , and from (XVI.11), = r - g.
V (0) V (0) 1- δ V (0)

So, in general, neither the earnings-to-price nor the dividends-to-price ratio is an unbiased
estimate of the cost of capital, r.

Does dividend policy "matter"? Empirical Evidence Graham & Dodd (early work)

As the result of a cross-sectional fit of companies, they found the following relationship:

307
 Robert C. Merton

E
P = m[ D + ]
3
where
E = earnings; D = dividends; change in retained earnings = ∆RE;
P = price of stock; m = constant.

Because E = D + ∆RE , we also have

m
P = [4 D + ∆RE ] .
3

The weighted average is important and the dividends have a large weight. Implied policy: make
the dividend as large as possible. The equation was "derived" by looking at the data (although it
did not do well for growth stocks, e.g., IBM). Regression or "fit" was done as follows:

Implication: Other things equal, the higher the payout ratio, the higher the price.
Is there any problem with this analysis?

308
Finance Theory

Suppose: P = price is a function of future earnings and managements choose dividends as a

function of future earnings. Does it follow that because P plotted against D gives a good fit,
one can raise price by increasing the dividend payout if the anticipated future earnings stream
remains the same? I.e., is the Graham-Dodd result a causal relationship?

Suppose: the price-earnings ratio properly computed, using long-run "smooth" earnings (i.e.,
π (0) and the "target" payout ratio, D /π (0) , are independent of each other). At a point in time, some
firms' earnings will be transitorily lower than their long-run average. Realizing the transitory
nature of the lower earnings, management does not "cut" the dividend which is based on "long-
run" earnings trend. Hence, D / E > D /π (0) . Similarly, the market, recognizing that price is

dependent on "long-run" earnings, will not bid down the price. Hence, P / E > P /π (0). At the same

time, some firms' earnings will be transitorily higher than their long-run average. For the same
reasons, management does not raise the dividend nor does the market bid up the price. Hence
D / E < D /π (0) and P / E < π (0) .

In a cross-section, the strong positive fit between D/E and P/E could merely reflect
transitory earnings coupled with managements having a target payout based on long-run
"smoothed" earnings.
Because of their concern over the information effect of dividends, management may well
"smooth" dividend payments to match their long-run expectations about the earnings of the firm.

Suppose: (as seems to be the case empirically), that dividend payout policy and the risk (as, for
example, measured by beta) of a firm's underlying assets are not independent. I.e., that high (or
low) dividend payout policies are not randomly distributed across firms. Moreover (as seems to
be the case), suppose that low-risk firms tend to also have high payout policies. Then in a cross-
section of firms, one would expect to find that high-payout ratios would be associated with high
price-earnings ratios. Yet, such a finding does not imply that a firm can raise its PE ratio by
increasing its payout ratio if it maintains the same risk level for its assets.

309
 Robert C. Merton

Black-Scholes Dividend Paper

As was discussed in the beginning of this section, the only way that dividend policy can
affect the value of the firm (given, a fixed investment policy) is if alternative choices for dividend
policy affects the required expected return on the firm (i.e., the {k% (t )}) . In their dividend paper,
Black and Scholes provide a test of the hypothesis that alternative dividend policies differentially affect
required expected returns. To overcome the inherent difficulties with simple cross-sectional analysis,
their test is a combined time-series and cross-sectional analysis. Moreover, their test attempts to correct
for the different risks inherent in a cross-section of stocks. In constructing the test procedure, they begin
with a (generalized) Capital Asset Pricing Model specification for expected returns on securities:

(XVI.13) E ( Z j ) = γ 0 + [E ( Z M ) - γ 0]β j + γ 1[δ j - δ M ] / δ M

where δj = current dividend yield on security j; δM = current dividend yield on the market; γ0 =
expected return on a "zero-beta" portfolio; and γ1 is the "expected return" on the dividend factor.

