Robert C. Merton
Table of Contents
I. II. III.
Introduction ............................................................................................................. On the Arithmetic of Compound Interest: The Time Value of Money................ On the Theory of Accumulation and Intertemporal Consumption Choice by Households in an Environment of Certainty ............................................... On the Role of Business Firms, Financial Instruments and Markets in an Environment of Certainty............................................................................. The "Default-Free" Bond Market and Financial Intermediation in Borrowing and Lending .................................................................................................. The Value of the Firm Under Certainty ................................................................. The Firm's Investment Decision Under Certainty: Capital Budgeting and Ranking of New Investment Projects........................................................... Forward Contracts, Futures Contracts and Options............................................... The Financing Decision by Firms: Impact of Capital Structure Choice on Value............................................................................................................. The Investor's Decision Under Uncertainty: Portfolio Selection ......................... Implications of Portfolio Theory for the Operation of the Capital Markets: The Capital Asset Pricing Model ........................................................................ Risk-Spreading via Financial Intermediation: Life Insurance .............................. Optimal Use of Security Analysis and Investment Management .......................... Theory of Value and Capital Budgeting Under Uncertainty................................. Introduction to Mergers and Acquisitions: Firm Diversification ......................... The Financing Decision by Firms: Impact of Dividend Policy on Value ............ Security Pricing and Security Analysis in an Efficient Market.............................
1 8
34
IV.
57
V.
76 115
VI. VII.
134 151
VIII. IX.
165 185
X. XI.
Copyright 1982 by Robert C. Merton. These Notes are not to be reproduced without the authors written permission. All rights reserved.
I. INTRODUCTION
Output Product Markets Consumption Labor Markets Manufacturing or Business Firms Investment Capital
Households Savings
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 welldeveloped 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.
Households (Personal Finance) Taken as Given: 1. A criterion function for choice among alternative consumption programs 2. Initial endowments
To be Determined:
1. 2.
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. 2.
An operation criterion for measuring good management Investment decision in physical assets (capital budgeting) a. b. Which assets to invest in How much to invest in total
3.
The long-term financing decision a. b. Dividend policy Capital structure decisions and the cost of capital
4.
Robert C. Merton 5. 6. Mergers and Acquisitions: Firm diversification 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. 2. Why they exist and what services they provide How the management of financial intermediaries differs from the management of business firms 3. 4. Efficient management and measurement of performance The role of market makers
Capital Markets and Financial Instruments (Capital Market Finance) To be Determined: 1. 2. 3. Why they exist and what services they provide The characteristics of an "efficient" capital market 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
Finance Theory Basic Methodology and Approach of the Course 1. 2. 3. 4. How should the system work? Does it work that way? If not, is there an opportunity for improvement (and hence, a profit opportunity)? 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.
Robert C. Merton Summary 53-Year Return Experience: Stocks and Bonds (19261978) Source: Stocks, Bonds, Bills, and Inflation: Historical Returns (19261978), R.G. Ibbotson and R.A. Sinquefield, Financial Analysts Foundation (1979).
Average Annual Return 11.2% Standard Deviation 22.2% Growth of $1000 (Average Compound Return) $89,592 (8.9%) $ 7,807 (4.0%) $ 5,342 (3.2%) $ 3,728 (2.5%)
4.1%
5.6%
3.4%
5.7%
2.5%
2.2%
1.6%
NA
0.9%
NA
0.0%
4.6%
Finance Theory
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
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. No Satiation: Individuals always strictly prefer more money to less. 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. 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.
(A.II.2) (A.II.3)
(A.II.4)
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
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. individuals, "deposits." Indeed, for expositional convenience, we will call loans made by
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, redeposits V1 dollars for the
2
rV o ,
or
second
period
V 2 = (1 + r) V 1 = (1 + r)[(1 + r) V o ] = (1 + r ) V o .
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 (II.1)
V n = (1 + r ) V o ,
n
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
where "log" denotes the natural logarithm (i.e., to the base e). Two "rules of thumb" used to approximate n in (II.2) are: (II.3)
* n 72/100r
and (II.4)
* n 0.35 + 69/100r
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 PV n (x) = x/(1 + r ) ,
(II.5)
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 )
(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 (II.7)
PV( x1 , x 2 ,..., x N ) = xt /(1 + r ) .
t t =1 N
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.
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 .
1 + i = eR
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
N
1 2 5 10 15 20 30
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 ,
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, (II.13)
PV(x1 , x2 ,...,xn ) x(t)e - r ctdt,
0 N
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 )
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 .
N -3
Hence, the total amount accumulated, S N , will be the sum of all N terms. I.e., SN =
y(1 + r )
t =1
N -t
N -1
develop
mathematical
N -1
formula.
The
sum
of
geometric
progression,
N -1
x = (x
t t=0
- 1)/(x - 1).
x = x( x
t t =1
- 1)/(x - 1).
20
Finance Theory SN
N is called the compound value of an annuity, and [(1 + r ) - 1]/r
of
an
t
annuity
N
(denoted
t
by
AN )?
From
(II.7),
we
have that
A N = y/(1 + r ) = y 1/(1 + r ) .
t =1 t =1
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
N therefore A N = S N /(1 + r ) . The reader may verify that this is the case by inspection of (II.16).
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 (Nt) period annuity. Hence, the change in the present value of an N-period annuity over one period is equal to
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.
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 Amount % of Total $2,400 2,367 2,252 1,990 1,605 1,039 208 85.4% 84.2 80.1 70.8 57.1 37.0 7.4 Principal Reduction Amount % of Total $ 410 443 558 820 1,205 1,771 2,602 14.6% 15.8 19.9 29.2 42.9 63.0 92.6 Amount of Loan Outstanding $29,590 29,147 27,589 24,052 18,855 11,220 0
Year 1 2 5 10 15 20 25
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
A N - t = y[1 - 1/(1 + r )
N -t
]/r.
mortgage, AN , as
23
Robert C. Merton
(II.20)
N t N 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 (II.21)
N -t AN -t -1 - AN -t = - y/(1 + r ) ,
and the percentage of the total payment used to reduce principal between t and t + 1 written as (II.22)
[ AN -t - AN -t -1 ]/y = 1/(1 + r )N -t .
can be
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 continuouslycompounded 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.
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)
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 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
years.
contribution to the plan takes place one year from now, what is the economic benefit of the taxdeferred 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.
$q = rS N /[1- 1/(1+ r )n ] per year for n years from this accumulated sum. However, he must pay taxes of $q 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)
If, instead, he had chosen an ordinary saving plan, he would have had to pay $y 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 each year. Moreover, the interest accumulation period, he could only contribute $(1 - )y
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 - )r after tax. Again using formula (II.15), his total amount accumulated at retirement from the ordinary saving plan, S 2 ,
N is $(1 - )y[(1 + (1 - )r ) - 1]/(1 - )r.
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 . Thus, from formula (II.16), he can generate an after-tax withdrawal plan of $ q 2 = (1 - ) rS 2 /[1 - 1/(1 + (1 - )r )n ] per year for n years which can be rewritten as (II.27) $ q 2 = (1 - )y[(1 + (1 - )r )N - 1]/[1 - 1/(1 + (1 - )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 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 - )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 - )PT. Hence, the cost of owning the house, PCO, can be written as
PCO = I + [M + (1 - )PT]/(1 + (1 - )r )
t =1
(II.28)
= I + PT/r + M/(1 - )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 - )r )t
t=1
(II.29)
= X/(1 - )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
tax cash flow is (X-M-PT)(1 - ) + D . Discounting these after-tax cash flows at his aftertax interest rate, have that X must satisfy
I = [(X-M-PT)(I - )+ D]/(I - )r or
(II.30)
X = rI + M + PT - D/(1 - ).
From (II.28), (II.29), and (II.30), we have that the cost saving of owning over renting can be written as
28
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.
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 (II.32)
N 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
(II.33)
D N -1 - D N = rM/(1+ r ) = 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
t N I N = C/(1 + r ) + M/(1 + r ) t=1 N
(II.34)
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)
(II.36)
n = 1+ r ( ) V n t V o 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,
(II.37)
(1+ R n ) (1+ r t ),
n t=1
and therefore, 1 + R n is the geometric average of the {1 + r t }, t = 1,2,..., n. rewrite (II.36) as (II.38)
V n = (1 + R n ) V o .
n
Hence, we can
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 (II.39)
n 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 )
(II.40)
t =1
= x /(1+R
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 " . entire path of interest rates from time t = 1 to time The former depends upon the
period rate that obtains between t = n 1 and t = n. For example, from (II.37), we have that
32
Finance Theory
> R n = R n -1 < > r n = R n -1 . <
(II.41)
if and only if
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 .
dC 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.
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
Max U[ W
C1
- PC 1 , C 1 ]
dU dC 1
* *
* * * = 0 = - U 1 [ W 0 - PC* 1 ,C 1 ]P + U 2 [ W 0 - PC 1 ,C 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. I.e.,
Co = yo and C1 = y1 .
*
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:
= y.
* *
If
(C0 , C1 )? From (III.3), we have that U1[C0 ,C1 ]/U2[C0 ,C1 ] = (1+)C1 /C0 = 1+r, or that
the the
budget
constraint, for
into
budget
constraint
39
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, can be interpreted as the consumer's rate of time preference. If > 0, then he has positive time preference. If = 0, then he has no time preference, and if < 0, then he has negative time preference. Note that in that example, if the interest rate exceeds his rate of time preference (r > ), then C0 < y , and he will save some of his current period's income to consume next period. If r < , then
*
C0 > y and he borrows against next period's income to consume more than his current income.
If r = , 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 = . Hence, by this example, we have illustrated one of the possible explanations for a positive rate of interest: namely, consumers' impatience to
*
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
X1
max
= f(y0)
which corresponds to X0 = y0 and C0 = 0. Hence, f(y0) X1 0. His next period's consumption can be written as (III.7)
C 1 = y1 + X 1 = y1 + f( y 0 - C 0 )
41
Robert C. Merton
Figure III.3 Production Function
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
, y 1 + f( y 0 - C 0 )]
42
Finance Theory
* * * 0 = U 1 [ C* ( y 0 - C* 0 , y 1 + f( y 0 - C 0 )] - f 0 )U 2 [ C 0 , y 1 + f( y 0 - C 0 )].
(III.8)
where
*
C1 = y1 + f(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
( dC 1 / dC 0 )U =U * = - f ( X* 0)
* * *
* *
(III.10)
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
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 }
+ 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) and (III.12b) where (X0 ,C0 ,C1 ) denotes the quantities chosen for the optimal investment-consumption program, and
* * *
* * * U/ C 1 = 0 = - PU 1 [ C* 0 ,C1 ] +U 2 [ C0 ,C1 ] ,
* * U/ X 1 = 0 = U 1[ C * 0 ,C 1 ](Pf '( X 0 ) - 1)
45
Robert C. Merton
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,
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,
X0
**
investment which maximizes the current wealth of the consumer, then from (III.11),
**
Max [ y
{ X0}
+ P y 1 + Pf( X 0 ) - X 0 ]
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. Specifically, at each point in time, there are (T+1) different claims traded in units where the
th
48
Finance Theory
such claim gives its owner the right to one unit of the consumption good payable periods from the date at which it is issued, = 0,...,T. In effect, these claims are pure discount loans as defined in Section II. Let Pt() denote the price at date t of a discount loan which pays one unit of the consumption good periods from date t (i.e., at date t + ). 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 (III.15)
W 0 = P0 (
T
=0
) y .
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 feasible, it must satisfy
P ( )C W
0
=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 ( ) C
T
Noting
=0
that P0(0) = 1, we can substitute for C0 from the budget constraint, and rewrite the problem as
{ C 1 ,C 2 ,...,C T }
Max
U[ W 0 - P0 ( )C ,C 1 ,C 2 ,...,C T ]
T
=1
where U U[C0,C1,...,CT]/C-1 denotes the partial derivative of U with respect to its argument and (C 0,C 1,...,C T)
* * *
th
is
the
optimal
consumption
program
with
49
Robert C. Merton
C =W 0 -
* 0
P ( )C .
* 0
In words, (III.16) says that at the optimum, the ratio of the marginal
=1
utility of consumption should just equal the ratio of the marginal cost of consumption in period to the marginal utility of current consumption in period , P0(), to the marginal cost of current consumption, P0(0) = 1. From (III.16), we have that
(III.17)
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,...t1), 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
W 0 = P0 ( ) y + P0 ( ) f ( X 0 , X 1 ,..., X -1, )
T T
(III.18)
=0
=1
T -1
- P0 (
=0
)X
where X
j= +1
plan {ft}, = 0,...,T1. Define the net increment to the consumer's current wealth of production plan {ft}, V0, by (III.19)
V 0 P0 ( ) f ( X 0 , X 1 ,..., X -1, ) - P0 (
T T -1
) X .
