Professional Documents
Culture Documents
Answers to all exercises are given here. In the case of purely numerical exercises
only the final resiilt is given, while in other cases to lielp the reader better to
understand the technicalities, a full derivation is provided.
Chapter 3
3.1.a)
(i) yt = A(0.5) t + 20; A = --20.
(ji) y =
t A(-2) t + 15; A = 3.
t
(iii) yt = A (-1) c+ 45; A=5.
(iv) yr = A () + z,; A=
3.1.b)
y
(i) = 20, monotonic convergent movernent.
(i) J = 15, oscillatory convergent movement.
(i.ii) y = 45, oscillatory witli constant amplitude movement.
y
(iv) = 5, oscillatory divergent movement.
3.2. S
(i) y t = A(2) t + 2 + 2t; A = 1.
t+1
(u) y t = A(3) t + 4 ; A = 4.
t
(iii) y
t = A( ) + (cos 4 t + sin 4 t) ; A=
T
3.3. Starting from Eq. (3.46) of the text, we can see that . w t = wo k, but
w 0 = bo/b = bo having assumed b = 1.
Performing all the passages we obtaiii
co Ti
r ). ca
o sa
To evaluate this expression, let S be the quantity of,
an
S=k+2k2+3k3+...
4.^
then ula
kS=k2+2k3+3k4+...
and
S-kS=k+k`'+k3+... =k(1+k+k2+...) _ k
so that
(1 - k ) S=
k
(1-k)
or
'1
8= k
( 1 - k) 2
,es Tllus we obtain the mean l(1 9 for the Koyck distributed lag model
to
E (1) = (1-k)
We know from the text tliat the Koyck distributed lag equation is equivalent to
a partial adjustrnent equation with a = (1 - k) and tliat the adjustment speed is
a-1
inversely related to tlie mean time lag which is 1 - 1 _
tx a
Chapter 4
4.2. Rewriting the model with the new assumption we have
S t = sYt,
/+J tr
l k r
t = (1 c - t-1) >
S t = It .
Suhstitutioti from tlie two first equations into the third one gives
Yt (s-k)+kYt - i = 0.
Y=Y
c o k k t -^' 1 S
s
t
.
'I'here is growtll if s/( k - s) > 0 This liolds only for k > s. Iil tlie affirinative
case, it is easy to see tliat s/(k - s) > s/k. The reasorl is t.11at, when income grows,
(tlie sarrie fraction
saving based oil current iI1coIIle is greater than saving based oIl
of) lagged income, hence (in equilibrium) investment is correspondingly greater
and income grows faster.
4.3. If la'yI < 11 + a'yI Nve apply forinula (3.37) which yields the following partic-
ular soltltion
pe= 1 ^^ ^ti
Answers to Exercises 613
In the opposite case, i.c. a ry l > 11 + a yJ We apply formula (3.39) and obtain the
particular solution
I 1 + o:-^ f
mt f-i
i =1 a
Cr -1
Pt = Po Pe
a
Ia I
Since a < 0 , it follows that 1 > 1. This makes the rnovement tinstable.
a
Solving forward we obtain the particular solution
a i
i_ 1 a-1
4.6.a) In this case investment is constant. Hence, after the appropriate substitu-
tions, the income equation becornes
Yt=a+bYt+cYt_1+Io,
therefore
(1 b) Yt cYt i = a + Io,
whose solution is
c Y+ a+10
5
Yc = Yo
1--b 1 bc
The model is stable if the marginal propensity to save in the current period, i.e.
1 b, is greater than the marginal propensity to consume referred to the previous
period.
4.7. Since demand arad supply are evaluated at time t, we have to take p t = Tr
pt+1 pt. The solution of the model is
t
aal
l a
Pt =p
o[ + ( bb
i)] Ap '
roc
in the
If (b - b 1 ) < 0 (which is always the case when the demand curve is downward
sloping and the sul ) i>ly curve upward sloping, narnely b < 0, b 1 > 0) this condition
requires
2
a<b,-b,
i.e., tlie adjustinent speed inust be lower than the critical value 2/(br - b). If, on
the contrary, (b - b l ) > 0, the stability condition can never be satisfied and the
equilibriurn is unstable.
able.
Chapter 5
5. a)
y t
(i) t = (1.55) (A 1 coswi A 2 sinwt) 38.'16;
wherecosw = 0.26 andsinw = 0.96.
. (ii) yy t = (0.77) t ( A 1 coswt + A2 sinwt) + 2.200;
;itu-
wherecosw = 0.71 andsinw = 0.71.
i (ii.i.) h - (0.5)' (gi l coswt. + A 2 sinwt) + 180;
where cos w = 0 and sin w = 1.
t
(iv) J t = (2.37) (A 1 c:os wt -}- A,, sin wt) ;
wherecosw = 0.97 and sirl w = 0.24.
(y) yt = (A1 cos sin wt) + 500;
wt + A2
wherecosw = 0.9 andsinw = 0.43.
ami
(i) , = A 1 (2)' 2 A = 1 . A2 = 2.
Z
t
i.e. (u) Y t = A l (3) + A 2t (3)'; A l -_= 3, A 2 =
)us
Chapter 6
6.1.a) After the introduction of taxation in Samuelson's model, we obtain the
following difference equation for income
Yt -b[(1+k-)(1-T)jYt-l+bk(1 -T)Yt-2-G=0.
L The characteristic equation is
-\b((1 +k)(1-T)J+bk(1-T) =0.
Applying Descartes' tlieorem ori the sequence of signs we can exclude negative
roots and hence improper oscillations. vloreover, we know that the roots will have
modulus less than 1, if and only if, the stability conditions (5.19) hold. In this
case they are:
The first inequality, given the assumptions made on tlie parameters, is certainly
satisfied; the third is also satisfied being a sum of all positive quantities. The
crucial inequality is then the second one, which can be rewritten as
b< 1
k(1-T)
We can see that the introduction of the taxation parameter makes the stability con-
dition easier to meet. Taxation (as a proportion of income) makes the model more
stable because it reduces disposable income and hence consumption (on whose
changes induced investment is based).
Therefore
Yt = b(1+k)Yt _ 1 -bkYt _ 2 +G/(l -T),
whose homogeneous partand hence the stability conditionis the same as in
equation (6.6) in the text.
In the case in which there is a lag between tax receipts and expenditure, the
stability conditions become:
Yo=
Ao(1+,q)2
(1-1-g)2--(b-}-f) (1-i-g) -f - k
Ao(1+9)2
Y
(1+g)t.
(1+g) 2 - (b+k)(1+g)+k
i nly no real roots. In sucli a case f (A) > 0 because it entire-ly lies abo ye the horizontal
rhe axis, lience (1 + g)' (b + k) (1 + g) rt k > 0.
6.2.b) F rorn tlie previous exercise we know tliat tlie characteristic equation of
tlie corresponding liomogeneous equation coincides with the denorninator of tlie
particular solution when A = (1 + g) . If the roots are real, the parabola f (A)
on- intersects the horizontal axis at A = A l and 1\ = a 2 , lying below it for A l < A < )2
y
ore and abo e it elsewhere. Hence (1 + g) 2 (b + k) (1 + g) + k> 0 for those values
ose of g such that (1 + g) < A l or (1 + g) > A2.
1 (b+k)+k = 1b>0,
in
1 k > 0,
he
1 +(b+k)+k > 0.