Possibilities:
(i) The classical security market line relationship of the CAPM would predict γ0 =
R, the riskless rate γ1 = 0 .
Thus, if they could not reject γ0 = R and γ1 = 0 , we cannot reject the CAPM and
we cannot reject the hypothesis that dividend policy "does not matter."
(ii) If γ1 ≠ 0 , then the data suggest that dividend policy does differentially affect
returns. Further, if γ1 > 0 , then this would imply that investors prefer low-
dividend yielding stocks. If γ1 < 0 , then this would imply that investors prefer
high-dividend yielding stocks.

Their findings were that while they could reject the hypothesis that γ0 = R, they could not reject
the hypothesis that γ1 = 0 .

310
Finance Theory

Their results seem somewhat surprising in the light of our proof that dividend policy does
not matter in the absence of transactions costs and personal taxes. Since both exist in the real
world, one's prior might be that γ1 > 0 . I.e., investors prefer low-dividend yielding stocks.
The Black-Scholes explanation of this result is as follows: because payout policies are
not randomly distributed across the firms and risk classes, to achieve dividend yields that are
significantly different from the market's, the investor must hold a less-than-well-diversified
portfolio. Thus, to achieve a higher (or lower) dividend-yielding portfolio, one must pay a price
in the form of increased variance. Because dividend-yield is only a small fraction of the total
return on the market and the maximum tax-saving is even smaller, it does not pay to adjust one's
portfolio to avoid dividends. Moreover, unless a taxpayer is in the maximum tax bracket, he
does not know if he would prefer high or low-dividend paying portfolios unless he knows the
"spread" between pre-tax yields. Hence, they conclude that for stock portfolios (in the world as it
is) investors neglect tax differentials between dividends and capital gains.

311
XVII. SECURITY PRICING AND SECURITY ANALYSIS IN AN EFFICIENT
MARKET

Consider the following somewhat simplified description of a typical analyst-investor's

actions in making an investment decision. First, he collects the information or "facts" (both
fundamental and technical) about the company and related matters which may affect the
company. Second, he analyzes this information in such a way so as to determine his best
estimate (as of today, time "zero") of the stock price at a future date (time "one"). This best
estimate is the expected stock price at time one which we denote by P (1) . From looking at the
P(1)
current stock price, P(0), he can estimate an expected return on the stock, Z , which is Z = .
P(0)

However, his analyst's job is not finished. Because he recognizes that his information is not
perfect (i.e., subject to error, unforeseen events which may occur, etc.), he must also give
consideration to the range of possible future prices. In particular, he must estimate how
dispersed this range is about his best estimate and how likely is a deviation of a certain size from
this estimate. This analysis then gives him an estimate of the deviations of the rate of return from
the expected rate and the likelihood of such deviations. Obviously, the better his information, the
smaller will be the dispersion and the less risky the investment.
Third, armed with his estimates of the expected rate of return and the dispersion, he must
make an investment decision and determine how much of the stock to buy or sell. How much
will depend on how good the risk-return tradeoff on this stock is in comparison with alternative
investments available and on how much money he has to invest (either personally or as a
fiduciary). The higher the expected return and the more money he has (or controls), the more of
the stock he will want to buy. The larger the dispersion (i.e., the less accurate the information
that he has), the smaller the position he will take in the stock.
To see how the current market price of the stock is determined, we look at the
aggregation of all analysts' estimates, and assume that on the average the market is in
equilibrium. I.e., on average, the price will be such that total (desired) demand equals total
supply. Analysts' estimates may differ for two reasons: (1) they may have access to different