=1
=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 that W 0 = P0 ( ) y + V 0 = P0 ( ) C . Substituting for C0 from the budget constraint, the
=0 =0
T T
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 ( =0
) y + V - P ( ) C , C
0 0 =1
,..., C T
(III.20a)
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 pureexchange 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.
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 ( ) C
T
=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
52
Finance Theory
wealth at date t = 1. Note:
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 (
T
(III.21)
+ U [ C ,C ,...,C ] , = 2,...,T,
+1 0 1 * T * * C 1 = W 1 - P1 ( - 1) C .
=2
- 1) C ,C ,...,C ] P ( - 1)
* * 2 * T 1
where
* *
=2
(C1 , C 2 ,...,CT ) will depend upon the amount of wealth {P1 (1),...,P1 (T-1)}, C0 ,
and the form of the utility function U.
W1,
the
prices
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()} 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
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
,( W 0 - C 0 )/ 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/ P 0 (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
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 twoperiod 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, (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 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
t * *
( )
*
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
(IV.1)
PV 0
(IV.2)
However, P 0 ( ) = 1/(1 + R ) . Therefore, from (III.19) and (IV.2), PV0 = V0. Hence, the set
of
choices
for
Xjt
which
maximize
PV0
will
satisfy
60
Finance Theory the necessary funds to purchase the inputs, the founder could actually sell either all or part of his holdings for V_ where 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
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 is the percentage ownership given to the new shareholders, then the value of their holdings as of next period will be f(X0). Therefore, to raise $X0 today, must be chosen so that f(X0) = (1+r)X0 or = (1+r)X0/f(X0). Under this financing method, the original owner's holdings will be worth (1)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
+ +
( )
*
*
Moreover,
by
supposition, V _ X 0
( )
+
V _ X0 .
( )
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-frommanager business firms does not rest upon the delicate and naive assumption of a mutualadmiration 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 (melteddown) 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: (IV.3) and
I I I y0 = 300 ; y 1 = 1500 ; and y 2 = 2700 .
I I I I I I I U ( C 0 , C 1 , C 2 ) = log( C 0 ) + log( C 1 ) + log( C 2 )
Similarly, for person #II, we assume that the preference and "wage income" endowment are given by (IV.4) and
II II II y0 = 2700; y1 = 1500; and y 2 = 300. II II II II II II II U ( C 0 , C 1 , C 2 ) = log( C 0 ) + log( C 1 ) + log( C 2 )
Note that in this example, both people have identical preferences and similar magnitudes of income except the time patterns of their receipt are reversed.
69
Robert C. Merton
In this case, both individuals have no choice but to consume their current income each period because
I I
"no
I
production"
I I
implies
I
"no
storage." U
II
I
Hence, = 9.08.
= 9.08.
Suppose now that there is an exchange market with market interest rates (r1,r2). The current wealth of person j is given by W0 = y 0 + y1 /(1+r1 )+ y 2 /(1+r1 )(1+r2 ), j = I,II. As
j j j j
discussed in Section III, person j will choose a consumption program as follows: Max { log( C 0j ) + log( C 1j ) + log( C 2j )} subject to the constraint that solution is given by
j* j C 0 = W 0 /3 j* j* j 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
(IV.5)
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. investment,
I* II* I
and
aggregate
saving,
must
equal
zero
i.e.,
Ct + Ct
market
clearing
70
Finance Theory
II
j*
II
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.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 (i.e.,
I
f t+1 = X t , t = 0,1). Even though Person #I's technology provides for a 100 percent rate
II
of return per period, his optimal choice is to not use his technology to produce any goods. The 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,
I
71
Robert C. Merton he derives no benefit from his technology, and has the same consumption allocation as in the noproduction 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
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
I
to
the
present
(or
"capitalized")
value
II
of
his
wage
income.
Thus,
W0 = 300 + 1500/2 + 2700/4 = 1725 and W0 = 2700 + 1500/2 + 300/4 = 3525. From
(IV.5),
I*
the
optimal
I*
consumption
I*
program
I
chosen
by
Person for
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 Production Income Operating Income Interest Income (Expense) Net Income Consumption Savings Investment Savings Surplus (Deficit)
Balance Sheet t = 0+
Person #I Assets Capital Bonds Capitalized Wage Income 1,250 0 1,425 2,675 0 1,525 825 2,350 1,250 1,525 2,250 5,025 1,525 0 1,525 Person #II Aggregate Person #I Person #II Aggregate Liabilities Debt
1,150 2,675
2,350 2,350
3,500 5,025
Net Worth
Aggregate (I + II)
Wage Income Production Income Operating Income Interest Income (Expense) Net Income Consumption Savings Investment Savings Surplus (Deficit)
74
Finance Theory
Balance Sheet t = 1+
Person #I Assets Capital Bonds Capitalized Wage Income 2,000 0 1,350 3,350 0 2,200 150 2,350 2,000 2,200 1,500 5,700 2,200 0 220 Person #II Aggregate Person #I Person #II Aggregate Liabilities Debt
1,150 3,350
2,350 2,350
3,500 5,700
Net Worth
Wage Income Production Income Operating Income Interest Income (Expense) Net Income Consumption Savings Investment (Liquidation) Savings Surplus (Deficit)
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() denote the price at date t of a default-free bond which promises a payment of $1 at date (t + ). 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 1 and 2 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(j 1)/P0(j), j = 1,2. Suppose that P1(1 1)/P0(1) < P1(2 1)/P0(2). 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(1) bonds which will be worth $N1P1(11) next period. If instead he invests in the second bond now, then he will purchase N2 = I/P0(2) bond now which will be worth $N2P1(2 1) next period. By hypothesis, N2P1(21) > N1P1(11). 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(21) N1P1(11) 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(1 - 1)/P0(1) > P1(2 - 1)/P0(2), 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. equilibrium, we have that Thus, in
77
for all maturities 1 and 2. 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( - 1)/ P 0( ) = 1 + r 1
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 ( - 1) / Pt ( ) = 1 + r t+1
for all dates t = 0, 1, 2, ... and all maturities . 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 (V.4)
T 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
P0 (T ) = 1/ (1 + RT ) .
T
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 ( - 1)/ Pt -1 ( ) = P0 (t - 1)/ P0 (t) ,
for all future dates t = 1, 2, ... and all maturities . 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., (V.8)
RT = [ P0 (T) ]
-1/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
81
Robert C. Merton
Table V.1 Interest Rates, Yields, and Investment Returns Comparisons Actual Value of $1000 Initial Investment Value of $1000 Initial Investment at R=6.2% per Period
At Calendar Time, t
Yield, Rt
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
2% 5 10 8 4 2 3 4 5 6 7 8 9 10 11
2.0% 3.5 5.6 6.2 5.8 5.1 4.8 4.7 4.7 4.9 5.1 5.3 5.6 5.9 6.2
$1020 1071 1178 1272 1323 1350 1390 1446 1518 1609 1722 1860 2027 2230 2475
$1062 1128 1199 1273 1353 1437 1526 1621 1722 1830 1944 2065 2193 2330 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 oneperiod 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=1
)/T) .
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
log(1 + r T *+1 ) < [ log(1 + r t )]/ T * which for T* much larger than one would require that rT*+1
1 T*
<< 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 (V.10)
1 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 (V.11)
1 ( 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 - RT1)/ 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 ( - dt)/ Pt ( ) = 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 (V.13)
P 0(T) = exp [- r c(s)ds]
0 T
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 (V.15)
R c(T) = [ r c(s)ds]/T
0 T
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)
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 (V.17)
2 2 d R c(T)/d T = {d r c( )/dT - 2d R c(T)/dT}/T .
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.
+ + + *
Suppose that the yield curve at the current time is given by:
R c(T) = R + AT - BT ,
3 4 2 3
0T T
= R + [AT - BT ]/T, T T
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) ,
0t T
=R ,
From (V.16), we have that
t T .
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
2 2 d R c(T)/d T = 2(A - 3BT), 0 T < T
= 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 ) .
+
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
2 2 d r c(t)/d t = 6(A - 4Bt), 0 t T
= 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.
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 crosssection 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()} 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 fixedincome 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()}, then the current equilibrium price of the general default-free, fixed-income security must satisfy (V.18)
V 0 ( x1 , x 2 ,..., xT ) = x P0 ( ).
=1
T
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. Define by V 0( x 1 , x 2 ,..., x T ) x P ( )
0 =1 T
x P 0( ).
=1
and hence, > 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 [ x + /( P 0( )T)] bonds each of which pays $1 at its
maturity date periods from now and do this for bond maturities = 1,2,...,T. The cost of acquiring these bonds is
N P0 ( ) = x P0 ( ) +
=1 =1
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, > 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
Suppose instead that V0 were smaller than
x P ( )
0 =1 T
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 N discount bonds of maturity for = 1,2,...,T. As was shown, the total proceeds from issuing these bonds
N P ( )
0 =1 T
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, < 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 < 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 {N}. Thus, the hypothesized condition that < 0 is not consistent with equilibrium pricing, and this completes the proof that (V.18) must obtain in equilibrium.
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 x units of a -period discount bond for maturities = 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 x units of the discount bonds = 1,2,...,T and issue k units of the general security" and k <
0 means "issue | k | x 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 $ kx P0 ( =1 )
T $k $k V 0 - x P 0( ) =1
to the investor. So, by choosing the sign of k such that k > 0, the investor receives an immediate, positive cash payment of $k 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 $k. By assumption, the magnitude of k is not bounded. Therefore, as long as 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 = 0. Clearly, the latter is the sensible conclusion, and therefore, by his actions, the investor will "force" prices to adjust until = 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 > 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 = 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, > 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 = 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 = 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 (V.19)
V 0 = x
=1
T
/(1+ R ) .
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 (V.20)
B j = C j P0 (t) + M j P0 ( T j ) .
t=1
Tj
Equivalently, the price of the coupon bond can be written in terms of yields as (V.21)
Tj B j = C j /(1+ Rt ) + M j /(1 + RT j ) . t t=1 Tj
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 (V.22)
T T B j = C j [1 - 1/(1 + r ) j ]/r + M j /(1 + r ) j
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 (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.
-1
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
961 = 100 P0(1) + 1000 P0(2) + 0 P0(3) 968 = 100 P0(1) + 100 P0(2) + 1000 P0(3) 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
(iv)
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()} 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.
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 yieldto-maturity on a pure discount bond is simply its yield R T
j
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.
$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,
S15
j
*
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 3, and so on. At the end of year 15, this payment will have grown to $50 (1 + r t ) .
t= 2 15
Doing
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 given by (6049/1545)
1/15
yield-to-maturity. Thus, even if bond #2 had been the better investment, in general, its yield-tomaturity 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-tomaturity 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
# j, r j C j/ B j ) .
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)
+ * -T j ( r*j - r j )/( r*j - r + j ) = ( r j / r j )(1 + r 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 .
+ * + + * +
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 (V.27)
j D j t t t=1
j j j where t C j P 0(t)/ B j , for t = 1,2,...,T j - 1 and T j ( C j + M j )P0( T j )/ B j . Hence, t is
Tj
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
j with $ t B j invested in pure discount bonds which mature at date t, Dj is equal to the (value-
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 yieldto-maturity is the better buy. An alternative definition of duration sometimes used is (V.28)
j -t where t C j(1+ r*j ) / B j ,
D j t t
t=1
Tj
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 fixedincome 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 bidask 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 fixedincome 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 (V.30)
y = r AN /[1 - 1/(1 + r )N ] = $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
*
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 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
* 12
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
Obligation Treasury Bills U.S. Government Obligations Secondary Market Excellent secondary market. Maturities 3 month 6 month 1 year
30 days to 40 years
Promissory notes of issuing companies (industrial & financial) Obligation of bank accepting the deposit
No secondary market.
30-270 days
12.50
5.00
Certificates of Deposit
Bankers Acceptances
Obligation of bank against which draft is drawn and which accepts draft Obligations of the bank borrowing funds
Federal Funds
No secondary market.
30-180 days. 90 days is most common primary market maturity. Usually overnight.
$100,000 and up; $500,000 minimum trading unit; $1,000,000 most common trading unit. Issued in odd denominations; traded in $100,000 to $1,000,000 lots. Negotiated among participants, generally $1 million units. $10,000 minimum.
$70.6 billion outstanding in June, 1976 in CDs of over $100,000. $19.5 billion outstanding in May, 1976.
12.66
5.56
12.16
4.94
Discounted or occasionally interest bearing. Based on actual days in 360-day year. Yield basis. Interest paid on actual days in 360-day year. Interest & principal paid at maturity. Discounted. Interest based on 360-day year.
$26.9 billion average weekly volume for July, 1976 $188 billion outstanding in June, 1976.
13.55
4.73
Limited secondary marketgood for some short terms but very thin for long terms
6.71
Par basisinterest paid based on 360-day year. Interest & principal paid at maturity. Price basis; quoted in dollars per hundred dollars face value.
