Assignment Difference and Differential Equations

2015-2016
Group 15
Sylvia Cheng (S2701367)
Ruth Oude Avenhuis (S2766515)
Aho Mirza (S2238462)
Marc Achterbosch (S2766345)
December 18, 2015

Contents
Exercise 1
a. Macroeconomic model in per capita
b. System of difference equations . . .
c. Equilibrium solutions . . . . . . . .
d. Stability . . . . . . . . . . . . . . .
Exercise 2
a. Nonlinear difference equations:
b. Solutions and their stability .
c. -limit points . . . . . . . . .
d. Bifurcation diagrams . . . . .

form
. . .
. . .
. . .

plots
. . . .
. . . .
. . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

3
3
4
4
6

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

7
7
7
8
8

Exercise 3
10
a. Equilibrium solutions and stability . . . . . . . . . . . . . . . . . . 10
b. Bifurcation diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Appendix
Exercise
Exercise
Exercise
Exercise
Exercise

2a
2b
2c
2d
3b

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

12
12
15
18
20
22

Exercise 1
Consider the following stylized macroeconomic model that describes an economy
that is aected by environmental pollution:
Pt+1 = rPt + Kt At ,
Lt+1 = nLt ,
Kt+1 = Kt + It ,
Yt = Ct + It + At ,
It = sYt ,
Ct = eYt ,
Yt = F (Kt , Lt ), t 0,
where 0 < r < 1, > 0, n > 1, 0 < < 1, 0 < s < 1, 0 < e < 1, and s + e < 1.
In these equations Pt is the stock of pollution in year t, Kt is the capital stock
in year t, At is the abatement in year t, Lt is the amount of labour employed in
year t, It is the investment in capital in year t, Yt is total production in year t,
and Ct is comsumption in year t. The production function F : R2+ R is given
by
KL
F (K, L) =
L + (1 )K
, where 0 < < 1.
a. Write the model (including the production function) in per capita form.
t
Use lower case letters to denote the per capita variables. For instance, kt = K
Lt
is capital per capita.
Solution.
pt =

Pt
Kt
It
Yt
Ct
At
, kt =
, it =
, yt =
, ct =
, at =
Lt
Lt
Lt
Lt
Lt
Lt
Lt+1 = nLt

pt+1 =

Pt+1
rPt + Kt At
Lt rPt + Kt At
r

at
=
=
= n1 (rpt +kt at ) = pt + kt
Lt+1
Lt+1
Lt+1
Lt
n
n
n

kt+1 =

Kt+1
Kt + It
Lt
Kt + It
kt + it

it
=
=

=
= kt +
Lt+1
Lt+1
Lt+1
Lt
n
n
n
Yt = Ct + It + At = eYt + sYt + At =
yt =

Yt
at
=
Lt
1es

it =

It
Yt
=s
= syt
Lt
Lt

ct =

Ct
Yt
=e
= eyt
Lt
Lt

At
1es

The function F is a linear mapping since the following holds for arbitrary
R+ :
2 Kt Lt
F (Kt , Lt ) =
(Lt + (1 )Kt )

Kt Lt
= F (Kt , Lt )
Lt + (1 )Kt

It can be concluded that F is a homogeneous function where the production

function in per capita form F : R+ R is given by
f (kt ) =

Kt Lt
Kt
kt
F (Kt , Lt )
=
=
=
Lt
Lt (Lt + (1 )Kt )
Lt + (1 )Kt
+ (1 )kt

b. Determine a system of difference equations for (kt , pt ).

Solution.
By a)
r
pt +
n

= kt +
n

pt+1 =
kt+1

at = (1 e s)yt
it = syt , with yt = f (kt ) =
This implies,

at
kt
n
n
it
n

(1)
(2)

kt
+(1)kt

at = (1 e s)f (kt ) =
and
it = sf (kt ) =

(1 e s)kt
+ (1 )kt

skt
+ (1 )kt

Substituting at and it , respectively, in (1) and (2) gives:

pt+1 =

(1 e s)kt
r
pt + kt
n
n
n( + (1 )kt )

kt+1 =

skt
kt +
n
n( + (1 )kt )

c. Besides the null solution the system of difference equations for (kt , pt ) has
a second equilibrium solution. Determine this solution. Find two conditions
involving the parameters r, , n, , s, e, and such that this solution makes economic sense if the two conditions are met.
Solution.
By b)
pt+1 =