312
Finance Theory

amounts of information (although presumably public information is available to all); (2) they
may analyze the information differently with regard to its impact on future stock prices.
Nonetheless, each analyst comes to a decision as to how much to buy or sell at a given market
price, P(0). The aggregation of these decisions gives us the total demand for shares of the
company at the price, P(0). Suppose that the price were such that there were more shares
demanded than supplied (i.e., it is too low), then one would expect the price to rise, and vice
versa, if there were more shares available at a given price than were demanded. Hence, the
market price of the stock will reflect a weighted average of the opinions of all analysts. The key
question is: what is the nature of this weighting? Because "votes" in the marketplace are cast
with dollars, the analysts with the biggest impact will be the ones who control the larger amounts
of money, and among these, the ones who have the strongest "opinions" about the stock will be
the most important. Note: the ones with the strongest "opinions" have them because (they
believe that) they have better information (resulting in a smaller dispersion around their best
estimate). Further, because an analyst who consistently overestimates the accuracy of his
estimates will eventually lose his customers, one would expect that among the analysts who
control large sums, the ones that believe that they have better information, on average, probably
do.
From all this, we conclude that the market price of the stock will reflect the weighted
average of analysts' opinions with heavier weights on the opinions of those analysts with control
of more than the average amount of money and with better than average amounts of information.
Hence, the estimate of "fair" or "intrinsic" value provided by the market price will be more
accurate than the estimate obtained from an average analyst.
Now, suppose that you are an analyst and you find a stock whose market price is low
enough that you consider it a "bargain" (if you never find this situation, then there is no point
being in the analyst business). From the above discussion, there are two possibilities: (1) you do
have a bargain─your estimate is more accurate than the market's. I.e., you have either better than
average information about future events which may affect stock price and/or you do a better than
average job of analyzing information. Or, (2) others have better information than you do or

313
 Robert C. Merton

process available information better, and your "bargain" is not a bargain.

One's assessment of which it is, depends on how good the other analysts are relative to
oneself. There are important reasons why one would expect the quality of analysts to be high:
(1) the enormous rewards to anyone who can consistently beat the average attract large numbers
of intelligent people to the business; (2) the relative ease of entry into the (analyst) business
implies that competition will force the analysts to get better information and better techniques for
processing this information just to survive; (3) the stock market has been around long enough for
these competitive forces to take effect. Unfortunately, the tendency is to underestimate the
capabilities of other analysts. Ask any analyst if he is better than average, and invariably he
answers "yes." Clearly, this cannot be true for all analysts by the very definition of average. If
the analysts are so good, why aren't most of them rich? Precisely because they compete with
each other, the market price becomes a better and better estimate of "fair value," and it becomes
more difficult to find profit opportunities. To stay ahead, the analyst must develop new ideas
continually. As the limiting case of this process, one would expect that as market prices become
better estimates of "fair value" in the sense of fully reflecting all relevant known information, the
fluctuations of stock prices around the expected "fair return" will be solely the result of
unanticipated events and new information. Hence, these fluctuations are random and not
forecastable. And it is in this sense that the fluctuations in stock prices can be described by a
random walk.
This also explains why the performance of most "managed" portfolios will be no better
than the performance of an "unmanaged" well-diversified portfolio. In fact, the "unmanaged"
portfolio, because it takes market prices as the best estimate of value, is equivalent to a
"managed" portfolio whose manager is a no-worse-than-average analyst! The investor who buys
such a portfolio is simply "piggy-backing" on the actions taken by active analyst-investors
competing with each other.
This is essentially the story behind the "Random Walk Theory." It does not imply that a
better-than-average analyst cannot make greater than fair returns. It does not imply that all
analysts should quit their jobs, and in fact, its cornerstone is that enough analysts remain and

314
Finance Theory

actively compete so that market prices are good estimates of "fair" value. It is only in this way
that the "piggy-backing" by investors can be justified. Further, it does not imply that all investors
should hold "unmanaged" portfolios. If an investor can identify an analyst with above-average
capabilities and is willing to bear the risk of his capabilities, then a "bargain" can be struck so
that both are rewarded for the effort. The theory does imply that to make "extra" profits, one
must have superior techniques which process information in a way not generally known in the
market and that the longer that the market is in existence, the greater the number of participants,
the more difficult it is to make these "extra" profits.