There is a forward market or one to two weeks. No futures market. No forward market. No futures market.
113
Finance Theory
114
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) (ii) (iii) Rules of thumb intended to facilitate comparisons of value among similar assets. Approaches based upon the economic theory of market value under certainty. 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. approach.) (The "Growth Opportunities"
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
(VI.1)
Z(t)
where d(t+1) is the dividend per share paid at time (t+1) and S(t+1) is the price per share (exdividend 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
(VI.2)
1 + r(t) = Z(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 exdividend 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) n(t) S(t) V(t) D(t) d(t) Z(t) REV(t) O(t) (t)
market value of the firm at time t number of shares of the firm's stock outstanding at time t price per share of stock at time t (ex-dividend paid at time t) n(t)S(t) if the firm is all equity-financed total dividends paid by the firm at time t dividend per share = D(t)/n(t1) return per dollar to the investor in the firm total revenues in period t = stream of cash receipts total operating cash outflow in period t after-tax profits in period t depreciation in period t cost of goods sold in period t taxes paid in period t gross investment (both new and replacement) in period t net (new) investment in period t i(t) + DEP(t) "gross" profit or net cash flow in period t (t) + DEP(t) 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)
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
(VI.3)
1 + r(T - 1) = Z(T - 1) =
or
(VI.3')
S(T - 1) =
Consider an investor who at time (T2) is going to buy the stock; the total return in dollars for holding one share for one period [from (T2) to (T1)] will be d(T1) + S(T1), and again to avoid arbitrage, we have from (VI.2) that
(VI.4)
1 + r(T - 2) = Z(T - 2) =
or
(VI.4')
S(T - 2) =
S(T - 2) =
1 + r (T - 3) =
118
Finance Theory d(T - 2) + S(T - 2) [1 + r(T - 3)] d(T - 1) d(T - 2) + = [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)] S(T - 3) =
(VI.6')
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.,
S(0) =
d(1) d(2) d(T) + SALV + + ...+ [1+ r(0)] [1+ r(0)][1+ r(1)] [1+ r(0)]...[1+ r(T - 1)] S(0)=
t=1 T
SALV
(VI.7)
=
T
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
(VI.8)
V(0)
Robert C. Merton
(VI.7')
S(0) =
t =1
d(t) [1 + R(t) ] t
and
(VI.8')
V(0) =
t =1
d(t)n(0) [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(t1)d(t).2 So, in general, V(0)
t =1
D(t) unless n(t) n(0). I.e., unless the firm neither issues any t t [1 + R(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 subperiod of time even if no dividends are paid during that time. E.g., if d(3) = 0, then the return from period 2 to period 3, will be
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).
(VI.2), we have that
Note it is "n(t1)" because we have assumed that dividends paid at time t go only to shares outstanding as of time (t1). 120
Finance Theory V(t + 1) = n(t)[(1+ r(t))S(t) - d(t + 1)] + m(t + 1)S(t + 1) = (1+ r(t))V(t) - D(t + 1) + m(t + 1)S(t + 1)
(VI.9)
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)
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
(VI.11)
V(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
Robert C. Merton
O(t ) = I (t ) + (t ) + CGS (t )
where the terms are defined in the glossary. Moreover, from the accounting identity (VI.12), we have that
(VI.13a)
(VI.13c)
1 {V(t +1)+ REV(t +1) - O(t +1)} [1+ r(t)] 1 V(t) = {V(t + 1) + X(t + 1) - I(t + 1)} [1 + r(t)] 1 V(t) = {V(t + 1) + (t + 1) - i(t + 1)}. [1 + r(t)] V(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
(VI.15a)
V(0) =
t=1
Finance Theory
(VI.15b)
V(0) =
t=1
(VI.15c)
V(0) =
t =1
Provided that
lim V(T) { } = 0, we can rewrite (VI.15a) - (VI.15c) for a firm that is T [1 + R(T) ] T
V(0) =
t=1
(VI.15b')
V(0) =
t =1
(VI.15c')
V(0) =
t =1
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 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 the total additional cost required to generate the stream of earnings {(t)} to be
[1 + R(t) ]
t =1
i(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. Note:
[1 + R(t) ]
t =1
(t)
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 increase over time. So, unless the economy as a whole is stagnant,
[1+ R(t)] t
t=1
(t)
will be a biased
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) (2) (3) the competitive ("normal" or "alternative") rate he can make in the market which is r. the earnings potential of the existing assets of the firm. 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) dollars at the end of each year. Define: r*(t) period on projects taken in period t (and F(t) . Then r* (t) is the average rate of return per I(t) 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 * r (t) - r -t ](1 + r ) ; and the present value of all such r
I(t)[
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 ]. V(0) = X ce + I(t)[ t r r t=1 (1 + r )
(VI.16)
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) X (t ) = X (t -1) + r *(t -1) I (t -1) = X ce + r *(s) I (s).
s =1 t -1
V (0) =
=
[ X (t ) I (t )] X (1) I (1) [ X (t ) I (t )] = + [1 + r ]t (1 + r ) (1 + r )t t =1 t =2
t 1 X ce I (1) 1 X r *( s ) I ( s ) I (t )] + [ + ce t (1 + r ) t = 2 (1 + r ) s =1
V (0) =
X ce t 1 r *( s ) I ( s ) I ((t ) + t r (1 + r )t t = 2 s =1 t =1 (1 + r )
Note:
r (1 + r )
t = 2 s =1
t -1
( s) I ( s)
t
= r *(1) I (1)
k =2
+ r *( s ) I ( s ) 1 1 = k (1 + r ) s k = s +1 (1 + r )
1 + ... k k = s +1 (1 + r )
1 1 1 1 1 = = k -s s j (1 + r ) j =1 (1 + r ) r (1 + r ) s k = s +1 (1 + r )
So, r *( s ) I ( s ) r *( s ) I ( s ) = s (1 + r )t t = 2 s =1 s =1 r (1 + r )
t 1
or
126
Finance Theory
* * r (s) I ( s) r (s) I (s) or = t s t = 2 s =1 (1 + r ) s =1 r (1 + r ) t -1 * 1 r (t ) - r ] V (0) = X ce + I t ( )[ t r r t =1 (1 + r )
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" priceto-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 i (0) and V i (0) denote current earnings and firm value for the ith
firm, then PE i
1 n V i(0) and the industry average is PE PE i . To find the "fair value" for n i=1 i(0)
the firm, we set V(0) = ( PE ) (0). A variate of this "quick-and-dirty" method is to compute the average earnings-to-price ratio,
EP = 1 n 1 n i(0) (0) . and set V(0) = EP i n 1 n i=1 V i(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) (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 payment per period; k = required rate of return (or discount rate) on the stream, then A = Applying the analog to the ith firm, we have that
y . k
V i(0)
i(0) =
EP i k i and
1 V i(0) = PE i = . i(0) ki
all firms in the same industry (or risk class) is the same, then k EP is the required return for the stream of the firm, and therefore, V(0) should equal
(0) (0)
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 the firm's value could be written as
V(0) =
t=1
(t) - i(t)
[1 + R(t) ] t
128
(0)
k
where k is the return per period on the firm? If current earnings are an estimate of and the firm requires no net new investment to generate these
earnings (i.e., i(t) 0) and if the term structure is essentially "flat" (at least, in "real" terms), then t
V(0) =
(0)
r
where r is the per period rate of return. In simple terms, if the firm's investment
(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 to an equivalent equal annual flow { (0)}, then provided the other conditions hold, V(0) =
(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 higherthan-average PE ratio. I.e., many stocks may have current PE ratios, high (or even undefined, if (0) 0), but have long-run
V(0)
(0)
129
Robert C. Merton
stocks are not what is meant by growth or glamour stocks. Rather growth stocks are stocks with
V(0)
(0)
that are unusually high persistently. In earlier times, such stocks were thought to be outside the traditional mode of analysis and
were simply excluded from such discussions. The problem was that EP =
1 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 opportunities. From (VI.16), we can rewrite the expression for
PE =
V(0) (0)
(noting that
(0) = X ce ) as
PE =
(VI.17)
* 1 1 V(0) 1 r (t) - r) ] = + I(t) [ r (0) r (0) t=1 (1 + r )t 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.
(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?
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. t
From
the
previous
analysis,
we
have
that
X(t) = X(t - 1) + r* I(t - 1) ; X(1) = (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) = X(t) where 0 1 . Then,
131
Robert C. Merton
X(t) = X(t - 1) + r* X(t - 1) = X(t - 1)[1 + r* ] = (0)[1 + r* ] t -1 and X(t) - X(t - 1) * = r = rate of growth of earnings. Substituting for X(t - 1) I(t) = X(t) = (0)[1 + r* ] t -1 into (VI.16), we have that
r * r (0)[1+ r* ] t-1 V(0) = + r r [1+ r ] t t=1
(0)
(0)
From p. II-15, (y
1+ r* ) 1+ r
= y t 1 = y j =
t=1 j=0
So,
V(0) =
(0)
( r* - r) 1+ r 1+ r (1+ r) r r *
or
(1- ) (0) r - r*
V(0) =
(VI.18)
Finance Theory
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 alreadyexisting 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)
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 quasiuncertainty 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 t define K(t) by [1+ K(t) ] [1+ k(t)], and use "[1 + K(t)] " everywhere in the formulas when j=1 t
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.
135
Finance Theory If I0 is the initial investment, then the payback period is that value of T such that
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 T
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)
*
X(t)
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 .
PV = - I 0 +
t =1
X(t) [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 0= - I0 +
t =1 N
X(t) [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.
T X(t) / I0 Profitability Index = PI = t 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.
1. 2.
Neglects the time value of money (no discounting) 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 I0=
t =1
T
X(t) (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, PV = - I 0 +
t =1 N
t X (t)
(1 + r )t
where 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
0 1 2 3
0 1 2 3
Example (continued)
Project A: Let x = 1 + i
139
( x1 , x 2 , x3 ) are x1 = 1 + i1 = 1.05 x 2 = 1 + i 2 = 1.05 so i A = .05 3 roots: x 3 = 1 + i 3 = 1.05 Project B: Let x = 1 + i PV = - 1,000,000 + 3,210,000 3,433,800 1,224,080 + =0 2 3 x x x
140
Robert C. Merton
Example (continued)
"Switching Points"
at k = .05 at k = .08
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.
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,
PV A = - 1000 + PV B = - 1000 +
1 15
= 3,177
and
= - 1000 + (15x214)
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 compound return per year for both. I.e., (1 + R A ) =
4177 has the solution RA = .10. So, the 1000 3210 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.
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
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.
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 Machine A costs IA and has operating costs per year C A ,C A ,...,C T A and has salvage value with
2 Machine B costs IB and has operating costs per year C B ,C B ,...,C T B . Assume replace each
For machine B
PB = I B + CB t t =1 (1 + k )
T2 t
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
I A+
A , because it is not used for its full life (1+ k )t (1+ k )T -T t=1
1
Ct
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,
IA +
T T1
' PA = PA +
(1 + k )T
t =1
t CA
(1 + k ) t
T T1 1 k + ( ) S
and
147
Finance Theory
' PB = PB +
(1 + k )
T2
1 PB = PB 1 + T (1 + k ) 2
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 replacements.
P= I +
n
If
is the present value of costs of the machine over one cycle, i.e.,
Ct , then, as above, the present value of costs over the product life will be t t =1 (1 + k )
P' = P +
(1 + k )
(1 + k )
2n
(1 + k )
3n
(1 + k )
4n
+K
1 1 1 1 = P 1 + + + + + K 1 2 3 4 (1 + k ) n (1 + k )n (1 + k ) n (1 + k )n
'
or
P' = P
N -1
X
j =0
where X
(1 + k )
So substituting for
X, we have 1 P' = P 1 1 1 + k n ) ( n = P (1 = k ) n (1 + k ) 1
'
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
y(1 AN = 1 (1 + k ) k
N
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.
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 = length of time between replacements (0 < 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() = present value of costs for one cycle for the machine replaced every years. Then P( ) = I +
t =1
Ct - S . (1 + k )t (1 + k )
Let P() be the present value of costs of the machine (replaced every years) over the product life. Then
1 1 P ( ) = 1 (1+ k )
*
P( ) .
The optimal replacement time, , will the be such that P( * ) P( ) for all possible = 1,2,3,...,T
We can also write the conditions in terms of equalized annual costs by:
y = kP( ) and then * is such that y * y for all possible .
Note:
will depend on k, the structure of the operating costs, and the salvage values, and
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 -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 + . 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 -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 P 0 ( T ) , and at date
t,
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, ) 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, ).