The first inequality is satisfied since we assume a propensity to consume smaller
than iunity; the tliircl is satisfied since the left-hand side is the sum of positive
r
magnitudes. fhe crucial inequality is then the second orie, and so tlie stability
condition in this model is
k<1.
In Samuelson's model the stability condition is, as we know, bk < 1. iNow,
since b < 1, the stability condition k < 1 is more stringent. Hicks' model is then
intrinsically less stable than Samuelson's model. This is not surprising, since in the
former, induced investment depends on the variations in total demand, whereas in
the latter it depends on the variations in consurnption demand, which are evidently
smaller.
2 2
6.2.e-f) Let us corisider the parabola f (k) = k (4 2b) k + b , and let k l and
2 2
k 2 be the roots of the equation k (4 2b) k + b = 0, which are
k 1 , k2 = 1
{4-2b((4-2b) 2 4b2]2}.
2
Simple manipulation yields
0<1- < 1, so that (1 - ) 2 < 1. The following table shows the complete
results.
If Roots Regions of Behaviour of the deviations
Fig. 6.2 from the trend, as t -a oo
t [Yt - (b + k) Y_ 1 + kYt _ 2 ] = 0,
Y
o (1+g)
we can rewrite it in the following way
k
Yt b + k Yt-1 + Yt-2 =0
nit-1 2
Yo ( 1 +0) t
(1 -o)Y,(1._j- (
, ^..1 V !, _^_ \t-2
,
or
D
b+k D
k D
t - t-1 + t-2 = 0,
(1+g) (1+g)
where D t - Yt /Yo (1 + g) t represents relative deviations. Applying the stability
conditions to the last equation, we find that the crucial condition is k/ (1 + g) 2 < 1.
It follows that in the opposite case, i.e. when k/ (1 + g) 2 > 1, D t will be divergent
anyhow. Note that with empirically plausible values of k and g, the instability
condition is certainly satisfied, and so Hicks' assumption (that in his model relative
deviations are explosive) is not unwarranted.
6.3.a) After the introduction of forcign trade into the model we obtain the follow-
ing second order difference equation
kY t t
Yt - (b + k - rn) Yt-1 + t-2 = Ao ( 1 + g) + Xo (1 + gz) .
Ao (1 +9)2
H0
ns (1+g)` - (b-}-k-ru)(1+y)+k.
Xo (1 + qx)2
0 =
(l+g ) - (bk-rn)(l+g,)+k
x
2
1 (b+krn)+k = 1 b+m> 0,
1 k > 0,
l +(b+km)+k > 0.
The first and third inequalities are always satisfied given the assumptions we made
on b and m (respectively b < 1 and m < 1). The crucial inequality is then the
scoon d on e, i.c. k < 1, whicli is the same as in the case of closed economy,
because the introduction of imports and exports does not influence the accelerator
coefficient.
G.4.a) After introducing the new assumptions we obtain the following difference
equation
ty
1. b
Yt- ( + k
- m
, -
b m
2 ) c-1+( -
k Y k m
2 ) t-2 = ( -
k Y 1 m
2) A
o( +9) + o( +9x)t
1 t X 1
nt
ty As in the previous case the known functions of time is a sum of exponential func-
re tions, lience we can try as particular solution
v= Ho(l+g)t+Bo(1+gx)t.
Ao (1 - m 2 ) (1 + g)2
H
^ _ (1+9)2[b(1 m1) +k(1m2)](1+9) +k(1 m2)
r Xp(1+ )2
0
B (1 + g^) 2 - [5(1 - 7n l ) + k (1 - rn
2)] ( 1
+ gx) +
k (1 - m2)
Answers Lo Exercises 619
As in the previous case the crucial stability condition is the second one, which is
now 1 k (1 m 2 ) > 0. This is less stringent than 1 k > 0 as (1 m 2 ) < 1.
6.5.a) Starting from the non homogeneous difference equation given in the text,
we see that the known function is a constant, i.e. a l a. We can then try as a
particular solution
P=C.
i
After substitution in the non homogenous difference equation we get
We can now consider the homogenous part of the difference equation which is
SI
b
pc b ( 1 + p) pt-1 + b 1Ppt-2 = 0.
ti
Dividing through by b we obtain ir
Pc bl (1 + p)
Pt-1 + b1P
Pc-2 = 0, 6;
b b
al
which has the associated characteristic equation
_ b1(1+P) ^+ biP
=0.
b b
A^
6.5.b) When p> 0, the discriminant in
1/
[b, (1
+ P) 2
blp M.
= b 4 b
is always positive given that b l > 0 and b < 0. Moreover, since the succession of
the coefcient signs is + + , by Descartes' theorem we know that the roots are Nc
bi(1+p) 2 1b1p
is :
bl 4p
th(
I b C
(1+p)2
C?'' (^
620 1tila:; 1
1 _ b l 1-f-P + -h p bL
> 0,
b11 -2
ichis 1 +b1 6 + =1-+- > 0.
1.
For p > 0 the first and second inequalities are always satisfied and so the
crucial inequality is the third, which may be written as
text,
asa 1 bl
b (12p)
Let us now compare the lasa inequality with the stability condition holding in
the original cobweb theorem, which is, in the case of normal demand and supply
functions,
L.
h is b < 1
Since 1/ (1 + 2p) is smaller than 1, the new stability condition is more restric-
tive. `chis last result confirrns that extrapolative
P expectations
P are an element of
instability.
(5 . 1) Por n < 0, the first inequality of the stability conditions is always satisfied,
_
and so the remaining two are the relevant ones. The second may be written as
bl
<.
1
b
p
As for the tllird, two sub-cases must be distinguished. If p < 1/2, then such
inequality is always satisfied, and so the only relevant inequality is b 1 /(b) <
1/(p). If p > 1/2 , then 1 + 2p > 0, and the inequality under consideration
may be written as
bl < 1
of b 1 2p
re Now, we must check which inequality is more stringent, which is equivalent to
checking whetlier
1< 1
p>1+2p
is in the range 1/2 <p < 0. From the inequality 1/(p) < 1/(1 + 2p) we have
1 + 2p
therefore
3p^-1,
Answers to Exercises 621
and
1 ^_1
P^ 3^
so
1 < 1 <1
for p>- 3.
._ p > 1 + 2p
1
The results for p < 0 are shown in table (6.2). Let us now compare, as before, the
crucial stability condition with that holding iri the original cobweb theorem. The
result is that, if -1 < p < 0, then the new stability condition is less restrictive
than b 1 /( -b) < 1; if p = -1, they are the same; if p < -1, then the new stability
condition is more restrictive than the original one. In fact, for -1/3 < p < 0, 7
the stability condition is b 1 /( - b) < 1/(1 + 2p), which is less restrictive than the
original one, since 1/ (1 + 2p) > 1. For -1 < p < -1/3 the stability condition is
b 1 /(-b) < 1/(---p) and, since 1/(-p) ,, 1, it is less restrictive tlian the original
one. For p < -1 the stability condition is again b 1 /(-b) < 1/(-p) and, if p = -1,
then 1/ - p < 1 and so b 1 /(-b) < 1/(-p) is more restrictive than the original
one. The economic meaning of these results is as follows. The fact that producers
expect price to reverse its movement is an elernent of stability, provided that the a
expected inversion is not too great, since in the opposite case it would have the
contrary effect, with the result that the stability conditions would have to be more
stringent than in the original cobweb theorem (whicli is tlie case when p < -1).
