Dr.

Maddah ENMG 624 Financial Eng’g I 01/14/11

Chapter 5 Applied Interest Rate Analysis

• Investment and Optimization

¾ Several investment problems can be solved via optimization.
¾ Examples of these problems are capital budgeting, bond
portfolio structuring, management of dynamic investments,
and firm valuation.
¾ An optimization problem is generally concerned with finding
the optimal solution, x*∈ Rn, that maximizes a real-valued
function, f : Rn →R, over a domain of Rn defined by a set of
constraints.
¾ That is,
max f (x)
subject to
g i (x) ≤ bi , i = 1, 2,..., m .
¾ If f(x) and gi(x) are linear then the optimization problem is
termed a linear program.
¾ In general, an optimization problem is called a mathematical
program.

1
• Capital Budgeting
¾ Capital budgeting involves allocating a fixed budget among a
set of investments or projects.
¾ Usually, there are no well-established markets for these
project or investments.
¾ They are lumpy requiring discrete lumps of cash (as opposed
to securities which can be traded in any number of shares).
¾ One type of capital budgeting problems is that of selecting
from a set of independent projects. That is, any subset of
projects can be selected if it is within the available budget.
¾ Consider a set of m projects. Let ci and bi be the initial cost
and the present value of project i. Suppose that a total budget
of C dollars is available.
¾ Define the decision variables as xi, i = 1,…, m, where xi = 1
if project i is selected, xi = 0, otherwise.
¾ Then, the problem is solved with the following integer-linear
program (ILP)
m
max ∑ bi x i
i =1
m
subject to ∑c x
i =1
i i ≤C

x i = 0,1, i = 1,… , m .
¾ The optimal solution to the above ILP can be found using an
optimization software (e.g. AMPL, www.ampl.com).

2
 ¾ Microsoft Excel has also a solver module which can handle
small ILP problems.

• Approximate (Heuristic) Capital Budgeting Solutions

¾ Approximate solutions to the capital budgeting ILP is
obtained by ranking projects according to benefit-cost ratio.
¾ This approximation method assumes that each project
involves an initial outlay of cash (first cost) followed by a
series of benefits.
¾ The benefit-cost ratio for project i is the ratio of the present
value of benefits, bi + ci to the initial cost, ci .
¾ The approximate solution is obtained by selecting the project
with the highest benefit-cost ratio, then the project with the
second highest ratio, and so on, until the budget is exhausted.
(Example 5.1)

• Capital Budgeting with Interdependent Projects

¾ In some situations, the projects are interdependent.
¾ E.g., assume that there are m goals. Goal i can be achieved
by one of ni project.
¾ Define the decision variable as xij, with xij = 1 if goal i is
chosen and implemented with project j, and xij = 0, otherwise.
¾ Then assuming a budget limit of C with bij and cij denoting
the present value and the first cost of project ij, the capital
budgeting problem can be stated as follows.

3
 m ni
max ∑∑ bij x ij
i =1 j =1
m ni
subject to ∑∑ c
i =1 j =1
ij x ij ≤ C

ni

∑xj =1
ij ≤ 1, i = 1, …, m

x ij = 0,1
(Example 5.2)

• Issues with capital budgeting

¾ The hard budget constraint is inconsistent with the
assumption than one can borrow unlimited funds.
¾ However, in practice, borrowing limits usually apply (e.g.
bank credit line, organizational budget).
¾ It is instructive to solve the budget problem with different
budget values to measure the sensitivity to the budget level.

• Optimal Portfolios
¾ The term optimal portfolio usually refers to the construction
of a portfolio of financial securities.
¾ A simple optimal portfolio problem is the cash matching
problem.
¾ This problem involves structuring a bond portfolio to meet a
series of future obligations from coupon payment and
redemption (face) values.

4
 ¾ Let y = (y1, y2, …, yn) be the cash flow stream representing
obligations and let cj = (c1j, c2j, …, cnj) be the cash flow
stream associated with bond j, j = 1, …, m.
¾ Here yi and cij represent cash flows at time period i.
¾ (There are n time periods and m bonds.)
¾ Define also pj as the price of bond j.
¾ The decision variable is xj the number of shares of bond j in
the portfolio.
¾ The cash matching problem can be solved with the following
linear program
m
min ∑p x
j =1
j j

m
subject to ∑c x
j =1
ij j ≥ yi , i = 1,… , n

x j ≥ 0, j = 1,… , m
(Example 5.3)

• Issues with cash matching

¾ One issue with cash matching is that it implicitly assumes
that surplus cash flows in a given time period are not
reinvested (as if surplus is thrown away).
¾ In practice, surplus cash flows are reinvested (in other bonds
perhaps).
¾ However, the reinvestment possibility can be accounted for
through introducing “artificial” bonds (see text for details).

5
• Dynamic cash flow process
¾ Many investment problems are not one time decisions. They
require ongoing dynamic management.
¾ A dynamic cash flow process can be described by a cash flow
stream x = (x0, …, xt, …, xn), where xt, depends on the
management actions at times 0, 1,…, t.
¾ A dynamic cash process can be represented by a graph (tree).
¾ In this graph, nodes represent different possible states of the
process. Each node is defined at a given time period (stage).
¾ Arcs connect nodes between one time and the next.