F t (L, M,T, ) = - L Pt (T - t) + M Pt (T + - t)
for t = 0,1,2,,T. Of course, the value of the forward contract to the borrower at time t is From the condition that
F 0 (L, M, T, ) = 0,
we
have
that
(VIII.2)
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 + )-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 + ), L/M is the T-period forward price for a default-free discount bond which
pays $1 at date (T + ). If we define TP0() to be the forward price for delivery at time T of a discount bond which pays $1 at date (T + ), then (VIII.3)
T
TP0()
Since
P 0 (T) is the amount one would pay today for one dollar delivered at date T and
P 0 (T + ) is the current price of the bond, from (VIII.3) the forward price TP0() 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 , the forward price associated with a forward contract for any default-free, fixed-income security can be computed. If
sV0 denotes
then
(VIII.4)
sV 0
t=s+1
sP 0 (t
- s) x t
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
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() rather than the forward price
TP0()
153
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 (VIII.8)
L t+1 - Lt =
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
(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
{ F
(VIII.12)
t =0
T -1
t +1
( Lt , T) - F t ( L t , T) } / Pt +1 (T - t - 1) = [ Lt +1 - Lt ] = LT - L0
t=0 T -1
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 .
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
[Note:
If
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 (VIII.13)
V T = N t [f(t +1) t =0 T -1
f(t)]/ P (T - t - 1) .
t +1
Consider the case where we choose N t = 1. In this case, from (VIII.13), the accumulated sum will be
V T = [ f (t + 1) - f (t )] + [ f (t + 1) - f (t )[
T -1 T -1
(VIII.14)
t =0
t =0
T -1
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
t + 1,
(VIII.15)
V T = [f(t + 1) - f(t)]
t=0
T -1
= 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. these combined positions, V T
'
can be written as
(VIII.16)
= L 0 - f(0)
f(T) = X(T)
because
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) and from (VIII.7), that (VIII.18)
f(t) = X(t)/ P t (T - t) . f(t) = Lt ,
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)
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.
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
General Asset Insured Asset's Current Value Term of Policy Stock Stock Price, $S Time until expiration of the put Exercise Price of the Put, $E
Maximum Insurance Coverage [Face Value of Policy (maximum loss to insurer) Amount of the Deductible (maximum loss to insured) Insurance Premium Important Differences
$[S E]
$3.875
$15.25
Three Ways of Reducing Risk 1) 2) 3) Diversification: "Mixing" less-than-perfectly correlated risky assets Substituting the riskless security for risky assets 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
Stocks Mixed With Commercial Paper Strategies Versus "Protective Put" Strategies July 1963- June 1977
Protective Semi-Annual: Stocks
1/
Protective Put
3/
Put
2/
(E = S)
(E = .gS)
Average Rate of Return Standard Deviation Highest Return Lowest Return Average Compound Return Growth of $1000
Semi-Annual:
1/
1/
Commercial Paper
Average Rate of Return Standard Deviation Highest Return Lowest Return Average Compound Return Growth of $1000
*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).
Equal-dollar Weighted Portfolio of 30 Dow Jones Industrial stocks rebalanced semiannually. Returns include reinvesting all dividends. No provisions for taxes or transaction costs. Same as footnote 1 plus a six-month put option with exercise-price-equal-to-initialstock price for each share of each stock.
161
2/
1/
1/
Finance Theory
3/
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
2/
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
NYSE Stocks Protective Put 30-Day U.S. Treasury Bills
Per Month:
Average Rate of Return Standard Deviation Highest Return Lowest Return Average Compound Return Growth of $1000 Average Annual Compound Return
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. 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.
2.
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)] = V(T) - B + Max [0, B - V(T)] = Max [0,V(T) - B].
(VIII.20)
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 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. 3. 4. 5. Are there costs to bankruptcies? Are there tax features unique to corporate debt? What are the effects on control of the firm? 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. 2. 3. 4. 5. growth rate of future sales stability of future sales the competitive structure of the industry the asset structure of the industry 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)
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.
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:
V0 = X k0
(IX.1)
Let
F
B0
= (expected) annual interest payments on debt outstanding = market value of debt outstanding
167
Robert C. Merton
E
= (expected) annual earning available to shareholders
= X -F S0 ki F B0
= market value of equity outstanding
= cost of debt =
(IX.2)
B0 S0 B0 S0 k0 = ki + ke= + ki + + ke . 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
Example: k e = 10% and k i = 5%; net operating earnings = $1000 Case 1: B 0 = $3000 which implies F = $150 and so, Net Operating Earnings $ 1000
B0 S0 k0 = ki + ke V0 V0
3000 8500 (.05) + (.10) = 11500 11500
(150)
$ 850
______
Leverage Factor
B0 = 3000 S 0 8500
.35
(300) $ 700
______
169
Robert C. Merton Market Value of Equity ( S 0 ) + Market Value of Debt ( B0 ) Market Value of Firm ( V 0 ) $7000 6000 $13000
Leverage Factor
B0 = 6000 S0 7000
.86
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. 3.
Assumes that beyond some point of leverage, k e rises at an increasing rate. 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:
or
B0 k e = k0 + ( k0 - k i ) S0
(IX.3)
172
Finance Theory (First) Proof of MM 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. debt (financed completely by equity) and company #2 has some debt.
Company #1
Company #2
X 1= X
1 B0 = 0 at rate k i = r
X 2= X
2 B0 > 0 at rate k i = r
1 1 S 0 (= V 0 )
S0
V0
V 0 (= B0 + S 0 )
Consider an investor who currently owns S 2 dollars of stock in company #2. If = % of total
shares of company #2 held by this investor, we have that S 2 = S 0 .
2
2 Suppose that V 0 >V 1 0 , 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
2 fractional claim, , times the portion of earnings available to shareholders, X - rB 0 . I.e.,
(IX.4)
2 Y 2 = (X - r B0 )
The investor could sell out his present holdings and choose an alternative portfolio as follows: Step 1: sell his current holding for S 2 (= S 0 ) dollars.
2 2 Step 2: borrow ( B0 ) dollars.
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 = S 0 + B0 = ( S 0 + B0 ) = 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
of #1's earnings.
S0 = V 0
and
1 ( S1 / S1 0 )X = ( S 1 / V 0 )X dollars of return.
( )
(IX.5)
1 V02 V02 V0 2 2 = = + X rB X rB X 0 0 1 1 V V 0 0
or
1 V02 V0 Y1 = Y2 + X V1 0
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.
1 2
"switch" to portfolio #2 from portfolio #1 until V 0 V 0 . The argument goes through in precisely the same way, if it was assumed that V 0 > V 0 . Therefore, to avoid arbitrage or dominance, (IX.6)
1 2
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):
Case I:
174
Return
2 X - rB0
2 B0
2 r B0
Total
1 2
2 V0
X
1 2
So, if V 0 > V 0, the investor gets the same return for (V 0 - V 0) fewer dollars invested.
Case II:
Alternative Investment:
Transaction Investment
1 S1 0 V0
Return
) ) (
X
2 - rB0
2 B0
Total
1 2 V0 B0
2 X - rB0
2 1 2 1 If V0 > V0 , then the investor gets the same return for (V0 -V0 ) fewer dollars. Hence, to 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.
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 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
T
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
(IX.7)
Promised yield = R +
g(1 + 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.
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.
V0=
X (1 - T c )(1 - T p ) (1 - T p ) k 0
(1 - T c ) X k0
179
Robert C. Merton
If the debt is riskless, MM argue that R = rBc, then the value of the debt is
and they show, by their arbitrage argument, that
(1 - T p )r Bc = Bc , (1 - T p )r
(IX.9)
V=
(1 - T c ) X
k0 = V 0 + T c Bc
+ Tc r
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 )]
(IX.11)
dV dBc
= - (1 - T c ) + 1 = T c as MM claim .
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 )] ,
V(Y ) = V 0
and
dV
(IX.15)
=1-
dB 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.
Loan Commitment = Maximum that can be Borrowed = "Line" L Principal Amount Borrowed = Gross Borrowings B Stated Interest Rate on Loan R = r + where r = "prime" rate and amount "over prime" charged.
181
Robert C. Merton CB
required (by the bank) amount to be kept on deposit in free balances in the
= 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[CBCB] where Rp penalty rate and CB compensating balances actually maintained. D d amount of noninterest-bearing demand deposits maintained by firm 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., LB]. Let RL rate paid as a line fee M actual amount of money available for corporate purposes
(IX.16)
M = B - D+d
I = R B + R L (L - B) + R p [ CB -CB]
182
Finance Theory
(IX.18)
R + R C2 RL B + RL + R C1 L RPCB / M RT = P P R + R C2 RL B + RL + R C1 L RP .84 D / [ B D + d ] RT = P P
(IX.18')
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 [ dB dD ] M
(IX.19)
dR = ( R R L ) (.84 c 2 ) R p /M dD M fixed
dR T dD
< 0 if R p >
M
(R RL ) (.84 c2 )
D optimum = 1.19CB + d
dR T dD
= 0 if R p >
M
[ R RL ] (.84 c2 )
dR T dD
> 0 if R p >
M
[ R RL ] D optimum (.84 c2 )
=d
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 2,000,000 3,000,000 4,000,000 5,000,000 6,000,000 7,000,000 8,000,000 9,000,000 10,000,000
53.49% 37.81 32.59 29.97 28.41 27.36 26.61 26.05 25.62 25.27
[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.
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 called "shares". Let Pi
o
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"). Denote by N i the number of shares (or units) of the i security that he chooses to purchase. His set
0 of feasible choices is determined by his budget constraint: W o = N i Pi . i=1 n
th
1 probability distribution for the end-of-period price per share, Pi , for each asset. Then, his end-of-
N P
i i=1
1 i
. Define wi
N i Pi to be the fraction of Wo
his wealth invested in the i security and define the (random) variable return (per dollar invested)
in the i asset to be Z i
th
(X.1)
w = 1 .
i i=1
{ N 1 , N 2 ,..., N n }
or equivalently
(X.2)
Max { w1 , w2 ,..., wn }
E{U( W o wi Z i )} subject 1 = wi .
1 1
186
Finance Theory
Suppose that one of the available securities is "riskless" and offers a return per dollar of R with
certainty. If, by convention, we choose the n security to be riskless, then we have that Z n =
n n -1 with certainty. Note that 1 = 1 wi = 1 wi + wn m or wn = 1 - 1 wi where m n - 1 .
th
Pn =R o Pn
Define:
n Z 1 wi Z i =
W1 = return per dollar on the portfolio. With a riskless asset, we can rewrite Z as Wo
n m m m m Z = 1 wi Z i = 1 wi Z i + wn Z n = 1 wi Z i + (1 - 1 wi ) Z n = 1 wi ( Z i - Z n ) +
(X.3)
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
(X.4)
or
(X.4')
Call
wn = 1 - 1 wi )
* m * *
(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
borrowing. Since the n
m
th
= w E{U ( W )( Z - R)}+RE{U ( W )} .
* i 1 i 1 1
(X.5)
E{U ( W 1 )Z * } = RE{U ( W 1 )} .
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
= 1W1 = 1W = fraction of his wealth invested in the "safe" asset #2, then the investor's portfolio problem can be described by:
E{U (W 1)} = Max E{U (W o[ w( Z 1 - R) + R ])} Max {w} {w} b 2 2 = Max E{W o[ w( Z 1 - R) + R] - W o [ w ( Z 1 - R) 2 + 2 wR( Z 1 - R) + R 2]} 2 {w}
* Maximizing, by the usual calculus, gives the condition that w = w , the optimal portfolio, when
0=
dEU 2 2 = E{ W o ( Z 1 - R) - bW o [ 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
(X.6)
* w=
(1 - bW o R)E[ Z 1 - R] 2 bW o E[( Z 1 - R ) ]
If W o R >
1 (i.e., if it is not possible to achieve the "satiation" level of wealth for certain), then a b
* w > 0)
necessary condition for him to hold some amount of the "risky" asset (i.e.,
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 .
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-ofperiod 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 wellapproximated 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.
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 probability P n is defined as
E(X) = X = Pi xi
i=1
2.
is defined by
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
2 2 X = 0 = X . An alternative useful formula for X is:
= P i( x i -X )
2 X i=1
where X = Pi x i
i=1 n n n i=1 i=1 i=1
= P i( x i2 -2 x i
i=1 n
X + X 2 ) = P i xi2 - 2X Pi x i + X 2 P i
= P i xi2 - 2 X 2 + X 2 = E[ X 2 ] - ( X )2
i=1
3.