Chapter 7
7.a) sc
t t t
(i) yt = A l (2) + A 2 (1.45) + A 3 (0.55) + 500;
where A l = 37.5, A 2 = - 88.19, A 3 = - 49.3. If
(ii) y t = ( A 1 + A 2 t) ( 1 .5) t + A 3 (0.6) t + 1000; b1
where A l = -511, A 2 = 184, A 3 = 667.
7.b) 8.
i y
( ) = A l (1.41) t + A 2 ( -1.41) t + A3;
t di
where A 1 = 4.13, A 2 = -0.13, A3=-4.
t t
(i ) Yt = A l + A 2 ( -1) + A 3 (2) + A 4 ( -2)t
where A1= 11 3
, A2=_,A3=_,A4=_1
3 6 6'
Yt = (0.70) [(A 1 + A 2 t) coswt + (A 3 + A 4 t) sinwt] ;
A
(iii) t
W.
Chapter 8 ec
11
8.1. The proof is the following. Assume that Yt-1, Yt-2 and Yt - 3 Nave been
t 1
equal to the corresponding values of the ceiling, i.e. Yt - 1 = Bo (1 + g) - hE
t-2
Bo (1 + g) and Yt-3 = 13o (1 + g) t-3 . Substituting these values into Eq. (8.17)
oz :^ti:;,ver., ^^ ^:,.;^i^;t:3es
we have
t-1
Y
t B
0(1 + g) (b1 + k1) BO (1 + g)t-2 ( k2 b2 k-1)
k2B0 (1 + g)t-3
= Ao (1 g) t ,
.s
is true. Simple manipulations sliow that the last inequality is actually satisfied if,
.e asid oiily if, Yo < B 0 . Iii fact,
e
[(j g 2 b k b k2 ]
e Bo {(1 + g) t (1 + g)t-3 + 1) ( 1 + g) ( 2 + 2 k1 ) +
) ( 1+ k
}
>Ao(1+c^)t^
hence, on the assumption that Yo > 0,
I3 0 Ao(1+g)3 =Yo>
(1 +g) (b1 k 1) (1 +g) (k2 + b2 k1 ) + k2
so
Bo>Yo.
y
If, on the contrary, Yo > Bo (i.e. the trend lies abo e the ceiling), then Yt would
be greater than B t and so income would continue indefinitely along the ceiling.
Y A
Yt b 1 Yt-1 b 2 t-2 = o ( 1 + g)t a.
Hi(1+g)t+ H2,
where 11 1 and H2 are undetermined constants. Substituting into the difference
equation ave obtain
t-2_
t t- 1 b1H2b2H1 ( 1 + g) b 2 H 2 = Ao (1 + 9)ta,
H 1 (1 + g) +H2 biI i (1 + g)
hence
1
rg) 2 - ) (1+ 9 ) b2 ] +H2 (1blgi 2 ) =Ao(1-+-g)ta,
jI1(1+ !J) t - 2 [(1
1
Answers to Exercises 623
therefore
c- 2
+ { a+H (1-b - b )} = 0.
{H l [(1 + g) 2 - b 1 (1 + g) - b2] - A o ( 1 + g ) 2 } ( 1 + g) 2 l 2
This equation holds for any value of t if and only if both expressions in curly
brackets are zero, i.e.
+'g)2 b, (1 + g) g)2 =
H1 [ ( j b2 1 - Ao (1 + 0, a+H2(1-bl-b2)=0.
Thus wc get
Ao (1 + g)2 a
H1= (1+9)2 - b 1 (1 +g) -b2,
H2= - 1 b1 b2'
and therefore
Ao (1 +
g)2
Y
t= 2
(1+ 9 ) - b l ( 1 + g) - b2 b1 2
is a particular solution of the basic difference equation of the inodel when I _ -a.
8.3. To show that in the descending phase the movement converges to the lower
li mit (or `floor'), given by the particular solution, we have to solve the homogeneous
equation, i.e.
b Y
Yt - b,Yc-i - 2 c -2 = 0.
2
-blA -b2 =0.
Since the succession of the coeflicient signs is + - -, both roots will be real, one
positive and the other negative. Given that b i +b 2 = b < 1, the stability conditions
(5.19) are certainly satisfied, hence the movement will be to the floor.
Putting this together with the results of the previous exercise, we note that
the lower limit is seen to be equal to tlic output corresponding to the multiplier
2
1/[(1 + g) - S i (1 + g) - b 2 1 applied to autonomous investment, less the output
corresponding to the multiplier 1/(1 - b l - b 2 ) applied to (the absolute value of)
depreciation. This has an economic meaning. In the descending phase, total 1
investment equals autonomous investment less induced disinvestment, the latter
being at worst equal to the amount of depreciation; therefore income cannot fall
below the minimum level obtained applying the multiplier to that. difference.
We note from the particular solution tliat the floor increases at a proportional
rate approximately equal to g. The `approximately' is due to the fact that, owing
to the presence of the terrn - a /(1 - b l - b 2 ), the actual rate of growth is slightly
different from g. al
In the descending phase, income moves toward the floor. But since the floor is T
increasing, in the course of the approach to it income must sooner or later begin
to increase. At this point the accelerator comes back into gear and brings about
O4 :'.' :'erg;
curly 8.4. Considering the general forni of Hicks' model in which investment is evenly
distributed over n successive periods, i.e. k l = k 2 = .... = k n = (1/n)k, the income
equation becornes
0. k k t
Yt bl + Y-r b2Y1-2 b 3 YY-3 ... b n Y -n + Y-n-1 = Ao (1 + g)
k
A ni _ bl + \m-1_b2^m-2bn,\+-=0,
7t n
Y b p
t ( + 2 ) Y_1 b(1-2p)Yt_2bpYt_3 =II.
ine We can now apply the stability conditions (7.19) on the coefficients of the equation
ins
1 b(p+2) b(1-2p) bp > 0,
.at 3b(p+2)+b(1-2p)+3bp > 0,
er
1 +b(p+2)b(1 2p)+bp >0,
2
ut (bp) + b(p + 2) bp + b(1 2p)+1] > 0.
8 [-
ff)
al
After simple manipulations the inequalities can be written, for p> 0, as
er
1 -3b > 0,
L11
1 +4bp+b > 0,
3b > 0,
al 2
1 +b+2b p-2bp > 0.
ig
Given the assumption we made on the parameters, the last three inequalities
are always satisfied. The crucial inequality then is the first one, 1 3b > 0.
ts Therefore the system will be stable if, and only if
b< 1
3
K
Answers to Exercses 625
1 -3b > 0,
1 +b > 0,
3b > 0,
1 +b > 0.
Thus the crucial inequality is still the first one.
8.6. In Metzler's model, when p = 1, the stability condition becomes more restric-
tive. In fact, the first inequality in (8.9) is unaffected. The second one becomes
( 1 +k)(2+k)b2-3(1+k)b+1 > 0.
It is satisfied for b < b 1 and for b > b 2 , where b 1 , b 2 are the roots of the equation
(1 + k) (2 + k) b 2 3 (1 + k) b + 1 = 0. It can be checked that such roots are real,
distinct and both positive. The inequalities (8.9) are then satisfied in the following
intervals
b < 3
2k+3
b < b1,
b > b2.
