Chapter 5: Difference and Differential Equations (Dynamic Systems)

1.1.Introduction

In real world, in most real economic problems we have to study the development of variables
over time.The intertemporal development of this variable is usually expressed in the form of
equations. If time is taken to be a discrete variable, these,equations are called ordinary difference
equations ( OD/ceE) and if time is taken to be continous , we peak of ordinary differential
equations (ODEs) .Both techniques solve the same problem but depending on how the model is
developed, one technique may lend itself more naturally to the solution than the other.An
economic variable could be a function of some other variable(s) in some period or in the same
period.Also may , an action taken in one period have their consequences in some other periodsor
in the same period.The incorporation of time in to economic model is known as dynamics.Thus,
dynamic analysis is the study of the time path due to the change in different variable(s) as time
changes.

The time path can be studied in to discrete time analysis and continous time analysis.In discrete
time analysis, time is treated as in discrete period and variables change once in a period.An
example for this type is when the price of certain agricultural products change from one season
to the next but are fixed through each season.

Continous time analysis on the other hand, time is treated in continuity and variables change at
every moment of time.The price of oil, effectively change continousely and are not fixed on a
seasonal bases is an example.Therefore, a tool for analyze the discrete analysis and continous
analysis is Difference equation and Differential equation respectively.

1.1.1. Difference Equation

Adifference equation expresses a relationship between a dependent variable and a lagged

independent variable(s) which changes at discrete intervals of time, forexample I t  f y ,
t 1

where I ,& y are measured at the end of each year.

In other way, difference equation is an equation that links endogenous variable in one period to
its value in to some other period aswell as the value of other variable.It is also a recurrence
relation, which relates consecuitive terms of a sequence of numbers.

Example:

1) y  ay
t t 1
 ct

2) y  ay
t t 1
 by
t 2
 ct , where y -is dependent variable; c
t t
-is independent variable.
 a) Order of Difference Equations

Based on the complexity nature of the socio-economic events of our real world’s activity
functions or equations may goes upto from simple to complex or from linear to non-linear or
from 1st order to higher order.So, order of difference equation is the highest lagged period in a
given difference equation.A1st order difference equation expresses a time lag of one period; a
2ndorder, two periods, etc.

Hence, a difference equation from the above expression say equation of example (1) stands for
1st order and equation of example (2) is 2nd order difference equation.

b) Solution for Difference equation

Through repeated iteration we can give an alternative solution of 1 st order difference equation:
Suppose y  ay  c t t 1 t
be first order equation.For simplicity let we assume that c c
t

(constant), then y  ay  c .Thus, take period t 1 ,

y  ay  c , putting this in the
t t 1 t 1 t 2

y  aay  c  c  y  a y  ac  c. Again take period

2
original difference equation.i.e.,
t t 2 t t 2

t 2, y  ay  c , putting this in the above difference equation

t 2 t 3

y  a ay 
 c  ac  c  y a y  a c  ac  c .If we continue this process for t
2 3 2
t t 3 t t 3

periods and assume that t  0 ,initial value in y , we would get a sequence like:
0

1 a
t 1 t

y  a y  a c ,
t t
y a y 
t
then the general solution can be formulated as: .c is an
t 0
i 0
t 0 1 a
alternative way.
t 1
To prove this relationship, let’s first define: S   a  1  a  a  a  ...  a      ( I )
t 2 3 t 1

i 0

Then, Sa  a  a  a  a  ...  a                              ( II )
2 3 4 t

equation(I)-equation(II)= S  Sa  S 1  a   S 1  a   1  a
t
and apply subtraction,

1 a
t
t 1
S   a .
t

1 a i 0

c) Steady State Value (SSV)

The steady stae value of the difference equation is when the particular solution expresses the
intertemporal equilibrium level of difference equation.It is the equilibrium value / long-term
value of the difference equation.

Suppose, y  ay
t t 1
 c , when t  , lim y  lim
t 
t
t 
ay t 1

c  y  ay
 
c

 c  1  a  y  c 
1
 y  ay y  .c
    1  a 
d) Nature of the Time Path

The time path of the equation dependent on the sign of difference of two time periods, i.e., if
y  y  t t 1
may tend to the steady stae value as t changes then the sequence is convergent
(stable), if not it will be divergent (unstable).


1  300, y  400 , then find the
Example1. If the difference equation with its initial is y t
y 0
2 t 1

SSV.