)( Z -Z )
j
191
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)
(3)
Cov(X, Z) = Pij ( xi - X
[
n i=1
i=1 j=1 X i
( x - X ) P ( Z -Z )
i Z j j j=1 X i
][
n i=1
)( Z -Z ) =
j
but
PiX ( xi - X
i=1
)= P
i=1
xi - X PiX =
X - X =0
X
and
XZ
Cov(X, Z)
XZ
XZ
XZ = 1 , then X
and
are
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
case, if wi = fraction (of each dollar invested in the portfolio) invested in the i
th
n define Z = 1 wi Z i as the (random) variable return (per dollar invested) on the composite security or
( w1 , w2 ,..., wn )
n
E ( Z ) = Z i , is E ( Z ) = E[ wi Z i ] = wiE (Z i)
1 1
= w Z
1
expected returns on the individual assets. Let i be the variance of the return on the i
2
asset (i.e.,
2 i2 = E{[ Z i - Z ] } and ij be the covariance between the returns on the ith and jth assets (i.e.,
Note :
wi( Z i - Z i ) = wi Z i - Z
i=1
n
i=1
wi( Z i - Z i ) i=1
n
=
n
( (
n i=1
n
wi Z i - Z i )
) (
n
i
w j( Z j - Z j )
j=1
= wi w j( Z i - Z i )( Z j - Z j )
i=1 j=1
Hence, E { [ wi ( Z i - Z i ) ]
i=1 n n
}= E [ w w ( Z
i j i=1 j=1
- Z i )( Z j - Z j )]
= wi w j E[( Z i - Z i )( Z j - Z j )]
i=1 j=1
= wi w j ij = 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
2 2 E[( Z 1 - Z 1 )2 ] = 1 = E[( Z 2 - Z 2 )2 ] = 2 2 = v , i.e., they have the same variance of return. Let
12 = E{( Z 1 - Z 1 )( Z 2 - Z 2 )} =
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 2 = E[(Z - Z ) ] = E[( w1 ( Z 1 - m) + w2 ( Z 2 - m) ) ] 2 2 = E[ w1 ( Z 1 - m )2 + 2 w1 w2 ( Z 1 - m)( Z 2 - m) + w2 2( Z 2 - m ) ] 2 2 = w1 E[( Z 1 - m )2 ] + 2 w1 w2 E[( Z 1 - m)( Z 2 - m)] + w2 2 E[( Z 2 - m ) ] 2 2 2 = w1 1 + 2 w1 w2 12 + w2 2 2 2 2 2 2 2 2 2 2 = w1 v + 2 w1 w2 12 v + w2 v = v [ w1 + 2 w1 w2 12 + w2 ].
Hence, as the relative proportions ( w1 , w2 ) are varied, the variance of the portfolio is changed.
2 Since w1 + w2 = 1, to see the effect on , we first substitute w1 = w and w2 = 1 - w1 = 1 - w, to get 2 2 = v 2 [ w2 + 2w(1 - w) 12 + (1 - w ) ] = v 2 [1 - 2(1 - 12 )w(1 - w)] . 2 So , as a function of the "mix parameter" w, is a parabola. Since the expected return is the same
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,
calculus as follows:
* w , by using the
min
w 2
d 1 |w=w* = 0 = - v 2 [2(1- 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
2
for various correlation coefficients and mixes. From above, the minimum
variance (corresponding to w =
1 ) for a given 12 is 2
2 = min
v (1 + 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., 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.
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
2 E{( Z 1 - Z 1 )2 } = 1 and E{( Z 2 - Z 2 )2 } = 2 are not equal, and by convention, assume that 2 2 2 2 > 1 . Form a composite security from a combination of security #1 and security #2. Denote its 2 random variable return by Z and its expected return by Z and variance by . Let w = fraction
Figure X.3.a
If Z 2 < Z 1 , then
197
Robert C. Merton
Figure X.3.b
The rest of this section is devoted to finding the characteristics of Z changes and for different assumptions about Z 1 and Z 2 .
2 (i.e., Z , ) as w
Case 1: Suppose that Z1 and Z 2 are perfectly (positively or negatively) correlated, i.e.,
12 = + 1, then
2 2 = w 2 2 2 + 2w(1- w) 1 2 + (1- w ) 1 = [w 2 +(1- w) 1 ] 2 2
or
= | w 2 +(1- w) 1 |
i.e., is in a linear relationship to 1 and 2 ; and linear in w for 12 = + 1.
198
Finance Theory
Figure X.4a
The only other case where is in a linear relationship to 2 and 1 is when 1 = 0, i.e.,
2 2 2 security #1 is riskless. Then = w 2 and = | w | 2
Figure X4.b
199
Robert C. Merton
Proof:
1. 2.
Then,
2 + 2 1 2 < 2 12 1 2 .
But
12 1,
so it must be that
( 1 - 2 )2 < 0
2 + 2 1 2 < 2 1 2
or
2 + 2 1 2 - 2 1 2 < 0.
Hence,
2 + 2 1 2 - 2 12 1 2 0.
3.
Suppose 1 + 2 - 2 12 1 2 = 0. Then
2 2 2 2 + 2 12 1 and 1 2 - 2 1 2 = 2( 12 - 1) 1 2 , or ( 1 - 2 ) = 2( 12 - 1) 1 2 . But
and
2 [ ] w2 - 2[ ]w + 1
(X.7a) (X.7b)
d2 2 2 = 2 {w[ 1 + 2 2 - 2 12 1 2 ] - [ 1 - 12 1 2 ]} dw 2 2 + 2 d w[ 1 2 - 2 12 1 2 ] - [ 1 - 12 1 2 ] = dw
200
(X.8b)
2 of w is a hyperbola. From (X.8a) and (X.8b), both and are strictly convex functions
of w.
Because
to proportion, wmin . Of course, this value of w minimizes as well. The minimum point will occur where the 1 have that (X.9)
wmin =
2 ( 1 - 12 1 2 ) 2 [ 1 + 2 2 - 2 12 1 2 ]
st
d2 d < 0 (and < 0) and for Thus, for w < wmin , dw dw w > wmin , d2 d > 0 ( and > 0). dw dw
Can
wmin > 1 ?
wmin =
2 2 2 > 1 or 2 > 1 . So, 12 > 1 which is impossible. Hence, wmin < 1. Can wm < 0? From
201
Robert C. Merton
1> 12 ). So, if 2 is not too much larger than 1 then 0 < w min < 1 (Note : if 12 0, then 2
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 as a function of w as:
Figure X.5
Note in Figure X.5: because is a convex function of w, the curve will always lie below the
straight line (w = 0, 1 ) and (w = 1, 2 ). For our purposes, it will be much more useful to work with the relationship between and Z rather than with 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
d d = = dw = dw d Z dw d Z d Z ( Z 2 - Z 1 ) dw d d dw
(X.10)
202
Finance Theory
and (X.11)
d 2 2 d dw = . d Z 2 ( Z 2 - Z 1 )2
2
d dZ
d . From dw
2
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
min occurs where
Note: d
dZ
= 0 or where
dZ = d
The curve in Figure (X.7) traces out all the feasible expected return-standard deviation (or meanvariance) combinations possible from the two risky assets. And to each ( Z , ) point on that curve, there corresponds a unique portfolio of these assets described by (w,1 - w).
204
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
2 Z 1 has mean, variance, and covariances ( Z 1 , 1 , 12 , 13 , 14 ) and Z 2 has mean, variance, I 2 and covariances ( Z 2 , 2 , 21 , 23 , 24 ). If w is the fraction of composite security #I invested
in security 1 and
(1 - wI )
and variance
2 I
as
Second, consider the composite security #II constructed from other securities with returns
2 Z 3 and Z 4 , where Z 3 has mean, variance, and covariances ( Z 3 , 3 , 31 , 32 , 34 ) and Z 4 II 2 has mean, variance, and covariances ( Z 4 , 4 , 41 , 42 , 43 ). If w is the proportion of the II composite security #II invested in Z 3 and (1 - w ) is the proportion invested in Z 4 , then the II II 2 return of #II, Z II = w Z 3 + (1 - w ) Z 4 , and Z II has expected return Z II and variance II as
Third, one can compute the covariance between composite securities #I and #II, Cov( Z 1, Z II ) 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
a( Z I , 2 If I , I,II ).
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,
2 2 if I > 0 and II > 0 and Z I and Z II are not perfectly correlated, then Figures (X.5) -(X.7)
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]
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 meanvariance "tradeoff" will be like either Figures (X.5) - (X.7) or Figure (X.4) as in the two-security case.
Suppose that there are n securities with (random) variable returns Z i with expected returns
E( Z i ) = Z i ; variances of returns i2 ; covariances of returns ij , for i, j = 1,2,..., n . Further,
2 suppose that all n securities have uncertain returns (i.e., i > 0 for all securities). By mixing
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
2 Efficient portfolio set. The point ( Z min , min ) is called the minimum-variance portfolio, and is
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
209
Robert C. Merton
1 =
w
i=1
(X.12a)
and (X.12b)
Z = E(Z) = wi E( Z i ) = wi Z i = m.
i=1 i=1 n n
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 minimization problem which can be solved by using Lagrange multipliers, i.e., minimize subject to (X.12a) and (X.12b), or (X.13) Min{
n n 1 n n ] + [m + [1 wi wi Z i ]} 2 wi w j ij 1 2 i=1 j=1 i=1 i=1 n n
2
2
2 where 1 and 2 are the multipliers and remember that = wi w j ij . To determine a i=1 j=1
critical point, we differentiate (X.13) with respect to w1 , w2 ,..., wn , 1 , 2 and set each partial derivative equal to zero, to obtain the optimality conditions
0 = w j ij - 1 - 2 Z i for i = 1,2,..., n
j=1 n n
(X.14)
m = wi Z i
i=1
1 = wi
i=1
These are
(n + 2)
(n + 2)
unknowns
w1 , w2 ,..., wn , 1 , and 2 . Let [ vij ] be the elements of the inverse of the variance
210
Finance Theory
; D BC - A ,
2
n
m wi =
(CZ j - A) + v ij(B - AZ j )
v ij
j=1 j=1
(X.15)
2 =
D Cm- A D
1 =
B- Am D
(X.16)
=
2
Note: the variance of the Frontier portfolio set is a 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: (X.17a)
m= A 1 + D(C 2 - 1) Frontier 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., i > 0 ). What
2 st happens if a (n + 1 ) riskless security becomes available with (certain) return R? Before
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 ] = 2 P.
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 variable return on the new portfolio, then
Z = w( Z P ) + (1 - w)R = w( Z P - R) + R Z
is the (random)
2 or
=| w | 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
= | w | P , if w 0 , then,
( - R) Z = w( Z P - R) + R = Z P + R . Graphically,
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 P = ). 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 , ) 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)
v ( Z - R)
ij j
(X.18)
wi =
j=1
(A- RC)
, i = 1,2,...,n
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
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
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 , 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 * , * ) 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 , 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 , , R) to choose his optimal portfolio.
218
Super Efficient Portfolio of Risky Assets (Optimal Combination of Risky Assets) Optimal Portfolio of Assets Alter Shape of Payoffs on Underlying Optimal Portfolio Structured Efficient Form of Payouts to Client Client Households Entrepreneurs Endowment Corporation
Components of Best Performing Risky Assets Only Portfolio: Diversification Risk Modulation
Risk Modulation through Insurance or non-linear leverage Pre-programmed dynamic trading Building Block State-Contingent Securities to create specialized payout patterns
219
Robert C. Merton
Fixed-Income
Real Estate
Other
Implementing Diversification as one of the Three Risk Management Tools Indexing of portfolios
220
Finance Theory
Active Management:
Asset-Class Allocation: Macro-Sector Market Timing Long-Short combinations to change fractional allocations from Benchmark Weights
ASSET CLASS Small-Cap Equity Mid-Cap Equity Large-Cap Equity Emerging Market Equity Domestic FixedIncome Real Estate BENCHMARK WEIGHT 5% 10% 30% 15% 30% 10% LONG (SHORT) INCREMENTAL +5% 0% (10%) (5%) 5% 5% REVISED WEIGHT 10% 10% 20% 10% 35% 15%
100%
0%
100%
Optimal Weighting Security Analysis Technical Analysis Proprietary Derivative-Security Pricing Models
Super-Performing Micro Aggregate Excess-Return 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
0
Percentage of Optimal Portfolio Invested in OCRA
Optimum Percent
Risk
Optimal Risk OCRA Risk
100%
Hedge or Leverage OCRA to obtain Optimal Portfolio Implement Macro Market timing of Risky versus Riskless Asset Performance
222
Finance Theory
Insurance and non-linear leverage Transform Payoff Pattern to fit precise preferences: custom design
0 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 ones 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
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 w*
k
fraction of the
th
optimal portfolio; Wo
k
th
investor;
th k k d k w* W o = number of dollars invested by the k investor in the fund = demand for the fund.
Define:
M=
n
N
i=1
225
Define:
VR = equilibrium market value of the aggregate supply of riskless asset (which may be zero).
+V = N
R i=1
Pi + V R
d k = w*k W o = M;
k k =1 k =1
(1 - w
k =1
k *
k )W o =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
(XI.3)
* Di wi M, i = 1,2,..., n.