Consider now the parabola f (b) _ (1 + k) (2 + k) 52 3(1 + k) b + 1. From what
we have said abo y e it follows that f (b) > 0 for b < b l and for b > b 2 and
f (b) < 0 for b l < b < b 2 . Substituting in f (b) the value b* = 3/ (2k + 3), we
see that f (b*) takes on a negative value, and so b* les between b 1 and b 2 . The
interval b > b 2 must then be discarded, and, since the inequality b < 3/ (2k + 3) is
absorbed by the more stringent inequality b < b 1 , the latter is the crucial stability
condition. Substituting b = 1; (1 + k) in f (b) upe see that f (b) < 0, and so
b l < 1/ (1 + k) < b 2. . This proves that the inequality b < b l is more stringent then
the inequality b < 1/ (1 + k) .
8.7. In the case in which inventories follow a partial adjustment equation, the
income equation is
Yt = Ut + a (Q Qt_1) + Io.
9
After the substitutions, we obtain the following third-order difference equation 1
Ytb{1+p+a[k(1+p)+1]}Yt_1+b{p+a[kp+(1+k) (1+p))}Yt -2
ba
p( 1 + k ) Yt- 3 =Io 9
1 b{1+p+a[k(1+p)+1]}+b{p+a[kp+(l+k)(1 +p))}
bap (1 + k) > 0, u,.
1 + b{1+p+a[k(1+p)+1]}+b{pa[kp+(1+k)(1 +p)]}
+bap (1 + k) > 0,
3b{l+p+a[k(1+p)+1}} b{p+a[k.p+(1+k)(1 +p)J}
+3bap (1 + k) > 0,
2 2 2 2
-b a p (1+ k) 2 +b pa(1+k-){1+p+a{k(1+p)+1}} w1
b{p+a[kp+(1+k)(1 +p)J}+1>0.
U2^ :.:,Ter:: to . ises
Tlie first asid second iriequalities are always satisfied since the propensity to cori-
sume is srnaller than unity and the expectation coefficient is nori-negative. The
relevant inequalities are then the third and the fourth, that can be written as
restric- From these inequalities we can see that if p = 0, the stability conditions become
ornes
b
< 1+2a 1-}-k
b < a(i+k ,
cation
that are both satisfied for sufficiently low values of a.
'e real,
When p > 0 it will also norrnally be possible to find sufficiently low values of
,erial
a such that the stability conditions will be satisfied. In fact, if we let a > 0 the
stability conditions tend to
the
Chapter 9
9.A) Since the coefl^icients are non-negative, we can apply the stability conditions
n I of Sect. 9.2.2, Eq. (9.65), wliicli sliow tliat the system is stable.
9.B)
(i)
y ll t 2) t
it = A l ai (3) + A 2 ai (2) 3
= A1a2 (3) + A2a1 (2)t
t 2)
1) 2)
wliere al = 3, al = 2,a 22 = 1,a 2 = 1.
(22)
11 t
Jit = A 1 ai (14) + A 2 ai
2
^ (0.95)c
1) t 2)
Y2t = A1a2 (1.4) + A2a2 (0.95)t
wherea = 1,a(2) = 1,a2 1) =0.2,a2 =-1.1.
I 1
Answers to Exercises 627
(iii )
y = Ala, 13 ! t + A2ai2)
(* t + A3ai31 ( * ) t + 7
lt ) ( )
i
21 t
Jet = A 1 a 21) (2) t + A 2 a2 C-7ll + A3a23 +3
t ( i) (i)t
Y3t = A 1 a3' (0.67) + A2a 2) + A3a33) +1
where = 0.5, a1 2) = il, ai 3) = 0.66,a = 0.5, a(2) = 3.4, a2 3 ^ = 0.59,a'
(2) (3)
1 a3 = 1 a3 = 1.
(iv)
A t
Yit = ( A 1 + 2 ) (- 1)t (a at
iJ 2t = (A1 + A t
2 ) ( -1 ) t + a-+11 at
9.C) Performing the matrix multiplication ice see that the first (n - 1) rows define z
the new variables, and the last row expresses the equation in terms of the new l
variables. r
Chapter 10 T
10.1. The other root of Eq. (10.6) is 1/2 with multiplicity (n 1) . In fact, if
we put \=1/2 in the characteristic matrix, we obtain a matrix with all identical
elements. It is well known in matrix algebra that the determinant of a matrix
which has two lines (rows or columns) equal is mili. Iri our case this will be true
Ni
for all minors up to th_- la.->t onc of order 1. He rice. the matrix undcr considera ion
has rank 1 and degeneracy (n 1). This shows that the root 1/2 has multiplicity all
(n 1). an
We know that, in the case of a multiple root, this vill appear in the general
solution multiplied by a polynomial in t with degree equal to the multiplicity of
1 t
n-2 as
the root less o ne, say (A + A 1Z t + A2z t2 + ... + A n _ 2,t t ) 2 However, due to
po,
the extreme degeneracy of the matrix, the coefficients of all the non zero powers mo
of t identically vanish, so that the general solution turns out to be of i
t
1 1 n t +anci +^^ ^ ci
X1t A Bz z =1 2 ... n ,
= 2 + 2 (rl + 1) b
where ^Z A l = 0, B Z = B, with (n 1) of the A Z and B arbitrary.
For further detail see McManus (1962) .
10.2. Under the assumption that each country imports directly or indirectly from
all other countries and remembering that b i + h i m > 0 because m is part of
b 2 + h i , the matrix
m ..
(bl + hl m i ) 12 mi n
m21 1
(b2 + h2 11 2) ... 7712n
M
nl mn2 .. ( h n + /Z n m.n)
is a non-negative ndecomposable inatrix. In this case we can apply the stability
condition V given iii Chap. 9, Sect. 9.2.2. Thus we have
n
(bi + h i in) -{- rn^ i < 1,
j=1
7v-2
at least orle strict inequality.
1)
b + hZ < 1
Ylt = (b i +/t m
l) it-1 + n 2 (1+h3)Y2t-1+(h3Mo 9 M01 +Mo 2 )
Y
rrt 11, M
Y 2t = ( b
2 + 11.
2
2) Y
2t-1+ m
1 ( 1
+ /1
4) Y
11-1+( 1 01 MO2+M01).
1 b2-112+m2 -711
ml
1
f
Y
Answers to Exercises 629
Chapter 12
12.1
(i) y(t) = Ac -z ; A = 1.
(ii) y = Ae-3c; A = 1.
(t)
6t
( ii i)
y (t,) = Ac ; A = 1.
(iv) y (t) = Ac t ; A = 1.
v y (t) = Ac-zt; A = 1.
( )
12.2.
(i) y (t) = Ac -t + 2t; A= 1.
(ii) (t) = Ae-3c + 2 - A = - 31
J
1 3 3
(iii) y (t) = Ac 61 - cos 2t - sin 2t; A= .
2 2 2
(iv) y(t) = Ae t +1+t; A=2.
(v) y (t) = Ae-2t + e-2t; A= 2
.
3 3
-at
12.3. The primitive of the integrand ae is - e -at , so the definite integral in*
-
the interval (0, co) is [-e ] + 1 = 1.
(aT + 1)
12.4. By using the hint, we know that the primitive is -eQTa2 , hence the definite
1 _ 1
integral in the interval (0, oo) is 0 -
(__) a2 a2
Chapter 13
13.1. Taking the adjustment mechanism into account, we obtain the following Z
first order differential equation:
P' =c((Q-dl)+(b-bl)P]
T
where x = ln x.
The general solution of the systcm is-
= Ae^ (b-b
1 )t
+ (a
h
- a)
b
of
( 1) J 1^^
vvv n_, j' r.._
Tlie systern will be stable if, and only if, (b b 1 ) < 0, being c > 0. This is certainly
satisfied in the normal case (b < O, b 1 > 0).
