Differential equations

Basic concepts and definitions

Differential equations are functional equations (unknowns are functions)
which contain not only the unknown function but also their derivatives.
The form of the function which solves the functional equation is found by
solving the equation.

The general expressions of an ordinary differential equation in a function y

of one independent variable x, can be written as:
F(y, x, dy/dx, d2y/dx2, ..., dny/dxn) = f(x) (1)

Differential equations January 1, 2025 1

 The order of the equation is that of the highest derivative of it .
An equation relating the independent variable x, the unknown function y and
its first derivative, is called a differential equation of the first order.
If in addition, the equation contains a second derivative of the function we are
looking for, then such an equation is of the second order. Accordingly we can
define deferential equations of higher order.
The differential equation is linear with constant coefficients, if the
coefficients of y, dy/dx ... dny/dxn are all constants not involving x. So, the
general form of a linear differential equation with constant coefficients and
order n is:
a0(dny/dxn) + a1(dn-1y/dxn-1) + ...+ an-1(dy/dx) + any = f(x)
where f(x) is some function of x.

Differential equations January 1, 2025 2

 In the case where in place of f(x) on the right-hand side we have a constant
instead, the equation is called linear with constant coefficients and a
constant term.
The first-order case is written as:
dy/dx + a1y = f(x)

The second order:

d2y/dx2 + a1(dy/dx) + a2y = f(x)

The case where f(x) is absent altogether is called the homogeneous form, i.e.,
a0(dny/dxn) + a1(dn-1y/dxn-1) + ...+ an-1(dy/dx) + any = 0

Differential equations January 1, 2025 3

 Summarizing:
Differential equations

Linear
Non-linear
[Not examined]

Constant coeff. Variable coeff. [Not examined].

Homogeneous Non-homogeneous

Differential equations January 1, 2025 4

 Now, making the independent variable be time (t) instead of x, suppose we
have an equation like the following (2nd order):
d2y/dt2 + a1(dy/dt) + a2y = 0

Taking y to be an economic variable (for example national income, or

inflation) and t stands for time, then, since dy/dt describes how y changes over
time, and since d2y/dt2 describes how dy/dt is changing over time, it is clear
that the equation tells us something about the behaviour of y over time, so,
it is about economic dynamics.

Differential equations January 1, 2025

5
 Solving linear differential equations
By solution we mean an expression (“formula”) giving y at all points in time
such that it satisfies the equation.

The number e (2.718…) - base of the natural logarithm:

The number e plays a role in our analysis using differential equations (continuous
time). It will enter the solution of the differential equations.
Intuition:
Say that we invest $X at a rate of interest i, compounded annually. At the end of the
year the principal will be:
P = X(1 + i)
If instead the interest is compounded semi-annually:
P = X(1 + i/2)2

January 1, 2025
Differential equations 6
 Generalizing:
P = X(1 + i/n)n
In particular, if X=$1 and the rate of interest i = 100% (i = 1):
P = (1 + 1/n)n
[i.e., if interest is compounded daily, (n=365), P = 2.715]
If interest was compounded continuously, we would have a principal given by
the value approached by P as n → ∞.

The limit approached by P as n → ∞ is the number e, i.e., the principal of

one dollar compounded continuously at 100% interest; e ≈ 2.718….

Of course, money is not compounded continuously, but one could think of
events which behave very much like a sum on which interest is compounded
continuously.

January 1, 2025 7
Differential equations
 Solution of Linear 1st order Differential Equations
These can be written in the general form:
dy/dt + ay = f(t)
If f(t) = 0 (homogeneous form):
dy/dt + ay = 0

Since the above equation is separable (i.e., we can separate the terms with y
from those with t), the solution can be found using integration, as follows:
dy/dt = - ay → dy/y = -adt, then:
∫dy/y = -a∫dt
 ln(y) + c1 = -at + c2 → ln(y) = -at + c, then:
elny = e(-at + c) → y(t) = e-atec = Ae-at (solution)
[y(t) is same thing as yt]

January 1, 2025 8
First-Order: Solution
 Verify by substituting into the differential equation:
dy/dt = d(Ae-at)/dt = -aAe-at, so:
dy/dt + ay = -a Ae-at + a(Ae-at) = 0
We can find the same solution in a more general way:
Given: dy/dt + ay = 0:
Assume that the solution is in this form: y(t) = ceλt, then:
dy/dt = λceλt, and: λceλt + aceλt = 0 → ceλt (λ + a) = 0 → λ = -a,
and: y(t) = ce-at.