But, in equilibrium, the supply of asset i must equal the demand, i.e.,
(XI.4)
Di = N i Pi , i = 1,2,..., n .
wi =
N N i Pi = 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-ofrisky-assets portfolio allocated to security i must equal the ratio of the market value of the i
th
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 } 2 M=
n
In equilibrium,
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
Let
w1 = % invested in security i (any security not on efficient frontier) w2 = % invested in the market portfolio (optimal combination of risky assets)
(security i has expected return Z i and standard deviation 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
E(Z) = Z = w1 ( Z i - R) + w2 ( Z M - R) + R
228
Finance Theory
and the variance is
(XI.8)
2 2 2 Var(Z) = 2 = E{(Z - Z )2 } = w1 i + w2 2 M + 2 w1 w2 iM
th where iM = the covariance between the return on the i security and the market portfolio.
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
w2 =
-R m- R (m - R) - w1 Z i = (1 - w1 ) . ( ZM - R Z -R ZM - R
Substitute for w2 from (X1.9) into the expression for the variance (XI.8) to get
2 (m - R )2 2(m - R)( Z i - R) w1 ( Z i - R) 2 2 2 = + + w1 w1 i M 2 ( Z M - R )2 ( Z M - R )2 ( Z M - R ) m- R ( - R) + 2 w1 - w1 Z i iM ( Z M - R) ZM - R 2
(XI.10)
(XI.10')
( - R) 2 + 2 iM - Z i M ( Z M - R)
2 (m - R) 2 M (m - R ) w1 + ( Z M - R) ( Z M - R )2 2
229
Robert C. Merton
d2 = 0. d w1
(XI.11)
= 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 M 0. iM ( 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,
(XI.13)
iM Zi - R= 2 ( Z M - R) M
Z i - R = i ( Z M - R)
230
Finance Theory
where (" beta" ) i iM . 2 M
(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
and, in equilibrium, the market portfolio is an efficient portfolio, and hence, must be on the line. I.e., Z M = R + r e M , or
(XI.16)
re
ZM - R ,
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 iM i = 1,2,..., n
232
Finance Theory
Figure XI.3
Note: kM can be negative in which case the equilibrium expected return on that security, Z k , will
be less than R.
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, iM iM i M
between the return on the i
th
where
iM correlation coefficient
233
Robert C. Merton
(XI.18)
Z i - R = r e i iM .
For a fixed risk premium, Z i - R, and fixed price of risk-reduction, r e , what value for iM allows
i to be as small as possible? Clearly, the largest value of iM , namely, iM = 1. Note: when iM = 1, (XI.18) and (XI.15) are the same. So again, we see that all efficient portfolios (with > 0 )
are perfectly correlated with the market portfolio.
Figure XI.4
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 - R) + R
and its
* 2 * variance is Var( Z ) = [ w ] M . Consider the effect of a small change in this portfolio achieved by
increasing the fraction held in asset i by and decreasing the holding in the riskless asset by . The return on this portfolio can be written as
Z = w* Z M + (1 - w* )R + Z i - R = Z* + ( Z i - R) ,
Z = Z * + ( Z i - R)
and
(XI.19b)
Var(Z) = Var( Z * ) + 2 w* iM + 2 i2 .
The effect of this small change on the mean and variance can be determined by differentiating (XI.19) with respect to and evaluating the derivative at = 0. That is,
(XI.20a)
[d Z /d ] =0 = Z i - R
and
(XI.20b)
[d Var(Z)/d ] =0 = 2 w* iM .
Case (i): Suppose that iM < 0. If Z i - R 0, then by moving from = 0 to > 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
235
Robert C. Merton Case (ii): Suppose that iM > 0 . If Z i - R 0, then by moving from = 0 to < 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.
iM > 0, then Z i - R > 0.
Case (iii): Suppose that iM = 0. If Z i > R, then by moving from = 0 to > 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 variance. from = 0 to < 0, one could increase the expected return and not increase its
Because either possibility would violate the efficiency of the market, it follows that
Z i = R if 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
parts of the returns on asset k and the market portfolio, respectively. From the definition of k and M and from (XI.13), we can write the return on asset k as
Zk = R+
(XI.21)
kM ( Z M - R) + k 2 M = R + kM ( Z M - M - R) + k 2 M = ak + k Z M + uk
236
Finance Theory
that
= kM - k 2 M =0
2 since k = kM / M . That is, u k is uncorrelated with the return on the market. For the reader
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.
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" ( 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 k = 1 have expected returns = Z M ; securities with k = 0 have
expected returns = R.
is only a
heuristic. (XI.21) was derived simply by construction with no assumptions about the joint distribution of
uk
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
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, k , can always be written as the sum of systematic
and unsystematic risk: namely, as (XI.23)
k = k M + uk .
From (XI.22) and (XI.23), the total variance of the return can be written as
2 k =
2 2 k M
(XI.24)
Note: From (XI.14), the equilibrium, (expected) reward or risk-presmium or excess return, Z k - R, is proportional to k M , the standard deviation of the systematic part of the total risk, and not 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.
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
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.
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
amount of insurance coverage purchased by each person as a fraction of his wealth. Let random variable describing the death of the i
th
to be the
yi
be a
and yi = 1 if person i dies during the year. Assume that the mortality tables are such that
people. Hence, is the expected number of deaths per person (0 < < 1) . Let Y N =
y
i=1
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 ] = N ; 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
2 2 Var( C N ) = 2 N = Nc V
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 invested in equity of the insurance company is Z N
R( NP N + K N ) - C N KN , and
E[ Z N ] = Z N =
R[( NP N ) + K N ] - Nc KN
Var [ Z N ] = 2 ZN=
= Nc 2 V 2 2
K
2 N 2 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:
supply = K N = Demand for asset; or KN = 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
v Z j( Z j - R)
N
j=1 w* ZN=
(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
Cov( Z N , Z M ) =
j=1 n
w*j
Cov( Z N ,Z j ) =
w* Z N Var
( Z N )= (
KN
NW
)(
Nc 2V 2 K2 N
)=
Nc 2V 2 . NW K N
From Section XI, we have from the Security Market Line that c2 V 2 R = ( , ) = Cov r Z Z r S N M S ZN WK N Substituting for Z N , we have that
R[ NP N + K N ] - Nc KN c 2V 2 - R= rS or WK N
qW 1 c 1 2 2 2 + r S q WV = r S cq V and therefore , R NR R NR
ZN = R+
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:
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
equity, the customer will not receive the full benefits promised, but only
R ( 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
in the event of his death, conditional on the company being solvent, pays the
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. R N = reserves requires = R (NPN + K N )
Let
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.
max C N = Nc
I.e., or
or
R max = N
= Nc = R( NP N + K max N )
max N
= Nc(
1 cq V 2 . - ) - rS R R
Nc large and
capital required to meet the maximum reserves could be prohibitively large. Further, if 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
max capital as K N ( p ). Note: K N (0) = K N .
Prob{C N > R N } =
2
Prob{( yi)c
1
KN p
( *)
rs cqV
V c N
where X N
yi-N
1
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 ) + } 1 - [( ) N + ] = p* Vc Vc N N N N
Prob{ X N > (
where
[ ]
is the
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
245
Robert C. Merton
(p )
*
.5000 .1600 .0230 .0100 .0010 .0001 (1 chance in 10,000) .00004 (1 chance in 25,000)
* Vc R K N ( p ) r S qV * + ) ( p* )N - r SqV for large N. Thus, or K N ( p ) = ( Hence, ( p ) = ( ) R Vc N N * * * for a given p (or ( p )) , we have an expression for the required equity capital for large N.
* KN( p ) (
Vc )( p* ) N R
limit ( P N ) =
N
c ; R
lim ( Z N ) = R;
N
limit ( 2Z N ) =
N
R > 0. 2 ( p* )
limit
N
lim (
N
* KN( p ) ) = 0; NW
* Required Equity KN( p ) = ( )= 0 limit max Maximum Equity limit N N KN * Suppose Suppose = .0025; c = $30,000 p = .00004 =
1 25,000
2 V = .0025 R = 1
[ ( p* ) = 4]
246
Finance Theory
NPN
K max = K N (0) N
K N (.00004)
10,000 4 (1x10 ) 90,000 4 (9x10 ) 1,000,000 6 (1x10 ) one million 9,000,000 6 (9x10 ) 8 1x10 one hundred million
$299,250,000 $2,693,250,000 ($2.7 billion) $30,000,000,000 ($30 billion) $270 billion $3 trillion 12 (3x10 )
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
minimum premium for a $30,000 policy would be $225 versus $75 charged by the company.
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
max
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
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 M . 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 M 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 M . (R is, of course, observable) is to use historical data to estimate ZM - R and M or ( ZM - R)/ M .
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
k > 0, then security k is "under priced" k < 0, then security k is "over priced"
A portfolio with a consistent positive alpha () shows evidence of ability to forecast a security (or securities) better than the market.
Figure XIII.2
ZP - R > ZM - R . An inferior-performing R M
251
Robert C. Merton Consider a portfolio constructed by a manager with superior stock selection skills that has a positive alpha. I.e.,
ZP - R = P ( ZM - R) + P , P > 0 P = P M + Var[U P]
2 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:
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 = w1 P + [ w1 P + w 2][ Z M - R] + R
2 2 2 2 2 2 2 2 Var(Z) = w1 P + w 2 2 M +2 w1w 2 PM = w1 P + w 2 M + 2 w1w 2 P M 2 2 2 2 2 = w1 Var[ U P ] + [w1 ] M P + 2w1w 2 P + w 2 2 = w1 Var[U P ] + [ w1 P + w 2]2 2 M
2 M ( ZM - R ) 2
dVar(Z) dw1
*= w1
2 * Var[ * ] M P * = 2 w1 = 0 at w1 = w1 UP ] - [ ZM - R - w1 P 2 ( R ) ZM
P 2 U P ( ZM - R)[Var[ 2P ] + ( M ZM - R)
( Z-R)
( - R) > ZM Var(Z) M
when
* Z = w1 ( Z P -R) + w * 2( Z M -R) + R= Z M
2 M ( ZM -R) 2
Hint:
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 - k )R + k ZM + U k , k = 1,2,..., m M
or over-valued. I.e., where E( U k ) 0. Define: U k k + k where k is a random variable such that E(k ) = 0;
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 k and k 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 in the i
a
th
w R = fraction of the portfolio invested in the riskless asset. Further, restrict the portfolio weights
to satisfy
256
Finance Theory
m
(XIII.4a)
a w ia + w a M + wR = 1 i =1
(XIII.4b)
w ia i + w a M=0
i =1
(XIII.5)
= w ia [i ( ZM - R) + i + i] + w a M ( ZM - R) + R
i =1
from (XIII.3)
= R + w ia i + w ia i
i =1 i =1
from (XIII.4b)
Cov( Za , ZM ) 2 M
= 0.
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 (i , i) 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 - a )R + a ZM + a + a
where:
Note that:
a = 0; a
i=1
w ia i;
w ia i
i=1
E( Za ) = a + R = (XIII.7)
w ia i + R = Za;
i=1 m
2 Var( Z m) = a
w ia w aj ij
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
2 at a , then
a {w1 ,..., w a m}
Max
=
{ Za +
2 m m a a [ a - w i w j i j ]} 2 i = 1 j=1
Max
{ w ia }
{R +
i =1
w ia i +
2 [ a 2
w ia w aj ij]}
i =1 j=1
: 0 = i - w a j i j
j =1
i = 1,2,..., m
m m 2 : 0 = a - w ia w a j i j i =1 j =1
(XIII.9)
Za -R
2 a
The efficient part will have Za R or from (XIII.9), 0. So, from (XIII.8) and (XIII.9), we have that (XIII.8)
Za - R m a 0 = i - 2 w j ij , i = 1,2,..., m a j=1
258
Finance Theory
If
th th vij = i , j element of the inverse of the variance-covariance matrix [ij] , then [in an
In the special case where the unsystematic parts of the returns on the securities are uncorrelated
2 with each other (i.e., Cov(i , j) = 0, i j), then ij = 0, i j and ii = (i ) and (XIII.8') becomes
(XIII.11)
w ia = (
2 a i ) '2 Za -R ( i )
i = 1,2,...,m
From either (XIII.10) or (XIII.11), note that the ratio of risky assets
a a w i / w k = ( Vi j j)/( V k j j) or j=1 j=1 m m
i k 2 k i
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 i = 0, i = 1,..., m). However, it is also true that
a m wi wM =- a i a wk i =1 w k a
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
wR = 0
w ia + w a M = 1.
i =1
This can be done by requiring that the wia satisfy w a i (1 - i ) = 1 . (This is possible
i =1
provided that i 1 for all i; a technical point is that even if i 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
(i , ij) 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 , 2 M ) 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 1 be the fraction of that portfolio invested in the market; 2 be the fraction invested in the active portfolio; 1 - 1 - 2 be the fraction invested in the
because:
riskless
asset.