13.2. After tlie substitution, we obtain tlie following differential equation
(1 cc
1 b l b
l) b f b 1^'
al a
p (t) = Ac b -}-
bl b
The equilibrium is stable if, and only if, b ( l b) /b < 0. If b < 0, the condition
holds for b i b > 0, which is thc usual one.
13.3. After the introduction of expectations in the previous exercise, the differen-
tial equation becomes
bbl al a
+ _ P blc
b b1c b
L
L t
al a
p
(t)
Ae b -}- bb1
,
al in'
The equilibrium is stable if, and only if, (b b /(b b i c ) > 0. This condition is 1 )
satisfied only if the numerator and the denominator have the same sign. Then, in
inite the normal case, b b l < 0, so that the crucial stability condition is b b i c < 0. If
c > 0 (i.e., extrapolative expectations), the condition is always verified. If c < 0,
the condition liolds only if b > bici.
13.4. Substitiit.ing the definitional cquations for supply and demand in the equi-
librium condition and rearranging terms, we obtain
^x (s)] a 1 a b1
b
Ax (s)
Substituting for p(t) in tl i e expectation equation, we have
^' ( t ) = 13 al
a +--j5(t)
1
P (t)
b b
13 b l b j^ (t) + j3 al a
b b
A particular solution is obtained setting 13' (t) = 0, whicli gives (a ax) /(b l b) .
ing This is the same as pe, the static equilibrium price. The general solution is
b
whicli is the same as in exercise 13.2. Given the equation written at the beginning
of this exercise, the behaviour tlirough time of p (t) will he the same as that of
j) (t.) .
Answers to Exercises 631
13.5. In equilibriurn, the iiicrement iri oiitpiit is equal to the incremcnt in aggre-
gate dernarid
Y'=C'+I'.
Assuming the simple consumption fiinction C = (1 s)Y, we have
Y=(1s)Y+I,
which gives
1
Y' _ I' .
Assuming that in the initial period productive capacity is fully utilized, for Chis
situation to be continuad over tiple the increase in potential output must equal
the increase in output, i.e. Y' = P', where P' = al is the increase in potential
output brought about by investment. From P' = al, Y' = sI' and Y' = P', we
have
I' =sal,
that expresses the relation that must be satisfied by investment for productive
capacity to be fully utilizad ovar time. The soliition of the differential equation is
I(t ) = Ioesat,
where Io is the initial level of investment. Investment must grow at the constant
rate sor. It is easy to see that, if investment grows at that rate, output will grow
at the same rate. Integrating with respect to time both sides of
Y
' _ I^>
I.
we have
sY=I+A,
where A in an arbitrary constant, and given
I(t) = Ioesat,
we have T
sY = IocSat + A, rc
i.e.
1 A
Y= Ioe t+
s s
Assuming that we start from an equilibrium situation, we must have sYo = Io, so
that Yo = (l/s)Io and, consequently, A must be zero. Tlius we have Al
Y(t) = Yo esot,
and so Y also grows at the rata su. Therfore, income grows at a rata which is the
w1
sarne as the one which obtains in Harrod's modal (su = s/k).
tei
fiil.`:1'le s to 1 c C1scs
`^ gg re- 13.6. To calculate the tirase req u ired for 90% of tlie initial deviation of k (t) from
its equilibrium value to be elirniiiated it is possible to apply formula (13.54). We
obtain that t k, o .00 = 137.9. As far as r (t) is concerned, we first nave to find
the equilibrium value. With the given values of u, s and a, the basic differential
equation becomes
r' = 0.10x 1/3 - 0.025v
The substitution z = r 2/3 transforms the abo y e equation into
0.05 0.2
z-{ 3 z= 3 ,
this
whose solution is
. qual u.os t
z (t) = Ac 3 + 4.
ntial
we The inverso trarisformation r = z 3/2 gives:
from which
0.05
10.6e - 3 t
5.3467.
Using logarit}iins,
tr,o.90 =
3
r log 0.5044057 -- 41.062.
0.05
To calculate t, , o .00 we need yo and ye. Knowing the values of ro and r e and the
relations between y and r (see Sect. 13.2.2), we obtain
1 ,
y o _ (r o ) _ 3 .80,
Ye = (r) - 2.
;o
y (t) -2 =0.10(3.80-2),
1/2
where y (t) _ [10.6e- t
-t- 4] 1 substituting for y (t) and rearranging
terms we obtain
0.05
1 0.6c + ^- 0.7524.
Using logarithms,
tv,0 9O = 158.72
13.7. Applying the sa.rne proccclure as iii thc previous exercise, we obtain
t
k,O.50 = 26.25,
t
k,O.75 = 52.51,
t
r,0.50 = 11.06,
t
r,o.75 = 20.78,
t
y,0.50 = 20.63,
t
y,0.75 = 37.48.
1
13.8.a) Given the hypothesis on the labour supply, Eq. (13.33) becomes i
nt
K = rLoc h ;
13.8.b) Adopting a Cobb-Douglas production function, the real wage rate is equal Al
to ra
a
a a K
aY
=K (1c)L = (1cx)=ra(1a)=w.
L L p
a
Differentiating w/p = r (1 - a) with respect to time, we obtain C
w' c-l
=cx(1-cx) r r', 14
and so
w' w w' r'
p p w r
After a little manipulation, the fundamental dynamic equation can be written as 14.
r
' = [ s f (r, 1) - nr] (1 + (Yh) -1
which, as Solow (1956) points out, "gives l some insight into how an elastic labour
supply changes things. In the first place, an equilibrium state of balanced growth
still exists, when the right-hand side becomes zero, and it is stable, approached
from any iriitial condition. vloreover, the equilibrium capital-labour ratio is un-
changed, since r' becornes zero exactly where it did before. This will not always
happen, of course; it is a corisequerice of tlie special supply-of-labour schedule.
Since r beliaves rnuch iii the sane way, so will all those quantities which depend
orily on r, sucli as tlie real wage rate".
13.9. Startirig from tlie basic clifferential equation of the neoclassical growtli rriodel
r' =sf(r,1)-nr,
we can rewrite it as
nt
K'L - KL' K 1
L'K
Ll = sLoe f L ^^ c ^^t > IJ2
7L
-^ 1 - a)t i- .
K (t) = Kp - a _ S L -a + _ L e
n s
Chapter 14
14.1.
(i) y (t) = Ale o.62t + A2e - i.s2t + 10;
( 22 ) y
(t)= A l + A2e-t - 5t;
(iii) y (t) = Ale-1.76t + A2e-1o.24t + 3;
(iv) y (t) = e -t ( A l cos 7t + A 2 sin 7t) + 20 - 0.4 cos 47rt + 0.1 sin flirt.
as 14.2.
(i ) y (t) ` A le l.73t + A2e-1.73t; A l = -0.87, A 2 = 0.87.
-t 2
(u) y(t) _ (A 1 +A9t)e +6-4t+t ; A l = -6, A 2 = -1.
Chapter 15
15.1. After the introduction of the hypothesis on expectations in the standard
multiplier equation, i.e. Y = [1/(1 b)] ( K' + a), we obtain the following second
order differential equation:
1 aflka
1 a^3k-y 1 K" + ,Q nOk K' + a^3K =
1 b 1 b 1 b
The corresponding hornogeneous part is
K" + K' (1 b ak
) + ak ( 1 b ) _ 0,
1 bc43k-y 1 ba/3k7
(1bak) aOk(1b)
A2 + A+ 0.
