Department of Economics

University of Maryland

ECONOMICS 600
Patricia Tovar August, 2004
Office: Tydings 4118B
Email: [email protected]
Lecture 1

1. Differential Equations

1.1. Basic Concepts for Univariate Equations

We use differential equations to model situations which treat time as a continuous

variable.

An ordinary differential equation is an expression which describes a relationship

between a function of one variable and its derivatives.

Definition: An ordinary differential equation is an equation of the form:

x m (t ) = F [t , x(t ), x1 (t ), x 2 (t ), ... , x m−1 (t ); α ] (1)

where
dx(t ) d 2 x(t ) d m x(t )
x1 (t ) = , x 2 (t ) = , ... , x m (t ) = ,
dt dt 2 dt m

α ∈ Ω ⊆ ℜ p is a vector of parameters, and

F is a function ℜ m +1+ p → ℜ

Note: The solution is a function x(t ) that, together with its derivatives, satisfies (1) for
given values of the parameters.

• We call (1) an ordinary differential equation because x is a function of one argument,

t, only. If it was a function of more than one variable, partial derivatives would appear
and we would have a partial differential equation. We will only study ordinary
differential equations.

• A differential equation is linear if F is linear in x(t ) and its derivatives.

• A differential equation is autonomous if t does not appear as an independent

argument of F, but enters only through x.

1
• The order of a differential equation is the order of the highest derivative of x that
appears in it. Thus, (1) is an m-th order differential equation.

Note: Any differential equation can be reduced to a first-order differential equation

system by introducing additional variables. For example, given the third-order differential
equation

x 3 (t ) = ax 2 (t ) + bx1 (t ) + x(t )

Define:
y (t ) = x1 (t )
z (t ) = x 2 (t )

Then:
y1 (t ) = x 2 (t ) = z (t )
z 1 (t ) = x 3 (t )

and we have the first-order system:

 z 1 (t ) 
   az (t ) + by (t ) + cx(t ) 
 y1 (t )  =  z (t ) 
   

 x (t )   y (t )
1 

 

Definition: A particular solution to (1) is a differentiable function x(t) that satisfies (1)
for some subinterval I0 of the domain of definition of t, I. The set of all solutions is called
the general solution, x g (t ) .

To see that the solution to a differential equation is generally not unique, consider the
following example:

x1 (t ) = 2 x(t ) (2)

One solution to (2) is x(t ) = e 2t . But for any constant c, x(t ) = ce 2t is also a solution.

The non-uniqueness problem can be overcome by augmenting the differential equation

with a boundary condition of the form:

x(t 0 ) = x0

Definition: A boundary value problem is defined by a differential equation

2
 x1 (t ) = f [t , x(t )] (3)

and a boundary condition

x(t 0 ) = x0 , ( x0 , t 0 ) ∈ X × I (4)

The following theorem says that, under certain conditions, every boundary value problem
has a unique solution.

Theorem: (The Fundamental Existence-Uniqueness Theorem) Let F be C1 in some

neighborhood of (x0, t0). Then in some subinterval I0 of I containing t0 there is a unique
solution x(t) to the boundary value problem given by (3) and (4).

Example 1: If F is not C1 in some neighborhood of (x0, t0), the solution may not be
unique. Consider the boundary value problem:

x1 (t ) = 3 x(t ) 2 / 3 , x, t ∈ ℜ (5)
x(0) = 0 (6)

It is easily verified that both

x(t ) = t 3
and x(t ) = 0

are solutions to (5) and (6). Note that f ( x) = 3 x(t ) 2 / 3 is not differentiable at x = 0 .

Example 2: The solution may not exist globally. Consider the boundary value problem:

x1 (t ) = x(t ) 2 , x, t ∈ ℜ (7)
x(0) = 1 (8)

It is easily verified that

1
x(t ) =
1− t

satisfies (7) and (8), but it is only defined for t ∈ (−∞,1) .

We may be often interested in analyzing the long-run properties of a differential equation,

in particular, the properties of its steady state (long-run equilibrium) and whether the
solution eventually converges to the steady state (stability). We can do such analysis
without having to find an explicit solution of the differential equation.