2 2 2 2 2 Zp = 1 ( ZM - R) + 2 ( Za - R) + R; p = 1 M + 2 a
cov( Za , ZM ) = 0.
becomes
2 2 2 2 {R + 1 ( Z M - R) + 2 ( Z a - R) + [ 2 p - 1 M - 2 a ]} a a 2 {1, 2 , w1 ,..., w m}
Max
where is a Lagrange multiplier. Note: he is not only allowed to pick 1 and 2 , but in
a ,..., a . w m I.e., he can select which active portfolio he wants. So the investor's choice addition, w1
260
Finance Theory
i , ij
as well as an estimate of
The first-order conditions are (XIII.12a) (XIII.12b) (XIII.12c) and (XIII.13) Since
Za
a wi
: 0 = Z M - R - 1 2 M 1 2 : 0 = Za - R - 2 a 2 2 2 2 2 :0 = 2 p - 1 M - 2 a .
a wi
: 0 = 2 { i - 2 w a j i j } i = 1,..., m
j =1
= i and
2] [ a a wi
= 2 wa j i j .
j =1
Za From (XIII.12b), 2 = 2 a
-R
(XIII.14)
Za - R m a , i = 1,..., m provided 2 0 0 = i - 2 w j i j 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.
261
Robert C. Merton
Thus, the optimal combination of risky assets will have the active portfolio in the amount
2 ( Z a - R))/ a ( Z a - R) ( Z M - R) + 2 2 a M 2 ( Z M - R)/ M ( Z M - R) ( Z a - R) + 2 2 a M
Compare the new optimal combination of risky assets (using forecasting) with the old optimal combination (with no forecasting): Note
2 2 ( Z M - R)/ M ( Z a - R)/ a * ( Z a - R) + ( - R) Z = R + ( - R) ( ( Z a - R) ( Z M - R) Z M Za Z M - R) + + 2 2 2 2 a a M M
262
Finance Theory
Final Portfolio Manager good; bad 1. Form Final Portfolio by mixing Active and Passive Portfolio 2. Monitor Performance of Active Portfolio 3. Monitor Performance of Passive (Market) Portfolio
Z , 2 p p
good; bad
good; bad
Product
Product
Active Portfolio
2 Za ; Za , a
Passive Portfolio
Z M ; ZM , 2 M
Active Portfolio Manager 1. Form Active Portfolio 2. How good are beta estimates? 3. How good are i , ij estimates?
Z2 , a
good; bad good; bad
Control 1. Is Za R a .
Passive Portfolio Manager 1. Form a Market Portfolio at minimum cost (good; bad) good; bad Control
(ZM R ) > 0 ?
2. Is a 0?
, ' i ij
Z , 2 M M
Group Analyst
e i good; 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 , 2 p )) 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 2 M 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. responsibilities of the boxed-in sub-units are now described. The various roles and
In Unit 1 (Product: Final Risky Asset Portfolio for Individual Investors) A. The Final Portfolio Manager
1.
(a)
(b)
He computes Z p and 2 p 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 " ( 2 / 1)" rule? (b) the
2 , (= 0) active portfolio: were Z a , a accurate? Was Z a - R - a ( Z M - R) > 0? a
Were the estimates of Z M and 2 M accurate? He either complains to or compliments the passive portfolio manager.
B.
(Final) Control
1.
Manager.
In Unit 2 (Product: Zero-Beta Risky Asset Portfolio representing superior Micro Forecasting) A. Active Portfolio Manager
1.
(a)
He receives estimates of i from the beta analyst and of ( i , ij) from the group analyst and forms the active portfolio according to the rule (XIII.10) or (XIII.11):
m
a/ a = wi wk v i j j / v k j j ; i, k = 1,..., m;
a (1 - ) = 1; a wi wa i M = 1- wi . i =1
(b)
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 zerobeta ( a = 0), is it because the i estimates supplied were inaccurate? (c) the
group analyst:
were the
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 - a ( Z M - R) > 0?
2. 3.
Does the active portfolio have a zero beta ( a = 0)? 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 ( ie) of the i and correct them for bias (historical) of each analyst to get (an unbiased) estimate
i ;
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
2) . ( i / i The larger (in absolute value) this ratio is on average, the more valuable
the analyst.
E.
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 2 M showing superior Macro Forecasting capabilities)
A.
1.
He must form a well-diversified portfolio which is as nearly perfectly correlated with the market portfolio as possible at minimum cost. He receives estimates Z M and 2 M from the Macro Analysts and then reports them to his customer (unit 1) and to his control management.
2.
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 2 M 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 MacroAnalysts are (good/poor), he registers the appropriate compliment or complaint
with them. He measures their performance by comparing a strategy of a variable position in the market depending on the estimated hold strategy.
ZM - R ratio and a buy-andM
267
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 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., i2 large), then independent of i , his security does not get into the portfolio. Similarly, if the aggregate error in the active portfolio estimates are large, then
Za - R 2 a
268
Finance Theory
will be small and it will have little weight in the final portfolio. In the limit as
Za - R 2 0 , a
the final portfolio is the market portfolio as it should be for someone with no forecasting capability.
269
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
x i is the random variable x i where ~ project with end-of-period random variable cash flow of Ii ~
x i) and i2 Var (~ x i) . If Vi is the average cash flow per dollar of investment. Let x i E(~
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 xi %i I ix I i2 i2 . with E( ) = and Var ( ) = Zi Zi Zi Zi Vi Vi V i2 firm's equity with the market, iM , is given by
270
Finance Theory
I i xi , iM Cov[ Z i , Z M ] = Cov ZM = Vi
I i i M iM Vi
In equilibrium, the equity of firm i will be priced so as to satisfy the Security Market Line:
Zi - R =
(XIV.1)
iM ( - R) = e iM 2 ZM M M Ii i iM = e ( ) Vi
where e ( Z M - R)/ M is the Market Price of Risk and it does not depend upon the decisions made by firm i . Substituting for Zi into (XIV.1), we have that rearranging terms, that (XIV.2)
Vi = Ii [ x - e i iM ] R i
Ii x i Vi
- R = e (
Ii i iM Vi
),
or
(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 (XIV.3)
Vi - Ii = Ii [ x - R - e i iM ] . R i
271
Robert C. Merton
So, for a given Ii , if [ x i - R - e i iM ] > 0 , take it; if [ x i - R - e i iM] = 0 , be indifferent; and if
[ x i - R - e i iM] < 0 , do not take it.
ip
Cov[ x % i , Z M ] i M iM i iM = = . Var( Z M ) 2 M M
Note: The beta of the equity of firm i (its "market beta") is given by
i
> > as Vi - Ii = 0 . <
iM = ( i iM )( I i ) = I i p . i 2 M Vi Vi M
Hence,
ip = i <
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 = ( ZM - R) where x is the expected cash p flow per dollar of investment in the project and 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 (XIV.4a) (XIV.4b) (XIV.4c)
p x i > R + i ( Z M - R) V i > I i Take the Project p x i = R + i ( Z M - R) V i = I i Indifference to the Project p x i < R + i ( Z M - R) V i < I i Do not take theProject
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 XIV.5) k i R + i ( Z M - R) .
p
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 i , 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, i 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 (i ) that matters in making the decision whether to invest or not,
and not the project's total risk ( i2) . 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
x i has the identical distribution from both firm's perspectives) should use "same" we mean that ~
the same cost of capital [given by (XIV.5)]. To see this, note by inspection of (XIV.5) that k i
x i and its joint distribution with the market. It does not depends only upon the distribution of ~ x i with other projects that the firm may have in addition depend upon the joint distribution of ~
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.
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.)
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 Let Let
P{X = x j} = f(x j) be the probability that X = x j , j = 1,2,3,... . P{Y = yk} = g( yk ) be the probability that Y = yk , k = 1,2,3,... . 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
275
Let Let
(XIV.8)
yk p( x j , yk ) f(x j)
(XIV.9)
j k
= yk ( p(x j, yk )) = yk g( yk ) = E(Y)
k j k
For purposes of this course, we will be dealing primarily with random variables describing an
(t) may be a random variable describing profits for date t. outcome as of a given date t . E.g., ~
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)} ~ Y(0) = y0 . E t -1{X (t)} will be the conditional expectation, conditional on knowing that
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) or more generally, (XIV.11')
~ ~ E t -k {E t- j[X (t)]} = E t-k [X(t)] for k j 0 ~ ~ E t -2 {E t -1[X (t)]} = E t -2 {X (t)}
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, (XIV.12)
~ d (t + 1) + ~ s (t + 1) ~ 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 (XIV.13)
s(t) = ~ 1 ~ 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 (T1): 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(T1) must be such as to satisfy (XIV.13). I.e., (XIV.14)
S(T - 1) = ~ 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(T2) must again satisfy (XIV.13). I.e., (XIV.15)
S(T - 2) = ~ ~ 1 ET-2 [ d (T - 1) + S(T - 1)] . [1 + k(T - 2)]
Substituting for S(T2) from (XIV.14) into (XIV.15), we have that (XIV.16)
~ ~ 1 1 [ d (T - 1)] + 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 (T2) it may be uncertain (i.e., a random variable). Noting that
~ ~ 1 d (T) [ d (T)] = { } ET-1 ET-1 [1 + k(T - 1)] [1 + k(T - 1)]
because
k(T1)
is not a
278
Finance Theory % (T ) % (T ) 1 d d % = = { [ ( )]} { [ ]} [ d T E T -2 E T -2 E T -1 E T -2 % (T -1)] E T -1 % (T -1) % (T -1) ] using the fundamental [1 + k [1 + k [1 + k 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)] ~ ~ d (T - 1) d (T) = ET - 2 { + }. ~ [1 + k(T - 2)] [1 + k(T - 2)][1 + k (T - 1)]
(XIV.17)
At time (T-3), we have that for markets to clear that S(T-3) must satisfy (XIV.13) or (XIV.18)
S(T - 3) = ~ 1 ~ 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
S(T - 3) = ~ ~ ~ 1 d (T - 1) d (T) { d (T 2) + + } ET-3 ~ ~ ~ [1 + k(T - 3)] [1 + k(T - 2)] [1 + k (T - 2)][1 + k(T - 1)] ~ ~ d (T - 2) d (T - 1) = 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)]
(XIV.19)
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 d (t) } E0 { t ~ t =1 [1 + k (s - 1)] ~ d (T) = E0 { ~ t} t =1 [1 + K (t) ]
T s =1
(XIV.20)
-1
Comparing (XIV.20) with the certainty case, in Section VI, there are some obvious similarities. Moreover, if
k(t) = r(t) ; E0 [ ~ d (t) = d(t) ~ d (t) [1 + k(t) ]
t
]=
As in the
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., (XIV.20') and (XIV.21)
~ n(0)d (t) 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) ). ~
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)}
~ ~
% d(t) E 0 [ d(t)] =
expected
1 [1+ (t) ]
t
E0 [
d(t)
t t =1 [1 + (t) ]
and
281
Robert C. Merton
V(0) =
(XIV.23)
n(0)d (t)
t t =1 [1 + (t) ]
Warning:
Case B:
Suppose that the {k (t)} are nonstochastic and constant, i.e., k(t) k . Then (XIV.20') and (XIV.21) can be rewritten as
(XIV.24) (XIV.25)
S(0) =
d (t) n(0)d(t) [1 + k ]t
t t =1 [1 + k ]
and .
V(0) =
t =1
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 (XIV.24') (XIV.25')
S(0) =
t t =1 [1 + K(t) ]
and .
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
~ 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 t
% t] E [ % ] E 0[W = [1 + K (t )]t or 0 W t W0 W0
~
1/ t
-1 = K (t ) .
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
for y equiv
1+ g 1+ k
(XIV.26)
d (0)[1+ g ] k -g = d (1) k -g
(XIV.27)
% (t ) % (t ) n % (t -1)d V (0) = n(0) S (0) because D % (1)] % (1)} = D(1) = E 0[n(0)d E 0{D % (1)] = n(0)d (1) = n(0) E 0[d
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
1 % +1) - [ n(t % +1)+ n(t)d(t % +1)} % +1)S(t % +1) - n(t)]S(t E t{ n(t [1+ k(t)] 1 % +1)}. % (t +1)+D(t % +1) -m(t % +1)S(t = E t{V [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 (XIV.29)
~ ~ ~ ~ (t + 1)~ R (t + 1) + m S(t + 1) = O(t + 1) + D(t + 1)
Substituting from (XIV.30) into (XIV.28), we have that (XIV.31a) (XIV.31b) (XIV.31c)
V(t) = V(t) = V(t) = ~ 1 ~ ~ E t {V (t + 1) + R(t + 1) - O(t + 1)} [1 + k(t)] 1 ~ ~ ~ E t {V (t + 1) + X (t + 1) - I (t + 1)} [1 + k(t)] ~ 1 ~ ~ E t {V(t + 1) + (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 (XIV.32a)
V(O) = E0 {
(XIV.32b)
(XIV.32c)
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) (XIV.33b) (XIV.33c) (XIV.33d)
~ ~ ~ ~ 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)] ~ ~ E t { V t} = k(t)V(t) - E t {R (t + 1) - O(t + 1)} ~ ~ E t { V t} = k(t)V(t) - E t {X(t + 1) - I (t + 1)} ~ ~ E t { V t} = k(t)V(t) - E t { (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 % (t) at time t. Since, by definition, the market at time t or the random variable cash flow X % (t), it must be that would be willing to exchange X ce(t) dollars for certain for the X (XIV.34)
V(0) =
t =1
(t)X (t)
[1 + r ]t
% (t)] . where (t) X ce(t)/ X (t) and X ( t ) E 0 [ X While, in general, one might expect (t) < 1, it need not be as for example in the CAPM if X(t) has a negative beta. Moreover, (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.