1 b a/^k-y 1 b aoky
The coriditions for the roots to be negative if real, and have negative part if
complex, are
0(1bak) > U,
1 ba /3 k-y
n7 I. i \
> 0.
1 ba,Qky
The conditions are satisfied only if the numerator and the denominator in each
fraction have thc same sign. Hence, if
T
1 ba 3 k7>0,
then
(1 b ak) > 0 and apk (1 b) > 0
must hold.
Otherwise, if Gi
1 baf3k-y<0, 15
then
X3 (1 b ak) < 0 and ca,3k (1 b) <0
have to be satisfied. This is however impossible, since af3k (1 b) < 0 can never wh
be satisfied given that 0 < b < 1. Vence in the case in which 1 b a,8 k-y < 0 the
model is not stable.
In the previous case, the crucial conditions are the first two, being the third
one (i.e. a,3k (1 b) > 0) always satisfied.
From the crucial conditions, we obtain " Sub
s i s
a< ^^anda< .
k k
.^... 1lrlswer,; tu J :a i'C1SLS
We can note that the second inequality coincides with Eq. (15.11). Now, the first
inequality will be more or less or equally stringent tlian the second one depending
.rd on the value of ,3-y. We can now analyse the two different c as es, namely -y > 0 (i.e.,
nd extrapolative expectations) and -y < 0 (i.e., regressive expectations).
Case l:7>0
If 4?-y > 1 the case of extrapolative expectations is niore stringent than the case
of static expectations (i.e. 7 = 0).
If ,8'y < 1 the opposite is true, while for ft y = 1, extrapolative and static
expectations are equivalent.
Case II: 7<0
The condition 1 b a/3k'y > 0 is always satisfied, so that the crucial condition
is a < - , as in the presence of static expectations.
ae B cos wt
SR" (t) + b1 SR' (t) + al SR (t) _
b 2 mb2 mb2
^ 2 + ^1 \ + Q1
= 0.
h b 2 mb2
bl
b^ > 0'
al
m b2
>O.
Given that m> 0, a l < 0, the inequalities are satisfied only for b l <0 and b 2 < 0
15.2.b) To find the particular solution we try
This equation is identically satisfied if, and only if, the three expressions in square wi
brackets are all zero. Solving the system we obtain
15
al mb2w2
_ B ^ ^
C 2 2
(al mb2w ) + m2b1w2'
Bm.blw
2 Fo;
(al mb2w 2 ) Z + m2b1w2'
ao
C3 = --
al
To compare the particular solution with the element expressing the seasonal influ-
ences in En (t) , i.e. B cos wt, we can writc the particular soliition in the following
form. Let us make the substitutions A&
squ
Cl = C cos -y,
thr^
C2 = C sin -y,
from which
Cl cos wt, -i- C2 sin wt = C cos(wt + 'y),
where
C +(C1 + 2 ) 1/2 = B
(a l mb 2 w2)2 m2bw2
ao
SR"()t + mbl
+
mb2
f2
SR' (t) + al
+ f l SR( t) =mb2
mb2 fl al aa0 B cos wt wher
mb2 excha,
that can always be satisfied by a suitable -choice of fi, f2 .. For example, if b l >
0, b 2 > 0 (the basic model is unstable, see exercise 2.a), then
Again, the equation is identically satisfied if, and only if, the three expressions in
square brackets are all zero. Tlius, we obtairl a system of tliree equations in the
tliree unknowns C 1 , C2 , C3 . By solving it
a
fi b2nUI)2 )
( l+
C1 = 2)Z
(f2 + b1m)' w- + (a 1 + f 1 b 2
C2 = 13(f2+ bi n)w
m 2 w2
(f2 + b 1 ) + ( a l + fi b2mw2)2'
ao
C3 = --.
al
We can write the particular solution in a siinplified forin. Let us make the substi-
tutions
C1 = C cos -y,
C2 = C sin -y,
from which
CI cos wt -}- C2 Sil] wt = C cos(wt + 'y)
where
B
C = +(C + C2)112 =
(J 2
`
+ b1 m) w
+ (al fi b2mw2)' .
TIIe l>Olic:y l lr;lllters, f I all(l f', cause C: to cic(:r(sl;;e, so t.11c. ;u1iJ)Iltllde of tlle
1,11ctuatlon is sInaller thaii the oIle obtained in the previous part of the exercise. We
Cali coliClll(1(, tllat elle 1)resence of the Ino11('tary a11t11(>rli,l('S III tlie fOIel - Ii exc11,1.I1g,e
Tnarket will llave a stabiliziIlg Influence.
15.3.c) To find an explicit expression for the change in the authorities' foreign
exchange reserves we have to evaluate the following definite integral
I
E(t)dt = I
fl Si? (t) ( ao ) + f2 SR'(t) dt
fo Jo L L al
T a
= fi f[si? (t) ( ) dt + f2 (SR(t)]
c, L a1
A n swe i s Lo E xc i -c i ses 639
Assuming that the policy parameters have been chosen so as to make the model
stable, the homogeneous part will tend to zero, hence we can neglect it and consider
only the particular solution. Substittiting it in the previous expression we obtain
f 1 T
EG (t)dt = fi C l 1 sin wt - C2 cos wt
w
+f2 [C 1 cos wt + C2 sin wt ao/ai]
It is now easy to see that, if we take T = s(27r/w), s = 1, 2, ..., the last expression
is zero. This is also true for any interval t l , t 2 suclh tliat t 2 t 1 = T. Idi this mearas
that tlie chango in international reserves is zero over a full cycle (or multiple
thereof) of the exchange rate.
Chapter 16
16.1.
t
(i) y (t) = A l e -2t + e ( A 2 cos t + A 3 sin t) ;
(ii) y (t) = A e t + A2e3.8t + A3e-o.st;
l
(iii) y (t) = A l + A 2 t + A 3 e t + A4e-3t
16.2.
(i ) y ( t ) = 1+4t+lt2;
^Z ) (t) _ 98 23 t +2 3 t2 + I t4;
(iii) y (t) _ 1
1 t (3t + 8t + 28) + -e 2t + 2 (cos t + 2 sin t)
Chapter 17
17.1. In this case we must substitute G * = f fo Ydt fp Y in Eq. (17.12),
obtaining
t
Y" + (al +,3) Y' + a,Q (l + fp) Y + a ft Ydt = aa3.
lo 1
Differentiating S
By applying the stability conditions, i.e. Eq. (16.29), we can see that the crucial
stability condition is
It can be seen that the greater fi is, the more difficult it is to satisfy the inequality. i.E
)del effect of proportional policy. This aliows us to say that the joint use of derivative
der and integral policies reduces the probability of instability, i.e. explosiva oscillations.
am n
17.2. In this case we must substitute G* _
fi f" Ydt fd Y' in Eq. (17.12),
obtaiiiing
t
Y" + (al + 3 + aO fd ) Y' + af3 1 Y + a(3 f2 Ydt = a^3.
lo
Differentiating
Y " + ( al a Y
Q+ Qfd) " + 1Y' +a 3 fZ Y = 0.
ton
tns By applying the stability conditions stated in Eq. (16.29), we can sea that the
pie crucial stability condition is
n
( al + 13 + c3 fd) a^l of, >0,
Differentiating
By applying the stability conditions, i.e. Eq.(16.29), we can see that the crucial
stability condition is
(al+ +a,3fd)a,3(l+fp)aOfZ>0,
which is the serme as
(al +,3+a(3fd) ( l +.fp) > fz.