An Example to Illustrate the Efficient Market Concept

Consider a firm in a cyclical business whose earnings are completely predictable but vary
in the following fashion: If the earnings per share this period are $50, then next period's earnings
per share will be $100, and if the earnings per share this period are $100, then next period's
earnings per share will be $50. I.e., if Et denotes earnings in period t , and if
E0 = $50 then E1 = 100, E 2 = 50, E3 = 100,... or

t
E t +1 = E t + (-1) 50

If the firm pays out all earnings as dividends (Dt = E t ) and if the required return ("fair market

return") is 20% per period, then the correct price per share, St (ex-dividend) is given by
S0 = $386.36, S1 = $363.64, S2 = $386.36, S3 = $363.64,...

t +1
St +1 = St + (-1) 22.72

I.e., the return per dollar from investing in the shares from time 0 to time 1,
D1 + S1 100 + 363.64
Z1 = = = 1.20, and from time 1 to time 2,
S0 386.36

D2 + S2 50 + 386.36
Z2 = = = 1.20, and so forth.
S1 363.4

315
 Robert C. Merton

Suppose that investors are myopic and assume that current earnings (and hence, current
dividends) are permanent. I.e., their best guess of future dividends is that they will be equal to
'
current dividends. If St denotes price per share under this belief, then

D0 50 D 100
S0' = = = $250; S1' = 1 = = $500,
r .2 r .2
D2 50 D3 100
S 2' = = = $250; S3' = = = $500,...
r .2 r .2
St' +1 = St + (-1) 250
' t

or

The return per dollar from investing in the shares from time 0 to time 1 under this pricing is

D1 + S1 100 + 500
'
Z1' = '
= = 2.4 or 140%
S0 250
'
and from time 1 to time 2 , Z 2 is

D 2 + S2 50 + 250
'
Z 2' = '
= = 0.6 or - 40%
S1 500

and it continues to alternate.

316
Finance Theory

Empirical Studies of Capital Market Theory

In Sections IX and X, we developed a theory for the capital markets based on essentially
rational behavior and optimal portfolio selection. Specifically, by applying the mean-variance
model and aggregating demands, we deduced the Capital Asset Pricing model, which provided a
specification for equilibrium expected returns among securities. Based on this model, we
deduced a naive or benchmark portfolio strategy. From our analysis of an efficient speculative
market, we deduced a rationale for random selection of securities or the naive strategy as possible
portfolio strategies. Since these models have important implications for both corporate finance
and financial intermediation, it is most important that empirical testing of the models is
performed. Basically, there are three questions to be answered: (i) How does the "random walk"
theory hold up against the data? (ii) Is the security market line specification a reasonable
description of returns on securities? (iii) How does the performance of the naive strategy
compare with managed portfolio strategies?

317
 Robert C. Merton

The answer to (i) is simply that a large number of technical trading strategies (filtering,
serial correlation, charting services, volume analysis, etc.) have produced no evidence to refute
the random walk hypothesis. To the extent that any serial correlation in the returns were present,
it was of such small magnitude and "short-lived" nature that no profitable trading was possible.
Other studies of brokerage house and general service recommendations, dividend announcements
and earning reports have shown no evidence of providing trading profits. "Dart throwing" or
more careful random selection of portfolios provide no evidence against the random walk
hypothesis.
In the study of managed portfolio performance, both the random walk hypothesis and the
asset pricing model are implicitly tested.

Returns on the "Market"

NYSE index: value-weighted index of all stocks on the New York Stock Exchange
≈80% in market value of all securities)

S&P index: Standard & Poors 500-stock index including the largest companies (in
1965 representing ≈80% of market value of NYSE stocks)

Random Selection of Stocks (Fisher & Lorie): Equally-weighted portfolio of all stocks on
the New York Stock Exchange 1926-1965.

Average Return (1-year): including dividends, no taxes, or commissions

Average Annual Return Standard Deviation

Years (Arithmetic Average) (Annual)

1926-1945 17.8% 41.2%

1946-1965 15.1% 19.8%

1926-1965 16.5% 32.3%

318
Finance Theory

"Market" (in this sense) was much more volatile in the pre-war versus post-war period.