3
Steady States and Stability:

Definition: Consider the autonomous differential equation

x1 (t ) = f [ x(t )], f :X ⊆ℜ→ℜ

A steady state is a point x ∈ X such that f ( x ) = 0 . (See DLF pag. 410)

We can use phase diagrams to illustrate this definition:

• A steady state may not exist:

• The steady state need not be unique:

4
• Non-isolated steady states:

Consider an equation that is initially at rest at an equilibrium point x , and suppose that
some shock causes a deviation from x . We may want to know if the equation will return
to the steady state, or at least remain close to it, or if it will get farther and farther away
from it over time.

Definition: Let x be an isolated (or locally unique) steady state of the autonomous
differential equation

x1 (t ) = f [ x(t )], f :X ⊆ℜ→ℜ

We say that x is stable if for any ε > 0, ∃δ ∈ (0, ε ] such that

|| x(t 0 ) − x || < δ ⇒ || x(t ) − x || < ε ∀t ≥ t 0

That is, any solution x(t ) that at some point enters a ball of radius δ around x remains
within a ball of (possibly larger) radius ε .

Definition: A steady state is asymptotically stable if it is stable (as in the previous

definition) AND δ can be chosen in such a way that any solution that satisfies
|| x(t 0 ) − x || < δ for some t0, will also satisfy lim t →∞ x(t ) = x .

That is, any solution that gets sufficiently close to x not only remains nearby but also
converges to x as t → ∞ .

5
x1 (t ) > 0 implies that x(t ) is increasing (the “arrows of motion” point to the right), and
x1 (t ) < 0 implies that x(t ) is decreasing (the “arrows of motion” point to the left). Then:
x1, x3 : locally asymptotically stable
x 2 : unstable

x1 : globally asymptotically stable.

We can conclude that if for all x in some neighborhood of a steady state x :

x(t ) < x ⇒ x1 (t ) > 0 and x(t ) > x ⇒ x1 (t ) < 0 then x is asymptotically stable
x(t ) < x ⇒ x1 (t ) < 0 and x(t ) > x ⇒ x1 (t ) > 0 then x is unstable.

Therefore, we can determine the stability property of a steady state by checking the sign
of the derivative of f[x(t)] at x :

xi is (locally) asymptotically stable if f '[ x(t )]| x (t )= xi < 0

xi is unstable if f '[ x(t )]| x (t )= xi > 0

What happens if f '[ x(t )]| x (t )= xi = 0 ?

6
x1 : locally asymptotically stable
x 2 : unstable

This suggests that we can study the stability properties of a differential equation from the
sign of f '[ x(t )]| x (t )= xi as long as f '[ x(t )]| x (t )= xi ≠ 0 .

Definition: Let x be a steady state of the autonomous differential equation:

x1 (t ) = f [ x(t )], f : X ⊆ ℜ → ℜ, f ∈ C 1 (9)

We say that x is a hyperbolic equilibrium if f '[ x(t )]|x (t )=x ≠ 0 .

The previous analysis suggests that we can study the stability properties of a nonlinear
differential equation by linearizing it, as long as the equilibrium is hyperbolic.

Theorem (Grobman-Hartman): If x is a hyperbolic equilibrium of the autonomous

differential equation (9), then there is a neighborhood U of x such that (9) is
topologically equivalent to the linear equation

x1 (t ) = f '[ x(t )]|x (t )= x ⋅ [ x(t ) − x ] (10)

in U.

Note that (10) is a first-order Taylor series approximation of f around x :

f [ x(t )] = f ( x ) + f '[ x(t )]|x (t )= x ⋅ [ x(t ) − x ] + R[ x(t ) − x ]

where lim x (t )→ x {R[ x(t ) − x ] / || x(t ) − x ||} → 0 (and since f ( x ) = 0 ).

The theorem says that near a hyperbolic equilibrium x , the nonlinear differential equation
(9) has the same qualitative structure as the linearized differential equation (10). In
particular, if x is (locally) asymptotically stable for (10), then it is locally asymptotically
stable for (9), and if it is unstable for (10), then it is unstable for (9).

7
Application: The Solow Growth Model

Output (Y) is produced using capital (K) and labor (L) according. The production function
takes the form:

Y (t ) = F [ K (t ), L(t )] (11)

F is C1, it has constant returns to scale and positive and diminishing marginal products
with respect to each input.