Solution: let we consider the sequence of the given difference equation and its consequitive
differences in tabular form as follows:

y y  y y
t t t t 1

y  400 -
0

1 500-400=100
y 
y  300  500
1 2 0
1
y2  2 y1  300  550 550-500=50
1
y3  2 y2  300  575 575-550=25
1
y4  2 y3  300  587.5 587.5-575=12.5

. .
. .
. .
Here, from the above example y  y 
t t 1
implies the same sign, thus it is monotonic.Then; SSV
1 1
will be y .c   .300  600 .This shows that the above sequence values converge to
1 a   1
1  
 2
the value of 600.

1
Example2.If the difference equation with its initial is y  y
t
 300, y  400 , then find
0
2 t 1

the SSV.

Solution: let we consider the sequence of the given difference equation and its consequitive
differences in tabular form, like the first example above as follows:

y y  y y
t t t t 1

y  400 -
0
1
y 
y  300  100 100-400=-300
1 2 0
1
y2   2 y1  300  250 250-100=175
1
y3   2 y 2  300  175 175-250=-75
1
y4   2 y3  300  212.5 212.5-175=37.5
1
y5   2 y 4  300  193.75 193.75-212.5=-18,75
. .
. .
. .

Here in this case, we have alternate sign of two consequitive periods as t changes, the time path
is oscillatory as observed from the table abov.Though it is oscillatory its SSV=200.

Example3.Suppose, we have the difference equation as y  ay

t t 1
c :
 i) If  1  a  1 , then irrespective of the value of y , it will converge to theSSV as t  
0

,and thus the system is stable.

ii) If a  1, or a  1 ,and y  SSV ,then it will be divergent from the SSVas t   ,and
0

thus the system is unstable.


7  30, y  200.
Example4. Suppose, we have the difference equation as y t
y 0
8 t 1

7
Solution:  1  1,then the time path is monotonic, convergent and the system is stable, its
8
1 1
SSV  y  .c  .30  240. that is y  SSV too.
 1  a  1  7  0
 
 8

1
Example5. Suppose, we have the difference equation as: y  y t
 6, y  50 .
0
5 t 1

1
Solution: Since  1    1,then the time path is oscillatory, convergent and stable.Its
5
1 1
SSV  .c  .6  5.
1  a  1  (  1) 
 
 5 

e) General Solutions for Difference Equation

Suppose; y  ay
t t 1
c is first order difference equation.General solution (GS)
=Complementary function (CF) +Particular solution (PS).The complimentary function is also
called homogenous solution (part) which represents the deviations from the equilibrium.
Where as the particular solution experesses the intertemporal equilibrium level of y (ort
the
long-run equilibrium level).To solves the solution for the given difference equation, let’s
separate the equation in to two parts (homogenous and particular) as follows:

The Homogenous part: y  ay

t t 1
  A a , which is complimentary
t
Suppose the solution’s to this homogenous part is y t
t 1
function(CF) and A is deterministic.At t  1  y  A a .The particular solution (PS) can be
t 1

any solution:simplicity let For it be the SSV.Therefore,

1 c
GS  y  CF  PS  y  A a  .c ,but t  0 , then y  A a 
t 0
.
t t 1  a  0 1  a 
1 1  1  t 1
 y  A .c  A  y  1  a  .c and also, y   y  .c  a  .c after
0 1  a  0 t
 0 1  a   1  a 
1 a
t
1
the solutionto the difference equation:yt  A a 
t
y  y a  1  a .c .Thus
t
solving, .c, if
t 0 1 a
1 a
t

a  1and  c, if a  1 or definite solution will be expressed as: y  y a  1  a , for

t
 Aa
t
y t t 0

a  1 and yy  c, for a  1

t 0

Example1. Solve the following difference equation with the specified inititial
conditions.Comment on the qualitative behavior of the solution in each case.

a) y  4y
t t 1
 21; y  1
0

Solution: given: y  1; c  21; a  4  1

0

GS  y   y
t 0

c  t
a 
1 a 
c
1 a

 1 
21  t
 4 
 1  4 
21
1  4
8 4  7 .Hence, y increases
t

without bound as t increases infinitely, therefore, the time path diverges uniformly or it explods,
because a  1  4  1.


1  8; y  2
b) y t
y 0
3 t 1

1
c) Solution: given: y  2; c  8; a  1
0 3
  
  t t

 c  t c
  2 
8   1  8  1 
   1 
GS  y   y  a    10    12 .The time
      
 1     3  1    3
t 0 1 a 1 a 1

  3   3
path converges uniformly to the value of 12 which is related as the equilibrium value as t   ,
y t
tends to the equilibrium value which is equal to 12.