Average Compound Return: including dividends, no taxes, but including purchase commissions:

Average Compound Return

Years (Geometric Average)

1926-1945 6.3%
1946-1965 12.6%
1926-1965 9.3%

All Stocks on the New York Stock Exchange: Value-Weighted

Cowles (1871-1937): Average Compound Return: 6.6%

Since the Fisher Lorie results for average performance of randomly selected portfolios is as good
as managed portfolios on average over the same period, this is additional evidence in favor of the
Random Walk.

319
 Simulated Rate of Return Experience for Successful Market Timing*
Monthy Forecasts: P = Probability of Correct Forecast
January 1927 – December 1978

Market Timing NYSE

Per Month P=1.0 P=.90 P=.75 P=.60 P=.50 Stocks

Average Rate of Return 2.58% 2.17% 1.56% 0.94% 0.53% 0.85%

Standard Deviation 3.82% 3.98% 4.13% 4.19% 4.18% 5.89%

Highest Return 38.55% 38.27% 37.61% 36.41% 35.12% 38.55%

Lowest Return -0.06% -17.05% -22.02% -24.52% -25.64% -29.12%

Average Compound Return 2.51% 2.10% 1.47% 0.85% 0.44% 0.68%

Growth of $1,000 $5,362,212,000 $418,902,144 $9,146,722 $199,718 $15,602 $67,527

Average Annual
Compound Return 34.65% 28.32% 19.14% 10.69% 5.41% 8.47%

*Buy the market when the forecast is for stocks to do better than bonds.
Buy bonds when the forecast is for bonds to do better than stocks.

320
 Robert C. Merton

Average Annual Compound Return on the Market (value-weighted, including reinvesting

dividends, no commissions, or taxes) (Scholes)

NYSE S&P
Average Avg. Excess Average Avg. Excess
Years Return Return Return Return

Total: 1953-1972 11.98% 7.57% 11.63% 7.22%

10 Years: 1953-1962 13.11% 10.00% 13.38% 10.25%
1963-1972 10.86% 5.15% 9.90% 4.19%
5 Years: 1953-1957 12.25% 9.45% 13.50% 10.70%
1958-1962 13.99% 10.52% 13.26% 9.79%
1963-1967 14.53% 9.70% 12.34% 7.50%
1968-1972 7.31% 0.71% 7.52% 0.92%

Jensen Performance of Mutual Funds Study 1945-1964

Testing 115 Funds ability to Forecast (relative to Security Market Line):

Model Specification: Z j (t) = R(t) + β j[ ZM (t) - R(t)] + α j (t)

~ (t) = R(t) + β [ ~ (t) - R(t)] + (t) + ~ (t)

Test: Zj j ZM αj εj

E(~ε j (t)) = 0; Cov(~ε j (t), ~ε i (t - k)) = 0 i, j = 1,

k = 1,2,...

Assumes: β j is stationary. Suppose not:

~ ~
β j (t) = β j + U j (t)

~ ,~ ) > 0 estimated
if you can forecast the market, then Cov ( U j ZM which would imply E(β j ) < βj

and biases tests in favor of superior performance (i.e., larger α j ).

321
 Robert C. Merton

115 Funds Studied. Returns net of all costs including management fees.

76 funds had measured α j < 0

Average α = – .011
= – 1.1%
39 funds had measured α j ≥ 0

The statistical significance of the positive α j were no more than would have been expected by

chance when the true α j = average α .

Using Returns Gross of Management Fees

55 funds had measured α j < 0
Average α = – .004
= –0.4%
60 funds had measured α j ≥ 0

Statistical significance of the positive α j were no more than would have been expected by

change when the true α j = 0.

Conclusions: Funds taken as a whole do not show evidence of superior forecasting capability;
and, of course, do not show evidence of sufficient superior forecasting to cover costs.