Using the constant returns to scale property, we can rewrite output as:

Y (t ) = F [ K (t ), L(t )] = L(t ) F [ K (t ) / L(t ), 1] = L(t ) f [k (t )]

K (t )
where k (t ) =
L(t )

Then
Y (t )
y (t ) = = f [k (t )] (12)
L(t )

That is, we can write output per unit of labor as a function of capital per unit of labor. We
also assume that the so-called Inada conditions hold:

f (0) = 0, lim k (t )→∞ f [k (t )] = ∞, f ' (0) = ∞, lim k (t )→∞ f '[k (t )] = 0 (13)

The economy is closed, thus savings equals investment, and a constant fraction s of
output is saved. Assume that there is no depreciation of capital. Then:

K 1 (t ) = I (t ) = sF [ K (t ), L(t )], s ∈ [0, 1] (14)

where I denotes investment. Assume that labor grows at a constant, exogenous rate n:

L(t ) = L(0)e nt = L0 e nt (15)

Then:
L(t )
K (t ) = K (t ) = k (t ) L0 e nt (16)
L(t )

and thus
K 1 (t ) = k 1 (t ) L0 e nt + k (t )nL0 e nt (17)

8
Using (17) and (14) we get:

k 1 (t ) = sf [k (t )] − nk (t ) (18)

We can interpret (18) using the following graph:

The steady state:

k s
k 1 (t ) = 0 ⇔ sf (k ) = nk ⇔ = (19)
f (k ) n

The capital-output ratio: K / Y = k / f (k ) is constant, and the capital stock and output in
steady state grow at the rate of growth of the population, n.

Stability:

k 1 (t ) > 0 for k (t ) < k (since sf [k (t )] > nk (t ) ) and

k 1 (t ) < 0 for k (t ) > k (since sf [k (t )] < nk (t ) ) and

Therefore, the steady state is globally asymptotically stable.

Solving Autonomous Differential Equations:

Consider the following first-order autonomous linear differential equation:

x1 (t ) = ax (t ) + b (20)

9
All solutions to (20) can be written as:
x g (t ) = x c (t ) + x p (t ) (21)
123 1 2 3 123
general solution complementary function particular solution

The complementary function x c (t ) is the general solution to the homogenous equation

associated with (20):

x1 (t ) = ax (t ) (22)

and x p (t ) is any particular solution to the full non-homogeneous equation (20).

We can use the method of separation of variables to solve (22). Rewrite (22) as:

dx(t )
= ax(t )
dt

Then:
dx(t ) = ax(t )dt

and
dx(t )
= adt (23)
x(t )

Integrating both sides of (23) we obtain:

1
∫ x(t )dx(t ) = ∫ adt
ln x(t ) = at + c1

e ln x (t ) = e at e c1

x c (t ) = ce at , c = e c1 (24)

Now, to verify that (21) holds, we want to show that if x p (t ) is any solution to (20), then
ce at + x p (t ) also satisfies (20):

d dx p (t )
[ce at + x p (t )] = ace at +
dt dt
= ace + ax p (t ) + b
at

10
 = a[ce at + x p (t )] + b
Thus, if we add ce at to any solution, we get another solution. Moreover, we get all
solutions in that way:

Let x p (t ) and x g (t ) be two arbitrary solutions to (20).Then

dx g (t )
= ax g (t ) + b (25)
dt

dx p (t )
= ax p (t ) + b (26)
dt

Subtracting (26) from (25) and defining y (t ) = x g (t ) − x p (t ) , we obtain:

dy (t )
= ay (t )
dt

with general solution:

y (t ) = ce at

Therefore, the difference between any two arbitrary solutions has the form ce at , and thus
we can get all solutions by adding ce at to x p (t ) .

To solve (20), we still need to find x p (t ) . The simplest alternative is usually a steady
state solution:

x1 (t ) = 0 ⇔ ax (t ) + b = 0

b
x p (t ) = − , if a ≠ 0
a

Therefore:

b
x g (t ) = ce at − , if a ≠ 0 (27)
a

and
x g (t ) = bt + c, if a = 0 (28)

(Show the last equality)

11
We can pin down the arbitrary constant by means of a boundary condition. Suppose we
have:

x(0) = x0 (29)

Then, if a ≠ 0 ,

b
x(0) = c −
a

b
c = x0 +
a

and the solution to the boundary value problem given by (20) and (29) is:

 b b
x g (t ) =  x0 + e at − , if a ≠ 0 (30)
 a a

and
x g (t ) = bt + x0 , if a = 0 (31)

What are the stability properties of (30)?