This is a general way of solving the homogeneous equation and there is an arbitrary
constant, c, appearing. To remove the unknown constant requires additional
information – i.e., an initial condition.
y(t) = ce-at; at t=0: y(0) = ce0 = c = y(0). So, for homogeneous equations the constant
= y(0).
 y(t) = y(0)e-at or: yt = y0e-at
This is the solution to the homogeneous equation.
First-Order: Solution January 1, 2025 9
 The general solution:
Besides the solution to the homogenous equation there is also a
particular solution; we need to take into account the non-
homogeneous part, f(t). The particular solution will be any particular
solution to the complete equation.

The general solution will be the sum of the solution to the

homogeneous equation (call it complimentary solution, yc) plus the
particular solution, yp.
The way to find the particular solution is by setting the solution to be a
general expression of the same form as f(t); i.e., if f(t) is a constant k
(such as f(t) = 3), try yp = k. If f(t) = e3t, try yp = ekt and so on.

First-Order: Solution January 1, 2025

10
 Example 1: Solve:
dy/dt – 3y = -3, y(0) = 2.

The solution to the homogeneous form is:

yc =ce3t [we called this the complimentary solution]

To find the particular solution: let y p = k. Then:

dy/dt = 0 and we have: 0 – 3k = -3 → k = 1.
The complete solution is: yt = ce3t + 1 [general solution]
Definite solution (find value of c):
Using the initial condition,
y(0) = c + 1 = 2 → c =1
and yt = e3t + 1 [definite solution]
Verify:
First-Order: Solution
dy/dt – 3y = 3e 3t
– 3(e 3t
+ 1) = -3 January 1, 2025 11
 Example 2: y – 2(dy/dt) + 2t = 6, y(0) = 1
To use the standard solution, we first need to re-write the equation in
the standard form:
2(dy/dt) – y = 2t – 6, or in the standard form :
dy/dt – ½(y) = t – 3. (1)
Then:
yc = ce(1/2)t = ce0.5t (check solution)

To find the particular solution, set yp = ko + k1t, then:

dy/dt = k1 and from (1): k1 – ½( ko + k1t) = t – 3

→ k1 – ½ko – ½k1t = t – 3
→ - ½k1t – (½ko - k1) = t – 3

Then: - ½k1= 1 and - (½ko - k1) = -3, or: (½ko - k1) = 3

First-Order: Solution January 1, 2025 12
 Continued…
Solving, we get k1 = -2 and ko = 2, and:
yp = 2 – 2t, so:
yt = ce0.5t + 2 – 2t [General solution]
Using the initial condition, y(0) = 1:
y(0)= c + 2 = 1 → c = -1

Then: yt = -e0.5t + 2 – 2t [definate solution]

Check by substituting into y – 2(dy/dt) + 2t = 6, given that:

dy/dt = -2 – 0.5e0.5t:
-e0.5t + 2 – 2t – 2(-2 – 0.5e0.5t) + 2t =
2 – 2t + 2t + e0.5t + -e0.5t + 4 = 6
First-Order: Solution January 1, 2025 13
 Dynamic models of market price – Stability of equilibrium
 In (static) demand-supply models, equilibrium is determined at the intersection
of D and S. Demand and supply of a commodity are assumed to depend only
on price (other things unchanged).
 What happens when the system is not in equilibrium? In particular, the
question is whether the system will tend to equilibrium or not.

 When the question is simply whether the economic forces will finally bring
the system towards equilibrium, without being concerned about the time
path of price (which relates to existence and stability of equilibrium), then the
question concerns the static stability of the system.

First-Order: Applications 19
January 1, 2025
  But even if economic forces push the system towards equilibrium, we
cannot exclude the case where equilibrium is overtaken. This could give
rise to oscillations before approaching equilibrium, if we ever reach it.
So the more interesting question is that of dynamic stability.

 st, we need to adopt an assumption about the behaviour of the variables

in the system (in this case, what makes P change). Stability conditions
will depend on this assumption.
The usual behavioural assumption is the “Walrasian” assumption:
According to it, if there is excess demand, price tends to rise; if there is
excess supply, price tends to fall.