Exercises:

1. Solve the following difference equation with the specified initial conditions.
1
a) y   y  6; y  0
t 0
2 t 1

b) y   2 y  9; y  4
t t 1 0

A  1 and B  0 , then the component

t
Suppose y  A a  B and assume
t
2.
t a generate
various types of time path as t changes:
i. Recall all possible characteristics(cases);
ii. What is the magnitude and sign of A affect?
iii. What affects the magnitude of B ? and if B  0,then what is the effect of B ?

Solution:

Case:

(1) If a  1, a
t
increases also ast increases.Thus time path y 
t
will explode (uniform
divergent).
(2) If a  1, a
t
 1, for all t , then time path y  will represent a horizontal line.
 
t

(3) If 1  a  0 , then the time path y will oscillate between positive and negative values
t

and move towards the equilibrium (oscillatory convergence).

(4) If 0  a  1, a decreases as t
t
increase.Thus, time path y 
t
will dampen (converge and
make towards equilibrium).
(5) If a  0,  0 , for all t ,then it will represent the x  axis.
t
a
(6) If a  1, in this case one point will be oscillate between positive and negative.
(7) If a  1,in this case, time path y will oscillate and make further and thus diverges.
t
The magnitude of A reduces the scale effect with out changing the configuration of time
path.But its sign reduces both scale effect and mirror effect.

Example . If A  1 ,then the mirror image of the time path a with respect to the horizontal line
t

will be shown.The magnitude of B affects the vertical intercept of the curve.Thus, in noway
affects the convergence/divergence.

1.1.1.1. Economic Application

Known that many socio-economic phenomenon appear not in a continuous one and hence for
this evidence studies also focus not only on relating continuous changes of variables but is on
discrete changes.In planning models, forexample, the comparison is between the initial base year
and the terminal year and change in investment over the period is related to change in time over a
period.Both the changes are said to be discrete.

This indicates that the discrete dynamic equations have an enormous socio-economic application,
among those some will be considered as follows:

I . National Income Determination

Assume two-sector model with structured equations:

y cI
c  ay  b

II

Where b , and I -denote autonomus consumption and investment, a - is marginal propensity to

consume 0  a  1.The corresponding consumption function is given by:ct  ay  b .The
t 1

time lag between consumption c and income y is ct andy .If I  I is constant,finaly the flow
t 1

of money is inbalance in each time period.Then y c I t t t

.Substituting, we get

y  ay
t t 1
 b  I is difference equation of standard form.

Example1.Consider the two-sector model: y  c  I ;c  0.8 y

t t t t t 1
 100, I t  200 ,find an

expression for y , when y  1700. Is the system stable or unstable?

t 0


Soluton: for an equation: y is y , Given c  0.8 y
t t t t 1
 100, I t  200 , y  1700.
0
y  0.8 y
t t 1

 100  200

y  0 .8 y
t t 1
 300

y


t

 1700 

300 

1  0.8  0.8  1300
t

 0 .8 

y


t
 200 0.8  1500
t

Then, as t approaches infinitely large t  , the exponent 0.8 tends to converge to zero
t


and so the time path y settle down at the equilibrium level of 1500.It is therefore, stable, time
t

path displays uniform convergence.

Exercise1. Consider the two-sector model: y  c  I c  0.9 y

t t t t t 1
 250, I 0  350.

i. find the expression for y when y  6500.

t 0

ii. Is the system stable or unstable?

II . Supply and Demand analysis

Assume that the S and D functions are both linear, then we have the relation:
Q  aP  b;Q  cP  d . It is known that Q is depend on Pt 1 (example, farmers’ farm
S D St

activity when and why they sew the crop and sell it!!!). Hence, the corresponding time dependent
supply and demand equations are:

Q  a Pt 1  b
 Assume equilibrium, then QSt  QDt, that is: a Pt 1  b  c Pt  d 
st

Q  cP  d 
Dt 

 c Pt  a Pt 1  b  d   a
    Pt 1 
b  d   is Difference equation of the standard form.
P t
 c c

Q  4 P  10
st t 1
Example1. Consider the demand and supply equations:  assuming that the
Q   5 P  35 
Dt

market is in equilibrium. Find the expression for P and Q , when P  6 , is the system
t t 0

stable or not?

Solution: If Q  Q , then
Dt St 5P  35  4P t t 1
 10  Pt  0.8Pt 1  9.
 