(1 e s)f (kt )
pt + kt
n
n
n
4


sf (kt )
kt +
n
n
= pt and z2 = kt+1 = kt . This gives:
kt+1 =

Now let z1 = pt+1

(1 e s)f (z2 )
z1 + z2
n
n
n

sf (z2 )
z2 = z2 +
n
n
First solving (4) for z2 gives the following:
z1 =

z2 =
Substituting f (z2 ) =

z2
+(1)z2

(3)
(4)

sf (z2 )
n

and doing some simple computations gives:

z2 =

s (n )
(n )(1 )

Now substitute z2 in (3) and solve for z1 to find the following:

z1 = (

1 e s s (n )

)(
)
(n )(1 )
s(1 )
nr

Hence,
z=

  


1es
( (n)(1)
s(1)
)( s(n)
)
z1
nr
=
s(n)
z2
(n)(1)

is an equilibrium solution.
Two conditions involving all parameters such that the equilibrium solution
makes economic sense are given by the following:
s (n )

(n)

1es
s

d. Examine the stability properties of the equilibrium solution you have

found in part c.
Solution.
To examine the stabilty properties of the equilibrium solution, theorem 1.5.2.
can be used.
r

n pt

f (pt , kt ) =

(1es)kt 
n(+(1)kt )
skt
n(+(1)kt )

+ n kt

n kt

The Jacobian of f (pt , kt ) exists, since f (pt , kt ) is differentiable at c for all c Ck

and is given by:
!
(n(+(1)k )2 )n(1es)
Df (pt , kt ) =

r
n

(n(+(1)kt ))2
(n(+(1)kt )2 )ns
(n(+(1)kt ))2

Substituting the equilibrium solution in the Jacobian gives:

!
2
2
A = Df (z1 , z2 ) =

r
n

s (1es)(n)
ns2
s(n)2
ns2

The eigenvalues of Df (z1 , z2 ) are:

2
1 = nr and 2 = s(n)
ns2
By Theorem 1.5.2. if r (A) < 1, then the equilibrium solution is asymptotically stable.
If 1 > 2 , then:
r (A) = |1 | = | nr | < 1 since 0 < r < 1 and n > 1, and hence the equilibrium
solution is asymptotically stable for 1 > 2 .
If 1 = 2 , then:
r (A) = |1 | = |2 | < 1, hence the equilibrium solution is asymptotically stable
for 1 = 2 .
If 1 < 2 , then:
2
r (A) = |2 | = | s(n)
| < 1 if ns2 > s (n )2 and hence the equilibns2
rium solution is asymptotically stable for 1 < 2 if ns2 > s (n )2 . If
ns2 < s (n )2 , the equilibrium solution is unstable.

Exercise 2
Consider the 1-parameter family of nonlinear difference equations
p(t + 1) = f (p(t)) = p(t) exp p(t), > 0

(1)

This difference equation is used in demographics to model the dynamics of population densities (p is the population density in period t).
a. Write a program (in MATLAB) that plots for a given value of > 0
and initial population density p0 the solution p of (1) versus time t. Provide
such plots for three combinations of and p0 .
Solution.
See appendix page 13 for the Matlab code, and pages 14 and 15 for the figures
1, 2 and 3.
b. Choose an 1 (0, 1), an 2 (2, 6), an 3 (8, 12), and an 4 (13, 14).
For each of those s, do the following: Plot in a single figure the graphs of
x x, f , f2 and f4 . Use these plots to find approximate values for equilibrium solutions and any periodic solutions that can be deduced from the plots.
Estimate the slopes of tangents in order to examine the stability of solutions.
Solution.
For the Matlab code see appendix (page 16).
In figure 4 in the appendix (page 17), where = 0.5, the only equilibrium
solution is the null solution. Namely, the only point where all lines x 7
2
4
x, f0.5 , f0.5
, f0.5
intersect is at the origin.
Furthermore, the slope of the tangent line of f0.5 and x 7 x at x = 0 is less
than one in absolute value. Hence, the null solution is asymptotically stable at
x = 0.
There are no periodic solutions for = 0.5
In figure 5 in the appendix (page 17), where = 4, there are two equilibrium
solutions where all lines x 7 x, f4 , f42 , f44 intersect. One is the null solution and
the other one is approximately at x 1.3863.
Furthermore, the slope of the tangent line of f4 and x 7 x at x = 0 is greater
than one in absolute value. Hence, the null solution is unstable at x = 0.
The slope of the tangent line of f4 and x 7 x at x = 1.3863 is less than one in
absolute value. Hence, this second solution is asymptotically stable.
There are no periodic solutions for = 4
In figure 6 in the appendix (page 18), where = 10, there are two equilib2
4
intersect. One is the null
rium solutions where all lines x 7 x, f10 , f10
, f10
solution and the other one is approximately at x 2.3026.
Furthermore, the slope of the tangent line of f10 and x 7 x at x = 0 is greater
than one in absolute value. Hence, the null solution is unstable at x = 0.
The slope of the tangent line of f10 and x 7 x at x = 2.3026 is less than one in
absolute value. Hence, this second solution is asymptotically stable.
2
4
intersect at x 0.9346 and at x 3.6706, hence these
The lines x 7 x, f10
, f10
are 2-periodic solutions.
7