First-Order: Applications January 1, 2025

20
 Equilibrium is statically stable if a P increase caused by (positive) excess demand will
decrease excess demand:
P
 Excess demand function: E(P) = D(P) - S(P)
 Differentiate excess demand with respect to P:
dE(P)/dP = d[D(P) – S(P)]/dP < 0 [condition for static stability]
or, dD/dP – dS/dP < 0 (slope of supply > slope of demand) Q

 On the other hand, when examining dynamic stability we express the behavioural
assumption in the form of a functional equation, which is then solved to derive the
time path of price.
So, price will be changing in response to excess demand:
dP/dt = P’ = c[D(P) – S(P)],
[i.e., price responds to excess demand (or supply), with a positive adjustment coefficient c]
From which we can solve for the time path of the time path of price, P t.
First-Order: Applications January 1, 2025 21
 In the case of linear demand and supply:
D = a + bP, S = a 1 + b1 P [if D and S have conventional slope, b < 0, b1 > 0]
dP/dt = c[a + bP – a1 – b1P] = c(b - b1)P + c(a – a1)
So, in standard form: dP/dt - c(b - b1)P = c(a – a1) (1)
(Static) equilibrium price is derived by setting dP/dt = 0:
c(b - b1)P + c(a – a1) = 0
Pe = - (a – a1)/(b – b1)

 On the other hand, to find the time path of price, we solve (1):
The solution of the homogeneous equation is:
Pc = Aec(b – b1)t (complimentary solution)

First-Order: Applications January 1, 2025 22

 For the particular solution we let:
Pp = k → -c(b – b1)k = c(a – a1) → k = - (a – a1)/(b – b1) = Pe, so:
Pt = Aec(b – b1)t + Pe
The particular solution represents the inter-temporal equilibrium price.
Looking at the exponent in the complimentary solution, we need:
c(b – b1) to be negative for stability (with c > 0).
So, the stability condition is:
b – b1 < 0 or b < b1
i.e., the slope of the supply must exceed that of the demand.
This is the dynamic stability condition; however, note that this condition is
the same as the static stability condition: dD/dP – dS/dP < 0.
This is not always the case though. When they are different, we are more
interested in the dynamic stability condition.
First-Order: Applications January 1, 2025 23
 Example: Static stability
D = 80 – 4P, S = -10 + 2P, P0 = 18
Find the equilibrium point (market clearing) and examine static
stability.

At equilibrium: 80 – 4P = -10 + 2P → Pe = 15, (and Qe = 20)

Using the Walrasian (static) stability assumption:

E(P) = 80 – 4P – (-10 + 2P) = 90 – 6P
Or:
dE/dP = -6 < 0, i.e., static stability.

January 1, 2025
24
First-Order: Applications
 (Same) Example: Dynamic stability
dP/dt = P’ = c[80 – 4P – (-10 + 2P)] = c(90 – 6P) = -6cP + 90c
or: dP/dt + 6cP = 90c
which has the solution:

Pt = Ae-6ct + 15 (Pp = k, 6ck = 90c → k = 15)

Given that P0 = 18 → P0 = A + 15 = 18 → A = 18 – 15 = 3
Pt = 3e-6ct + 15
Given c > 0, the term 3e-6ct → 0 and equilibrium is dynamically stable.

First-Order: Applications January 1, 2025

25
 An alternative behavioural assumption:
Assume small shocks in market price.
(instead of: price adjusts according to excess demand or supply): Assume that the
market clears at any point in time. [Example: price of stock in the stock market]
In this case, to have a dynamic model, the derivative dP/dt must be in the
demand or the supply or both.
Consider the following demand and supply for a product:
Dt = D(Pt, dP/dt) [does it make sense? Yes]
St = S(Pt)
Dt = St
→ derive a dynamic equation (1st order differential eq)

First-Order: Applications January 1, 2025

27
 Assuming linear functions:
Dt = a1 + a2P + a3(dP/dt)
St = b1 + b2P
→ Dt = St, or: a1 + a2P + a3(dP/dt) = b1 + b2P = Q
→ dP/dt + [(a2 – b2)/a3]P = (b1 – a1)/a3, and the time path of P is:
Pt = Ae[(b2 – a2)/a3]t + P*, or
Pt = (P0 – P*)e[(b2 – a2)/a3]t + P* (since P0 = A + P* → A= P0 – P*)
For price to converge to the inter-temporal equilibrium value: P* (for
convergence/stability) we need:
[(b2 – a2)/a3] < 0, then as t → ∞: lim (e[(b2 – a2)/a3]t) = 0, and Pt → P*

First-Order: Applications January 1, 2025 28