GS  Pt   6 

9 

1  (0.8)  0.8  1  (90.8)  P  0.8  5.
t
t
t

Q   5Pt  35
Dt

and for Q , taking Q

t Dt
 0.8  5  35
 5

t

 5  0.8  10  Q
t

as t increases,  0.8 converges to zero and so

t
P t
and Qt
eventually settle down at the
equilibrium levelof 5 and 10 respectively.The system is then stable.And the time path display
oscillatory convergence.

9.1.2. Second Order Difference Equations (SODEs)

The general solution of second order difference equation is the summation of the particular
solution and the solution of homogenous part (complementary function) : y  CF  PS
t
.Thus
the particular solution which is the ssv.And thus as t ,
lim y  ay
t 
t t 1
 by
t 2

c  y  ay  by
  
 c. ;

 1a by 
 c,  y 

p
1
y  1  a  b  c; a  b  1;

1 
ifa  b  1, then y 
 ct ; a  2 
y y 
1
 c;
2  a   p

p  1  a  b 1 2
ifa  b  1, and , a  2, then y   c t 
p 2 

y  ay  by  c ,assume the solution as : y k

t
The solution of the homogenous part: ,
t t 1 t 2 t

then we can write the above as:

y  ay  by  0
t t 1 t 2
t 1 t 2
k  ak  bk  0
t

k ak  bk  0,
2 t 2
  4b
2
1
Sofor k  0, k  ak  b  0 is auxiliary quadratic equation.Hence, k 1 , k 2  a 
2 a ,here
2 4
we have three possible sets of the value of k.

Cases:
2

1) a  b  k 1  k 2 and two distinct real values;

4
2

2) a  b  k 1  k 2 and repeated (same real value)

4
2

3) a  b  the roots are complex number.

4

Case1.Whenroots are distinct and real.

y  Ak  B k 2 ,and the general solution can be

t t
The solution of the homogenous part would be 1
t

GS  y  A k1  B k 2 
t t
expressed as:
t
y 
, which is called the complete solution to the
t t
second order difference equation. A k 1 and B k 2 are linearly independent with the given values
say y 0
and y 1
,by putting these in the above we can get a pair of simultaniouse equations and
solving them we could get values of A and B and so we could obtain a definite solution.

 yt  A k1  B k 2  y ,where Ps  y  1  a  b .c; a  b  0
t t 1

Case2. When roots are repeated value of k ; k1  k 2  .

y   A  Bt k
t
The solution of the homogenous part would be , where A and B are constants
t

to be determined.So, the complete solution to the second order difference equation could be
y   A  Bt k  y
t
expressed as: t 

case3. When roots are complex value of k .

The complex number can be written as: r  m  1 ,where r is the real part,and m is the
imagenary part.Hence, the solution of the homogenous part and the Ps will be.
y  A b
t   t  at 
cos   B
 2 b 
 b  t  at 
sin   
 2 b 
y 
or complex conjugate roots, if the

k  r  mi& k  r  miand R cosmt , R sin mt  are

t t
roots are complex conjugate then 1 2

y  c R cost   c R sin t   y , where R  r  m ,

t t 2 2
solutions and the general solution is
t 1 2 

r m sin  m
cos   and sin   ,or tan    ,where R is assumed to be the radius of the unit
R R cos  r
circle. for example consider the 2nd-order difference equation y  4 y  16 y  0 .The
t 2 t 1 t

4  16  64 48
 4k  16  0 and roots will be   2
2
characteristic equation is k k 1, 2
,i.e.,
2 2
1 1 1
k 1
 2 48 i , k 2  2  48 i and r  2; m  48 the polar coordinates will be
2 2 2
2 1
 48

R  2 
1  r 2 1 m 2 3
48  = 4  12  4 , cos   R  4  2 ,and sin   R  4 
2
.Hence, it
2  2

  
implying that 
3
.Thus, y  c 4 cos  3 t   c 4 sin  3 t .If
t 1
t
2
t
k&k 1 2
are complex

conjugate ,then yt
oscillate because the cosine function oscillates. There are, however, three
different types of oscillation:

i) If R  1 , then the characteristic roots k &k 1 2

lie outside the unit circle shown in
fig(a).Hence y t
is oscillating, but increasing in magnitude. The system is unstable.

ii) If R  1, then the characteristic roots k &k 1 2

lie on the unit circle shown in fig (b), and
the system y
t
oscillates, with a constant magnitude.

iii) If R  1, then the characteristic roots k &k1 2

lie inside the unit circle shown in fig(c),
and the system y is oscillates, but converges to zero as t   . The system is stable.
t
Stability Criteria

(1) Two distinct values:

-If the absolute value of each is lessthan one, then it will be stable.