2
For these 2-periodic solutions, the slope of the tangent line of x 7 x and f10
is
less than 1 in absolute value, hence these 2-periodic solutions are asymptotically
stable.

In figure 7 in the appendix (page 18), where = 13.5, there are two equi2
4
librium solutions where all lines x 7 x, f13.5 , f13.5
, f13.5
intersect. One is the
null solution and the other one is approximately at x 2.6027.
Furthermore, the slope of the tangent line of f13.5 and x 7 x at x = 0 is greater
than one in absolute value. Hence, the null solution is unstable at x = 0.
The slope of the tangent line of f13.5 and x 7 x at x = 2.6027 is greater than
one in absolute value. Hence, this second solution is unstable.
2
4
The lines x 7 x, f13.5
, f13.5
intersect at x 0.6427 and at x 4.5627, hence
these are 2-periodic solutions.
2
For these 2-periodic solutions, the slope of the tangent line of x 7 x and f13.5
is
greater than 1 in absolute value, hence these 2-periodic solutions are unstable.
4
The lines x 7 x, f13.5
intersect at x 0.4677, x 1.0228, x 3.9552 and at
x 4.9651, hence these are 4-periodic solutions.
4
For these 4-periodic solutions, the slope of the tangent line of x 7 x and f13.5
is
less than 1 in absolute value, hence these 4-periodic solutions are asymptotically
stable.
c. Write a program that produces for a given interval [min , max ] R+
and initial value p0 0 a plot with the approximate -limit points of p0 (with
on the horizontal axis and on the vertical axis the -limit points). Of course,
it is in general impossible to find the exact values of the -limit points numerically. Carefully explain how you were able to find approximate values of the
-limit points. Choose a p0 > 0 and use your program to plot the -limit points
for the initial values 0 and p0 with between 0 and 12.
Solution.
For the Matlab code see page 19.
In figure 8 in the appendix (page 20), all -limit points are 0 for all between
0 and 12.
In figure 9 in the appendix (page 20), the -limit points are 0 for between
0 and 1. The -limit points are increasing from 0 to 2 for between 1 and
approximately 7.3. The -limit points are increasing with a steeper slope from
2 to approximately 3.5 for between 7.3 and approximately 9.6.The -limit
points are decreasing from 0.98 to 0.74 for between 9.6 and approximately 12.
d. Generalize your program of c. in such a way that it can produce bifurcation
diagrams. That is, for each it should plot the - limit points of sufficiently
many initial values, where the initial values should be chosen smartly. Construct
this bifurcation diagram for [min , max ] = [0, 16]. Discuss the bifurcations occurring in your plot (What kind of bifurcation? Where do they occur?).
Solution.
For the Matlab code see page 21.
See figure 10 in the appendix (page 22).
The initial value p0 = 0 has been excluded from the bifurcation diagram. This
is due to a horizontal line that has zero as limit point for every change in .
8

The following can be derived from the bifurcation diagram and by using the
solutions of question 2b.
If (0, 1) the null solution is asymptotically stable. When = 1 there occurs
a transcritical bifurcation, this means that the stability of the null solution is
transferred to a limit point between (1, 7.41). Furthermore, there occurs
a period doubling at an that is approximated to be 7.41. The stability of the periodic solutions are asymptotically stable for (7.41, 12.35). In
addition, at an that is approximated to be 12.35. There occurs two
period doubling bifurcation, this results in four periodic solutions which are
asymptotically stable for (12.35, 14.06). Chaos occurs if > 14.06.