It can be seen that the bigger is the value of f, the more difficult it is to satisfy the
inequality. But deis destabiiizirzg effect of the integral policy is counterbalanced by
a stabilizing effect of botli the derivative and proportional policy. This allows us
to say that the joint use of the three policies reduces the probability of instability,
i.e. explosiva oscillations.
Answers to Exercises 641
I = 77(kY I),
Y' _ alY + al + aG a^c,
G' _ /3G+QG.
From the second equation we can find an expression for I. We then differentiate
it, obtaining an expression for I'. We can now substitlute these expressions in the
first fundamental equation, obtaining
G, = _ 1
Y // 1
i k+ l
Y'+lY+ 1G'+
arl a rl 77
Substituting for G' the expression given in the third fundamental equation and
rearranging terms, we obtain
G 1
+
77
=_ Ir"
a7l
1 _}- +
a
l
rl
Y'-+-IY+ G*+.
r)
By differentiating it with respect to time we get an equation for G'. Substituting
these values of G and G' iri the third fundamental e quation and rearranging terrns
we obtain the basic equation of the model:
al+7/arlk+/3 >0,
arll + (al + rl arlk) + af3 fp > 0,
a 3r l
f l( +fp) >0,
a3
(al + -q arlk + f3) [arll + 3 (al + r/ arlk) + fp ] afjrl (1 + fp) > 0. II
c'
We can see that the only condition that is certainly satisfied is the third. Stability
is not certain: in particular, from the first condition we get
that tlie ilitrocliictiorn of tl e accelerator hito Plrillihs' niodel malees the model itself
intrinsically less stable.
If we set C = f,Y f 1 Y' i i i tlie lbasic equation, we obtain the characteristic
equation
\ 3 + (nl + 7l -- Qr k
) + + a 2
fd) ,\ + (orgl +a[3 l + 7) cx f.377k + r/ fd + a fjl ff)
ate +cx/3rl(1 + fr,) = 0.
che
Applying stability conditions (16.29) we obtain
(l+rl--cxrlk+,3+c 3 fd > 0,
wI l + c,8 1 + f3 17 cx/3 7/ k + a/37] fd + ca/3f f > 0,
a /3
7( l + fP) > 0,
a a
(cxl + "r) cxrlk + /3 + / f) (cr7ll + 3 l 1
/ +, l a/377k + rx/i fd + cxflff)
a
/77( l + .fv) > 0.
As in the case of pure proportional stabilization policy, tlie only condition certainly
satisfied is the third one. In this case, liowever, stability is ensured because the
other conditions can always be satisfied by choosing sufliciently high values of fd.
This conf rrns the stabilizing influence of the derivative policy.
17.4.b) We set G* _ f Ydt in the basic equation of the model (found in the
previous exercise) and obtain
To get rid of the integral we differentiate with respect to time. Collecting tercos
we obtain
/\4 +(cal+77cx77k+/3)X3+(carll+cr/l+/rlca/rlk)/\2
1
+ cx /3( 77 + fi ) + a/rl fi = 0.
As iii tlie case of pure proportional policy, stability of the model is uncertain.
In fact, necessary stability conditions are the positivity of all coefficients. The
coefficient of ) 3 will be positivo if, and orily if,
This is satisfied for sufficiently low values of k; irr the opposite case, it will be
satisfied for sufficiently low values of rl. Since no policy parameter appears in this
inequality, we cannot be sure that the model is stable. The introduction of the
accelerator into Phillips' model rnakes the model itself intrinsically less stable, and
pure integral stabilisation polic y does riot cont riblite to the stability of the rnodel.
Answers to Exercises 643
ii = i " " 9t
Y -y1^Y + ryka l Y + ^yk a. 2 Y -ysY -r- }Io e ,
from which
-yka2Y" + (`ykai 1) ' + kY' _ rysY = ryIee9t.
As a particular solution we try
Y=ce-q,.
After substituting in the equation and solving for c we obtain
-yIo
C =
(7
ka
2g 3
+ (^ ka
i 1
) 9
2 + -ykg
It can immediately be seen that the stability conditions (16.29) are not satisfied
since --ys < 0, so that the model is unstable.
C hapt er 18
18.1.
y l (t) = e - 2 A l ar cos t +A l ai sin t ,
Y2 (t) = e - 2 A1c cos t + A2a2 sin t
The movement is damped oscillatory since the roots are complex conjugate
with negative real part.
18.2.
t = Ala1eo.dit + A2a2e2.41t,
J1 ( )
1/ 2
t
( ) = Ala2eo.4lc + A2a2e2.4it
18.3.
(i)
-t
1/1 ( t ) = e ( A l al cos t + A 2 ai sin t) ,
J2 (t} = e -t (A 1 a2 cos t + A 2 a2 sin t) ,
where A l1 1; A
A2 =0;a1-2
0; 1 51 a 2 - 2 1 a1=1;a 2 =1.
1 5 , l , 2
:z s if ;rC rcisz:s
we
(u)
J1 ( t ) = Alaie -t + A2a2c4t,
J'? (t) = A 1 ale- + A2a2e4t, t
where A l = 3; A 2 = 2; al = 2, a 3; a l
2 = 1 a 2 = 1.
2
(222)
J1 ( ) _ ( A l ai +
t
A2a
i ) etat
Y2 (t) -_ (
l + A al A 2azt) et,
where Al=0;A 2 =0;ai-=2;a?=0; a2=1; a2=0.
(iv)
_ cos t + 10 sin t jet+ 2 (cos t + 3 sin t) et
y
cos t 5 sin t 4 sin t t '
(y)
1
t - t2 - t3
y= 4 2 et-
t 2 + It3
18.4.
J1 ( t
) = e 2
Ala, cos 4t + A l ai sin 4t , et,
d Y2 ( t
) = e Ala2 cos 2 t + A2a2
5
sin 2 t + 3et.
5
18.5.
J1 ( t ) = A
l al e t + A2aie -t (A 3 a cos t + A 4 ai sin t) + 5e2t,
Y2 ( t
) = A l a2e 7+ A 2 a2e t -t
5 e2t,
+ ( A 1 a2 cos t + A 4 a2 sin t) 1
wilere .i1 = 4 ; .i2 =
1 ; , = --
1'13 2 JI4 = J; al, ai, c1, a = j; a a _
-; a2, a2 = -1
A2
- ( a 11 + a
22) A
+ ( a 11 a 22 - a
,2 a 21) = 0.