• Example (investing in an oil well)

¾ A binomial tree is a tree such that exactly two branches leave
each node (excluding end nodes).
¾ E.g., in the oil well situation where in each period the
management must decide whether to pump 10% of the oil
reserve starting with a reserve of 10 M barrels.
¾ The state of the system (shown above the nodes) is the
number of barrels of oil reserve and whether a new crew of
workers has been hired.

6
¾ A binomial lattice is a binomial tree where “intermediate”
nodes at each stage can be combined.
¾ E.g., the graph below is a binomial lattice representing the
oil well situation where a crew can be hired at no cost.

¾ Cash flows resulting from a given decision are shown above

the arc representing the decision.
¾ E.g., suppose that the cost of hiring a new crew is $100 K
(think of this as a recruitment cost). Suppose also that profit
from oil production is $5/barrel.

7
 ¾ There could be a final reward or a salvage value associated
with a process termination. This is placed on the graph to the
right of the final nodes.

• Dynamic Programming
¾ Once the tree of a cash flow process has been developed, the
“optimal path” can be determined by enumerating all
possible paths.
¾ However, this process is computationally inefficient due to
the curse of dimensionality.
¾ Dynamic programming (DP) is a computational procedure to
search for the optimal path efficiently.
¾ DP recursively finds the optimal path from each node in the
graph to termination.
¾ E.g., the optimal decision and present value at time n−1, at
every node, are determined as follows.

8
¾ At time n−2, the optimal present value and decisions are
determined as follows.

¾ In general, Denote the value associated with node i at time k

by Vki .
¾ Suppose that a time k the set of decisions is A.

9
 a
¾ Let cki be the cash flow associated with decision a ∈A at
node i and time k.
¾ Then, the DP optimality equation is as follows
Vki = max(ckia + d kVk +1,a (i ) ) ,
a∈A

where a(i) is the node at time k+1 that the process moves to
if decision a is taken, and dk is the discount factor.
¾ DP starts with the terminal values, Vni, which are usually
known. It iterates until the optimal value Vk0, and decisions
(path), are found.
(Example s 5.4, 5.5)

• Valuation of a firm
¾ Different cash flow streams can be used to evaluate the worth
of a firm. E.g., dividends, net earnings.
¾ Different cash flows may lead to different valuations.
¾ This kind of analysis also assumes deterministic cash flows
which can be problematic.

• The dividend discount model

¾ In this model, the value of one share of a firm stock, V0, is
based on the dividend it pays.
¾ If the stock pays a dividend Dk in year k, and the interest rate
is r, then

10
 ∞
V0 = ∑ Dk /(1 + r ) k .
k =1

¾ A popular dividend model is the constant-growth dividend

model, which assumes that Dk+1 = (1+g)Dk .
¾ Then, assuming g < r, Dk+1 = (1+g)kD0 , where D0 is the
current dividend and
k −1
⎛ 1+ g ⎞ ⎛ 1+ g ⎞ ∞ ⎛ 1+ g ⎞
∞ k

V0 = ∑ D0 ⎜ ⎟ = D0⎜ ⎟∑⎜ ⎟
k =1 ⎝ 1+ r ⎠ ⎝ 1 + r ⎠ k =1 ⎝ 1 + r ⎠
⎛ 1+ g ⎞ ∞ ⎛ 1+ g ⎞ ⎛ 1+ g ⎞
k
1
= D0 ⎜ ⎟∑⎜ ⎟ = D0 ⎜ ⎟ .
⎝ 1 + r ⎠ k =0 ⎝ 1 + r ⎠ ⎝ 1 + r ⎠ 1 − (1 + g ) /(1 + r )
¾ Therefore,
D0 (1 + g ) 1
V0 = .
r−g

• The free cash flow approach

¾ This approach values the firm on the basis of net earnings
from the cash flow stream with the maximum present value
(maximum over different investing strategies).
¾ Suppose the firm earns Yn each year and decides to invest uYn
in growth. The growth rate is a function g(u).
¾ That is, Yn+1 = (1+g(u))Yn .
¾ Then, the annual capital Cn satisfies, Cn+1 = (1−α)Cn + uYn ,
where α is the depreciation rate.

1
This is called Gordon formula.

11
 ¾ It can be shown that
Yn (u)= (1+g(u))nY0 ,
⎧ −(1 − α ) n + (1 + g (u ))n ⎫ uY 0 (1 + g (u )) n
C n (u ) = (1 − α ) C 0 + uY 0 ⎨
n
⎬≈ .
⎩ g (u ) + α ⎭ g (u ) + α

¾ Then the free cash flow, at a tax rate of T per year is

FCFn (u ) =Y n −T (Y n − αC n (u )) − uY n (u )
= (1 −T )Y n + αTC n (u ) − uY n (u ) .
¾ Upon simplification,
⎡ αu ⎤
FCFn (u ) ≅ ⎢1 −T +T − u ⎥ (1 + g (u )) nY 0 .
⎣ g (u ) + α ⎦
¾ Then, the value of the firm at a growth investment rate u is
∞
FCF (u ) ⎡ αu ⎤ Y0 (1 + g (u ))
PV (u ) = ∑ = ⎢1 − T + T − u ⎥ r − g (u ) ,
n =1 (1 + r )
n
⎣ g (u ) + α ⎦
assuming r > g(u) , and following the same steps as before.
¾ Finally, the value of the firm according to the free cash flow
approach is
PV = max PV (u ) .
u ∈(0,1)

Examples 5.6 and 5.8

12