If a > 0 , then e at → ∞ as t → ∞ and x(t ) explodes unless x0 = −b / a

If a < 0 , then e at → 0 as t → ∞ and x(t ) is asymptotically stable (it converges
to the steady state x = −b / a , independently of the initial position).

Solving Non-Autonomous Differential Equations:

Consider the first-order non-autonomous linear differential equation:

x1 (t ) = a (t ) x(t ) + b(t ) (32)

where the coefficients a and b are functions of time.

Rewrite (32) as:

x1 (t ) − a (t ) x(t ) = b(t ) (33)

12
 t

Consider the function e −α (t ) , with α (t ) = a ( s )ds . Multiplying both sides of (33) by ∫

0
−α (t )
e :

x1 (t )e −α ( t ) − a(t ) x(t )e −α ( t ) = b(t )e −α (t ) (34)

Note that the left hand side of (34) is the derivative of x(t )e −α ( t ) . Therefore, (34) can be
rewritten as:

d
dt
( )
x(t )e −α (t ) = b(t )e −α (t ) (35)

The expression e −α (t ) that made the left hand side of (34) the exact derivative of a
function is called an integrating factor.

We can use (35) to derive two (equivalent) forms of the general solution of equation (32):
the backward solution and the forward solution.

• The backward solution is obtained by integrating (35) backward between 0 and s:

∫ ( )
s d s
x(t )e −α ( t ) dt = ∫ b(t )e
−α ( t )
dt
0 dt 0

s s
x(t )e −α ( t ) ∫ b(t )e
−α ( t )
= dt
0 0

s
x( s )e −α ( s ) − x(0) = ∫ b(t )e
−α ( t )
dt
0

s
x( s ) = x(0)e α ( s ) + ∫ b(t )e
α ( s ) −α ( t )
dt (36)
0

This expression gives us the value of x(s ) as a function of its initial value x(0) and a
weighted sum of past values of the forcing term, b(t ) . This form of the solution is
convenient when there is a natural initial condition for x. Otherwise, the second form of
the solution may be more convenient:

• The forward solution is obtained by integrating (35) forward between s and ∞ :

d
( )
∞ ∞

∫s dt
x(t )e −α (t ) dt = ∫s
b(t )e −α ( t ) dt

13
 ∞
lim x(t )e −α ( t ) − x( s )e −α ( s ) =
t →∞ ∫ s
b(t )e −α ( t ) dt

∞
x( s ) = e α ( s ) lim x(t )e −α ( t ) −
t →∞ ∫s
b(t )e α ( s ) −α ( t ) dt (37)

provided that the limit exists.

Application: Households’ Intertemporal Budget Constraint

Suppose that an (inter-generationally) altruistic household has the following budget

constraint:

a1 (t ) = ra (t ) + y (t ) − c(t ) (38)

where a(t) denotes assets (or wealth) held at the beginning of period t, y(t) is labor
income received at the beginning of period t, c(t) denotes consumption expenditure
incurred at the beginning of period t, and r is the interest rate.

We can find the forward solution. Multiplying both sides of (38) by e − rt and rearranging:

[a1 (t ) − ra (t )]e − rt = [ y (t ) − c(t )]e − rt (39)

Note that the left hand side of (39) is the derivative of a (t )e − rt . Therefore, (34) can be
rewritten as:

d
dt
( )
a(t )e −rt = [ y (t ) − c(t )]e −rt (40)

Suppose that a(0) is given, so let’s integrate (40) forward starting from 0:

∞ ∞
a (t )e −rt ∫0 [ y(t ) − c(t )]e
− rt
= dt
0

Then
∞
− a(0) = ∫ 0
[ y (t ) − c(t )]e − rt dt

∞ ∞
∫0 c(t )e −rt dt = a(0) + ∫0 y (t )e −rt dt (41)

(41) is the household’s lifetime budget constraint.

14
Note: If the interest rate changes over time, we need to replace r by r(t) and the
t
integrating factor becomes e ∫0
− r ( s ) ds
, yielding:

t t
∞ − ∫ r ( s ) ds ∞ − ∫ r ( s ) ds
∫0 c(t )e 0 dt = a (0) + ∫0 y (t )e 0 dt (42)

as the intertemporal budget constraint (instead of (41)).

15