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" () . 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
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
(XV.1)
Vi = Ii [ x - e i iM ], 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,
x 3 as the random variable end-of-period cash flow of define I3 as the investment in firm #3 and ~
firm #3 per dollar of investment. If no changes in the investment plans of the firms occur as a result of the combination, then where x2 I3 I1 + I2 and ~ x3 ~ x1 + (1 - ) ~
289
Robert C. Merton
(XV.2)
3M V
Cov[~ x 3 , ZM ]
(XV.5)
V 3=
I3 { [ x1 - e 1 1M ] + (1- )[ x 2 - e 2 2M ]} R
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 and the random variable end-of-period cash flow is IP = Ii , and total firm end-of-period cash
i =1 m
(XV.7)
= i x i
i =1
i =1 m
where i I i / I P and
i = 1 .
i=1
and
(XV.9)
Var( x P ) 2 P=
i j Cov( xi , x j ) .
i=1 j=1
m m
(XV.10)
= i iM i M .
i =1
291
(XV.11)
= =
IP m [ i ( x i - e i iM )] R i =1
Vi
i =1
where Vi = Ii ( x i - e i iM )/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
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:
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
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) (2) legal restrictions cash position (6) (7) profit rates access to capital markets (tradeability of equity) control of the firm tax position of shareholders corporate tax liabilities
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.
At time zero, buy % of firm #1 and sell short % of firm #2. Since V2(0) > V1(0), my total
297
Robert C. Merton position is at this point: (a) (b) (c) cash = [V2(0) V1(0)] > 0 long n1(0) shares of firm #1 short n2(0) shares of firm #2
Suppose that the portfolio policy is pursued of always maintaining a long position in firm #1 equal to % of its value and a short position in firm #2 equal to % 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) = n1(0) and N2(0) = n2(0) and
(XVI.1a) (XVI.1b)
Let C(t) = total cash flow from this portfolio strategy at time t. Then:
C (0) = n 2(0) S 2(0) - n1(0) S 1(0) = [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 < T1 , we have that
(XVI.2)
Finance Theory
Where D1 and D2 are the total dividends paid by the two firms respectively.
C (t + 1) = D1(t + 1) - D 2(t + 1)+ m 2(t + 1) S 2(t + 1)- m1(t + 1) S 1(t + 1) = {[ D1(t + 1) - m1(t + 1) S 1(t + 1)]-[ D 2(t + 1) - m 2(t + 1) S 2(t + 1)]}
X 1(t + 1) - I 1(t + 1)
Y 1(t + 1)
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 ) = {[ D1(T ) - m1(T ) S 1(T )] - [ D 2(T ) - m 2(T ) S 2(T )]} +V1(T) sale of shares long -V2(T) purchase of shares short + C(0)[1+R(T)] cash and interest on maturity of bonds.
T
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 = [V 2(0) - V 1(0)][1 + R(T )]T >0 if V 2(0) > V 1(0)
(XVI.8)
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 = T1 there is an investor who is considering buying % of firm #2 for $V2(T1). Suppose instead he bought % of firm #1 and did the following: at time T, he will receive D1(T) in dividends. Suppose he sells (ex-dividend) $[D2(T) D1(T)] of his stock for cash if D2(T) D1(T), or if D1(T) > D2(T) , then he buys $[D1(T) D2(T)] of the stock of firm #1. At this point, he will then have ${D1(T) + [D2(T) D1(T)]} = $D2(T) in cash and ${[V1(T) m1(T)S1(T)] + [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, [m1(T)S1(T) D1(T) + D2(T)] = m2(T)S2(T). Therefore, our investor would have: $ D2(T) , in cash, and $ [V1(T) m2(T)S2(T)] = $ [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 % of firm #2. If V1(T1) < V2(T1), then every investor (who prefers more to less) would be better off to buy firm #1 instead of firm #2. Hence, unless V1(T1) = V2(T1), 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(T1) = V2(T1) . Suppose at some date , V1() = V2() , then, by the same argument (with "" replacing "T"), we have that V1(1) = V2(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.
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
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:
(XVI.10)
V (0) =
(1- ) (0) r - r *
and = fraction of profits allocated to new investment. From (XIV.27), we have that:
(XVI.11)
V (0) =
D(1) 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) = X(t) , then D(t) m(t)S(t) = [1]X(t). Let r = fraction of current earnings retained (i.e., D(t) = [1r]X(t). Let e = the amount of external financing required expressed as a fraction of current earnings. It follows that [1r]X(t) eX(t) = [1]X(t) or e = r.
X (1) = (0), so, D(1) = [1 - r ] (0).
(XVI.12)
= V (0) =
[1- r ] (0) r - g
or
(XVI.12')
g=
r * (1- r ) r - e (1- ) 1-
303
Robert C. Merton Note: Unless e = 0 (i.e., no external financing), the rate of growth of dividends, g , is less
* than the rate of growth of profits, r . Further, even if the firm pays out all of its current earnings in
dividends, i.e., r = 0 , dividends and price per share will grow over time, i.e., g =
(r * - r ) . (1- )
Example: three firms all with (0) = $100 and identical investment policies:
Firm
(0)
III $100 .10 .20 .40 .20 .20 $3,000 $40 1,000 $3.00
r
r*
r e
V(0) I(1) n(0) S(0)
Firm I: Finances all its investment internally through retained earnings, i.e., r = = .40 and
e = 0 .
From (XVI.10), V (0) = (1- ) (0) (.6)(100) 60 = = = $3,000 * .10-.4(.2) .02 r - r I (1) = 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 the accounting identity. Dividends per share, d (1) =
have that
X (2) = X (1)[1 + r * ] = 100(1.08) = $108. Therefore, the value of the firm next period will be:
V (1) = X (2)(1- ) $108(1-.4) = = $3, 240 . Since no new shares are issued, n(1) = n(0) = r - r .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:
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 S (1) = V (1) - m(1) S (1) = n(0) 3240-40 or S(1) = $3.20 and m(1) = 12.5. The return to the 1000
305
.10+3.20-3.00 d (1) + S (1) - S (0) = = 10% = r since D(1) = X(1) = $100 S (0) 3.00
100 D(1) = = .10 . Note that the larger dividend of Firm II is offset by a smaller capital n(0) 1000
r * (1- r ) er (.4)(.2)(1-0) .4(.1) S (1) - S (0) = = 0.0667 = rate of growth of price per share = . S (0) 1- 1- .6 .6
Firm III: Uses a mix of one-half internal and one-half external financing. Hence, m(1)S(1) =
.5I(1) = $20 and again, S (1) D(1) d (1) g = = V (1) - m(1) S (1) 3240-20 = = $3.22 per share n(0) 1000 = 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 = = $0.08 and = = 10% = r n(0) S (0) 3.00 (1- r ) er (.8)(.8) (.2)(.1) r* = = .07334 (1- ) 1- .6 .6 S (1) - S (0) = rate of growth stock price . S (0) = d (1) + g = current dividend yield + growth . S (0)
Note: r = 10% =
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) = (1r)X(t) + gV(t1) , because (1r)X(t) is the amount of earnings not retained, and g, the rate of growth of dividends, is equal to the rate of growth of price per share. We have that V (t -1) =
and from (XVI.12'), that g =
(1- ) X (t ) r - r *
D(t ) + G (t ) = X (t ){(1- r ) + = X (t )
*
er (1- ) } (r - r * )(1- )
r-r rX (t ) = {1- r - e} = r - r *
X (1) (0) r - r * D(1) = = , and from (XVI.11), = r - g. Note: From (XVI.10), V (0) V (0) 1- 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
P = m[ D + E ] 3
where E = earnings; D = dividends; change in retained earnings = RE; P = price of stock; m = constant. Because E = D + RE , we also have
P = m [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.,
(0) and the "target" payout ratio, D / (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 "longrun" earnings trend. Hence,
D / E > D / (0) . Similarly, the market, recognizing that price is
dependent on "long-run" earnings, will not bid down the price. Hence, P / E > P / (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 / (0) and P / E < (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 crosssection 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
% (t )}) . In their dividend paper, policy affects the required expected return on the firm (i.e., the {k
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 ) = 0 + [E ( Z M ) - 0] j + 1[ j - M ] / M
where j = current dividend yield on security j; M = current dividend yield on the market; 0 = expected return on a "zero-beta" portfolio; and 1 is the "expected return" on the dividend factor. Possibilities: (i) The classical security market line relationship of the CAPM would predict 0 = R, the riskless rate 1 = 0 . Thus, if they could not reject 0 = R and 1 = 0 , we cannot reject the CAPM and we cannot reject the hypothesis that dividend policy "does not matter." (ii) If 1 0 , then the data suggest that dividend policy does differentially affect returns. Further, if 1 > 0 , then this would imply that investors prefer lowdividend yielding stocks. If 1 < 0 , then this would imply that investors prefer high-dividend yielding stocks. Their findings were that while they could reject the hypothesis that 0 = R, they could not reject the hypothesis that 1 = 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 1 > 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
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 current stock price, P(0), he can estimate an expected return on the stock, Z , which is Z =
P(1) . 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 bargainyour 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
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,... St +1 = St + (-1)
t +1
22.72
I.e., the return per dollar from investing in the shares from time 0 to time 1,
Z1 = Z2 = D1 + S1 100 + 363.64 = = 1.20, and from time 1 to time 2, 386.36 S0 D2 + S2 50 + 386.36 = = 1.20, and so forth. 363.4 S1
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
' S0 = ' S2
50 100 D0 D ' = = $250; S1 = 1 = = $500, r r .2 .2 50 100 D2 D3 ' = = = $250; S3 = = = $500,... r .2 r .2 t ' St' +1 = St + (-1) 250
or The return per dollar from investing in the shares from time 0 to time 1 under this pricing is
' = Z1 ' 100 + 500 D1 + S1 = = 2.4 or 140% ' 250 S0
' Z2 =
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 (Arithmetic Average) Standard Deviation (Annual)
Years
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:
Years
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 P=.75 1.56% 4.13% 37.61% -22.02% 1.47% $9,146,722 NYSE Stocks 0.85% 5.89% 38.55% -29.12% 0.68% $67,527
Per Month Average Rate of Return Standard Deviation Highest Return Lowest Return Average Compound Return Growth of $1,000 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 Average Return 11.98% 13.11% 10.86% 12.25% 13.99% 14.53% 7.31% S&P Average Return 11.63% 13.38% 9.90% 13.50% 13.26% 12.34% 7.52%
Years Total: 10 Years: 5 Years: 1953-1972 1953-1962 1963-1972 1953-1957 1958-1962 1963-1967 1968-1972
Avg. Excess Return 7.57% 10.00% 5.15% 9.45% 10.52% 9.70% 0.71%
Avg. Excess Return 7.22% 10.25% 4.19% 10.70% 9.79% 7.50% 0.92%
Jensen Performance of Mutual Funds Study 1945-1964 Testing 115 Funds ability to Forecast (relative to Security Market Line):
Z j (t) = R(t) + j[ ZM (t) - R(t)] + j (t) ~ (t) = R(t) + [ ~ (t) - R(t)] + (t) + ~ (t) j j Zj j ZM
321
Robert C. Merton 115 Funds Studied. Returns net of all costs including management fees. 76 funds had measured j < 0 Average = .011 = 1.1%
The statistical significance of the positive j were no more than would have been expected by chance when the true j = average .
Using Returns Gross of Management Fees 55 funds had measured j < 0 Average = .004 = 0.4%
Statistical significance of the positive j were no more than would have been expected by change when the true 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 j in the past tend to have positive j in the future? Jensen & Black studied the 115 funds for the years 1955-1964 computing the realized 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
1 2 3 4
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.,
p (1.085) M p = 1.085 pM p or pM 1 = .9216 1.085
Further, on the whole, funds tended to keep about the same level of p or p 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 nonsystematic 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 1 - 8%, 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 + 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" ( < 1) stocks were higher than predicted by the Capital Asset Pricing Model and the returns on "high beta" ( > 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- > R where Z0- is the expected return on the minimum-variance, zero-beta portfolio constructed from stocks. The specification that they fit was Z j - R = j( Z M - R) + ( j)( Z 0- - R) , where (1) = 0 and
d < 0. d
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