What about individual funds? Even if funds as a whole do not show evidence of superior
forecasting, what about the overtime performance of particular funds? Is it true that funds with
observed positive α j in the past tend to have positive α j in the future? Jensen & Black studied

the 115 funds for the years 1955-1964 computing the realized α j for each year (a total of 10 ×

115 = 1150 observations). The differential returns were computed gross of management fees.
The results were

322
Finance Theory

Number of Successive Percent of Cases

Years of Observed Number of Times Followed by Another
Positive "α" Observed Positive "α"

1 574 50.4%
2 312 52.0%
3 161 53.4%
4 79 55.8%

Conclusion: It appears that funds that did well in the past show little evidence of continuing to
do so.
Jensen also found that there was no significant evidence of serial correlation in the return
series in support of the Random Walk Hypothesis.
With respect to providing efficient (or well-diversified) portfolios, on average, Jensen
found that 85% of the variance of the funds' returns were due to market movements. I.e.,

1
σp ≈ (1.085) σM βp = 1.085 ρpM σp or ρpM ≈ = .9216
1.085

Further, on the whole, funds tended to keep about the same level of βp or σp through time.

Overall Summary

1. Over the last forty years, randomly selected portfolios have returns greater than or equal
to randomly selected managed portfolios.
2. Most mutual funds are reasonably well diversified (i.e., have reasonably low non-
systematic risk).
3. On average, funds did not perform, before expenses, any better than a naive strategy
portfolio with the same beta.

323
 Robert C. Merton

4. On average, funds did worse, after expenses, than the naive strategy portfolio with the
same beta.
5. Few, if any, individual funds showed any consistent performance superior to the naive
strategy over time.
6. Most funds spend too much money trying to forecast returns on stocks: either explicitly
in analyst salaries and support and implicitly through brokerage commissions and spreads
through excess turnover.

Investment prescription: Since these results did not include sales commissions on "load" funds
which run from 1½ - 8½%, clearly one should buy "no load" funds (with no sales commissions).
To achieve an efficient investment strategy, choose a mix of a few well-diversified, no load
funds. Select funds with the lowest costs (management fees and turnover).

(ii) Testing the Capital Asset Pricing Model

(Miller and Scholes; Black-Jensen-Scholes)
The capital asset pricing model specifies that

E( Z j) = R + β j [E( ZM ) - R] and E( ZM) > R.

I.e., investors are risk-averse; expected excess return on a security is proportional to its beta; it is
dependent only on beta; is linear in beta.
The Black-Jensen-Scholes paper is one of the most sophisticated tests of the capital asset
pricing model. Using monthly returns from 1931-1965 on 600-1100 securities, they found the
following:

1. The expected return on the market is greater than the riskless rate ( ZM > R).
2. Expected return on individual securities (portfolios) is an increasing function of its beta
and the excess returns are linear in beta.
3. Expected return depends on beta.

324
Finance Theory

4. The empirical Security Market Line is too "flat." I.e., the returns on "low beta" (β < 1)
stocks were higher than predicted by the Capital Asset Pricing Model and the returns on
"high beta" (β > 1) stocks were lower than predicted by the Capital Asset Pricing Model.

Results 1-3 are consistent with the capital asset pricing model, result 4 is not, and has
been the cause for much concern as well as new research in this area. To analyze this problem,
BJS constructed a "zero-beta" portfolio by combining stocks only (so it has variance), and this
portfolio had realized returns significantly greater than the riskless rate. I.e., Z0-β > R where Z0-β

is the expected return on the minimum-variance, zero-beta portfolio constructed from stocks.
The specification that they fit was Z j - R = β j( Z M - R) + γ(β j)( Z 0-β - R) ,

dγ
where γ(1) = 0 and < 0.
dβ

While there are many possible theoretical and empirical explanations for this finding, such
analyses are beyond the level of this course. It is evident that the simple form of the Capital
Asset Pricing Model as a means for estimating expected returns on individual securities is not
sufficient; however, the main results implied by that model (1-3) do seem to describe returns and,
as a good approximation, its specification is not unreasonable.

325