(2) One Value:

-If the absolute value of it is lessthan one, then it will be stable.

(3) comlex vale:

-If the absolute value of the real part of each one is lessthan one, then it will be stable.
Examples:

1. For Distinct real roots:

1 3 1
y  y  y 5 Ps  Ps  y  .5  16
t ; Particurar solution  1 3
2 t 1 16 t 2 1   
 2 16 

Roots will be emanate from the homogenous 2nd order difference equation,we get
 
 1
yt 2  =0
3

 yt y 16
 2 t 1 

 2 1 3
k  k    0
 2 16 

1 1  3
 4  1    
    16 
1 3


2
2 
1 1
  4 4  1  1  3 , 1 .Thus, the general
k ,k
1 2
2 1 2 2 2 4 2 4 4
t t
3  1
solution to the given 2nd order equation is y  A   B     16 .Now we can determine
 4  4
t

the constants A and B using the given initial values, assuming that at t  0 , y  1and at t  1,
0

y  2 , then substituting them in to thegeneral solution y respectively we form a pair of

1 t

3   1  
0 0

1  A    B     16 
4  4 
simultaneous equation and solving: 1  1
3  1 
2  A    B     16 
 4   4  
A  B  15

 11 71
3 1  B  ;A 
A  B  14  4 4
4 4 
Hence, the general solution for the specified equation is
t t

 71   3   11   1 
 y              16 .
 4 4 4
   
t
4
2. For Repeated real solution:

 
1 
5
Ps  Ps  
1 5
5
i) y t
y t 1
y 4 ; Particurar solution
y   1 4
4 t 2 1  1  
 4

Roots will be emanate from the homogenous 2nd order difference equation, we get
 1 
 yt  yt 1  yt 2   0
 4 

 1 
 yt  y  y 0
 t 1 4 t 2


 2 1
k  k    0
 4

1 1  4  1  14 

2

1
k ,k   .
1 2
2 1 2
t
 1
Thus, the general solution to the given 2 nd order equation is y   A  Bt     5 .Now we
 2
t

can determine the constants A and B using the given initial values, assuming that at t  0 ,
y  10 and at t  1, y  8 , then substituting them in to the general solution y respectively we
0 1 t

 10  A  B  5

form a pair of simultaneous equation andsolving: 1 
8   A  B    5
2 
A B  5


1 1   B  1; A  5
A  B  3
2 2 
 t
1
Hence, the general solution for the specified equation is  y  5  t    5 .
 2
t

y  6 y
1 1
ii)  9y 4 Ps  Ps  y  4 
t t 1 t 2 ; Particurar solution 1  6  9 4

Roots will be emanate from the homogenous 2nd order difference equation, we get
y   6 y  9 y 
t t 1 t 2

y  6 y  9 y   0
t t 1 t 2

k 2
 6k  9  0 


6 6  4  1 9  3 
2
36  36
 3 .
k ,k
1 2
2 1 2

y   A  Bt  3  4 .To
1
t
Thus, the general solution to the given 2 nd order equation is
t

determine the constants A and B addititional initial values should be given, say as such that at
t  0 , y and at t  1 ,y , then substituting them in to the general solution y respectively we
0 1 t

can form a pair of simultaneous equation andsolving it ,finally possible to expresse the general
solution for the specified equation like the above equation.

3. For Complex Roots:

i) If y  4 y 16 y
t t 1 t 2

y  4 y 16
t t 1 t 2
0

4 16  4 16
k ,k
1 2
  has complex roots.
2 4

Hence, y t
A  16  t  4t 
cos   B
 16 
 16  t  4t 
sin   
 16 
y  A4 cost   B 4 sin t 
t
t t

ii) y  2y  2 y ; at t  0 , y  0 and at t  1, y 1
t t 1 t 2 0 1
 2i
y  2y  2y  0 ; k1, k 2  1 
t t 1 t 2 2

Solution:
t  
y  A b  b  t  at 
cos   B
 2 b 
t  at 
sin   
 2 b 
y 


y  A t  2  2
t  t 
cos
y
 2
B
t  t 
sin 
 2


; but at t  0 , y  0 and at t  1,
0
y 1
1

o A 2 cos 02   B 2 sin  02 

0 0
A0 

  A B ; B  1, A  0
1 A 2 cos 12   B 2 sin  12 
1 1 


2
 2

2
 1


Thus, the required solution is: y t

  2
t
 
sin  t  
4 
y 
;

 2    
/// sin  t   sin  t  ///;  90
0

 2  4  4