Exercise 3
Consider the following difference equation:
y(n + 1) = h (y(n)), n 0,

(2)

where the function h : R R is defined by

1
h (x) = x3 + x2 + (1 + )x,
2
where R.
a. Compute for all R the real-valued equilibrium solutions of (2) and
examine their stability.
Solution.
The substitution of h in (2) is given by:
1
y(n + 1) = y(n)3 + y(n)2 + (1 + )y(n), n 0, R
2
Substitute y(n + 1) = y(n) = c to find the equilibrium solutions gives:
1
c = c3 + c2 + (1 + )c
2
Solving this yields the following fixed points:
p
p
c1 = 0, c2 = 1 + 1 + 2, c3 = 1 1 + 2
So, the real-valued equilibrium solutions are:
y1 0, for R
p
p
1
y2 1 + 1 + 2, y3 1 1 + 2, for
2
Theorem (1.2.5) can be used to determine the stability of the equilibrium solutions.
yi for i = 1, 2, 3 is (locally) asymptotically stable if |h0 (yi )| < 1
yi for i = 1, 2, 3 is unstable if |h0 (yi )| > 1
Calculating |h0 (yi )| for each equilibrium solution gives
|h0 (0)| = |1 + |
y1 0 is asymptotically stable if 2 < < 0 and unstable if > 0 and if < 2
p
p
|h0 (1 + 1 + 2)| = |2 + 1 + 2|

y2 1 + 1 + 2 is asymptotically stable if 21 < < 0 and unstable if > 0

p
p
|h0 (1 1 + 2)| = |2 1 + 2|

10


y3 1 1 + 2 is asymptotically stable if 0 < <
1
2 < < 0 and if > 32

3
2

and unstable if

b. Plot the bifurcation diagram of (2) for (min , max ) = ( 52 , 52 )

You can adapt your program of 2.d. for this. Provide this adapted program!
Describe as accurately as possible the bifurcations occurring in your plot. So,
indicate all bifurcation points and explain for each one what the bifurcation
entails. If you know the name of a particular bifurcation, then do mention this
name.
Solution.
For the Matlab code see page 23.
For the resulting bifurcation diagram see appendix figure 11 (page 24).
In the case that h (x) = ( 12 )x3 + x2 + (1 )x, and x0 lies between 2.5 x0
2.5. It can easily be seen that indeed the null-solution which is calculated in 3a
between 2 < 0 is an asymptotically stable equilibrium solution. The nullsolution is unstable if > 0, this can be seen by the transcritical bifurcation
which occurs at = 0. The stability of a fixed
p point is transferred to another
fixed point. This fixed point is equal to 1 (1 + 2) and is asymptotically
stable if 0 < < 32 and unstable if 21 < < 0 and if > 32 . Moreover, there
occurs a period-doubling bifurcation at > 32 . In addition, if 1
2 < there
occurs another
bifurcation
in
the
diagram.
This
has
an
asymptotically
stable
p
<

<
0.
It
is
unstable
if

>
0,
this
can
be
solution 1 + (1 + 2) for 1
2
seen by the period-doubling bifurcation after > 0. This will eventually lead
to chaos.

11

Appendix
Codes and plots from Matlab are given below
Exercise 2a

12

13

Figure 1: Plot for = 2, p0 = 5, n = 150

Figure 2: Plot for = 7, p0 = 5, n = 150

14

Figure 3: Plot for = 8, p0 = 5, n = 150

15

Exercise 2b

16

Figure 4: Plot for = 0.5

Figure 5: Plot for = 4

17

Figure 6: Plot for = 10

Figure 7: Plot for = 13.5

Exercise 2c

18

19

Figure 8: Plot for min = 0, max = 12, p0 = 0

Figure 9: Plot for min = 0, max = 12, p0 = 5

20

Exercise 2d

21

Figure 10: Plot for min = 0, max = 16, xmin = 0.1, xmax = 5
Exercise 3b

22

23

Figure 11: Plot for min = 2.5, max = 2.5, xmin = 10, xmax = 10

24