Herice, the derivatives of the characteristic roots with respect to the parameter a
are:
a) casing the standard formula
OA 1 _ 1 a ll aa, 1 '
) c -
`l i)cx + aa ) + z ` (al 1 -} a 22 ) ( ^ll
a + ad/
2 ( a il+ a 22) -i- 4 ( a l2a21-a11a22 ) L \ 3a aca
+2 (al2 (9a21
Da-i-
a^ au
a22 Q11
0121 jcr - aa - acx ^^ >
_1 c3u _ aa_>2 1
3a - 2 Da da ) - 2 a 11+a22) z-}-4(a12a21a11a22) [(all + a 22) ( a + _)
+2 (al 2 - a
+ a21.21 a
-0-22 c3
-X11 i
da )]
b) llsirlg the followiilg relations between the coefficients and the roots
A
l "f- A 2 = a
11 + a22,
/ A a a
`1 2 = l1 22 - a12a21,
Answers to Exercises 645
whence
Q_J. _ ( ^u +
a a
l_
da da dar l da
O_ ( ms
d
+a
d U
) ^ 1 A ^ 2 (a12 d + a21 d
C2. 22 e L
a
ll d 2
) 1^2
Chapter 19
19.1. After the introductiorl of price expectations, thc linearised dynamic model
is
m
P^ = k i
a Z p j +b +p j p^
k
i ( a 11 + bi ) k1 a
12 ... k1 al
k2a21 k2(a22 + b2) ... k2a27,
A=
kmaml kmam2 . k
m( a mm + bm)
which is stable, since it fulfils the Metzler conditions by assumption. When p 10,
the system can be written as
p = DAp,
d1 = (1
we can apply the results orl D-stability (see Sect. 18.2.2.1) and conclude that, if
A is Metzlerian-stable, DA is also stable.
PI = cbp cq + ca,
q' = kpbiq+di.
.\ s '.: ' er to Exercises
cbA c
k
k )
We can clieck the stability of tlie system without solving the characteristic equa-
tion. Wc know t.hat the necessary and sufficient conditions for the roots to be
negative if real, and Nave negative real parts if complex, are that the coefficient
of the characteristic equation are positive. We can then distinguish four different
del cases, depending on the values of b 1 and b, that are usted in the following table:
Parameter values Movement
Case I bl > 0 b < 0 Stable
Casell b 1 >0 b>0
a) b^ > cb 1 > b Stable
b) All other cases Unstable
Caselll b 1 <0 b<0
a) < cb 1 >- Stable
b) All other cases Unstable
Case IV b l < 0 b > 0 Unstable
19.3. Considering Eq. (19.63), we can see that Eq. (19.62) is the same as
Eq. (19.51), i.e. the characteristic equation of tlie physical system. Thus, since
,
q = r -- A, it follows that the characteristic roots of the price system are obtained
subtracting tlie characteristic roots of the pliysical systern from r. This allows to
tate . that +1 prices
^.^ the ces r rr^.^ronding
..o. csp the balanced
to ^..^. ^. growth path
p are constant if,
and only if, r = A1.
Let 71 1 = r A 1 . Then the prices corresponding to the balanced growth path
are those obtained \vIleIl p 2 (0) /p> (0) = a^ l) . It can be checked that when r = ^1
tliese prices are thie salve as tlie static prices. The system for the static prices
is homogeneous so that its determinant must be zero for it to give a non- triv-
f
ial solution. This deterrninant is the same as the determinant appearing in the
characteristic equation of the output system, with r instead of ); thus if r = A1,
then p i /P1 = Cl (. ' ^ . T his proves tliat the dynamic price system (19.61) does not
necessarily give rise to a changing price level on the balanced growth path. But if
r A A l (and t11is is surely the inost probable case), then the prices corresporiding
to the balanced growth path will be changing over time, converging or diverging
ac(or(hiIlg flti 7' C .^ 1 .
19.4.a) The dynamics of the model carl be analysed taking into account that the
output of sector 1 is equal to net investment, so that K' = X 1 . Differentiating the
equation
K = a 1 N1 + a2N2
with respect to time, we obtain
K^ = a 1 N1 + a2N2,
Answers to Exercises 647
N ' = Ni + N2.
Ni+N2=nNl+nN2.
a 1 N1 + a 2 N2
Nl = 0, b
i
Ni +N2 - nNi nN2 = 0.
19.4.b) To find a solution of the system we can either put it in normal form (for
this to be possible the determinant A =a, a 2 must not vanish), or apply the
l
al , bl a2)\
_ (a l a2) A2 (al a 2) n + ,l +
L 1 b,
^ T n
=0.
1
l = n; \ 2 = (al a2) bl
19.4.c) Once found the roots, we can write the general solution of the system
Using the initi al conditions, we can now determine the values of the arbitrary
bc
collstal lts 1111,1112,1121,1122
= eNo,
A 12 = 'V())dl ,
e --bi.,i
A 21 = ( 1 c)No,
A 22 = - o d = -Al2,
\ 2) where
alb1 n
C = 1 n(a l a2)bl
d
a'
19.4.d) From the soliltions for the arbitrary corlstarlts it follow that, if Ko is equal
to cNo/b i n, then A l2 and A 22 turn out to be zero. Thus in the solution only the
exponential term containing n remains, anc.l this means that N 1 , N2 (as well as
X 1 , X 2 , K 1 , K2, given the assumption of fixed technical coefficients), grow at the
sanee proportionate rate n (tlie growth rate of tlie labour force).
19.4.e) If (a l - a2) < 0, then thc second exponential in the solution tends to
zero as t increases, so tliat the solution approaches the balanced growtll patli.
ir Ori the otller lland, if (a l a 2 ) > 0, then the opposite conclusion liolds. Sirice
e (a l a 2 ) >< 0 means a l j a 2 , it follows that the balanced growth path is stable
(unstable) if tlie consumption good sector is more (less) capital intensiva than
the capital good sector. This concerns stability in the absolute sense. For relativa
stability it is enougll tliat 1/(a l a)b l < n, and this does not require (a l a 2 ) < 0.
19.5. The stability of the model depends on the roots of the matrix kA. If
the rnatrix A is stable on its owrl, by usirig Llie results on D-stability (seo Sect.
18.2.2.1) it will be possible to choose economically meaningful (i.e., positive) policy
parameters such that the matrix kA remains stable. More generally, it is possible
to apply the Fisher and Fuller theorem; this describes the conditions that a matrix
A must fulfil for a diagonal matrix k to exist capable of making the product matrix
kA stable (see Sect.18.2.2.1). The diagonal matrix k may further be chosen so that
the matrix kA has strictly negativo real eigenvalues. It follows that, if A is unstable
but satisfies the required conditions, it is possible to find a matrix k that not only
stabilizes the system but also eliminates possible oscillations.
In our case, we want k > 0, l ience the matrix A must satisfy conditions
(18.80).
19.6. The pairing fiscal policy-internal equilibrium and monetary policy-external
equilibrium gives risa to the following system of differential equations
G' = V l ( Yp Y),
r' = V 2 [m(Y) xo Eo K(T)) ,
Y' = V3 [A(Y, r) + xo - m(Y) + G - Y].
The first two equations are the formal counterpart of the adjustment rules: gov-
ernment expenditure increases (decreases) if income falls short of (exceeds) the
full ernployment level and the interest rate increases (decreases) if there is a deficit
Answers to Exercises 649
(surplus) in the balance of payinents. The third equation describes the usual pro-
cess of adjustment of national income in response to excess demand; the constants
v Z > 0 indicate the adjustment velocities. Expanding in Taylor's series at the point
of equilibrium, we have
x'=Ax,
where x is the column vector of the deviations ( G, r, Y) and
0 0 vl
A . 0 V 2 Kr V2my
v3 v3 Ar v3 (Ay my 1)
0 k i Kr klmy
Al e 0 0 k2
k3 k 3 Ar k3 (A y my 1)
is the matrix of the system of differential quations. If we expand the characteristic
equation, we have
650
ro-
Que can see immediately that, as the constant terna is negativo, orle of the stability
1ts
condltions is violated, aIld thus the assignIIlent in questiorl gives rise to a Inove.Inent
int
tliat diverges from equilibrium.
Chapter 20
20.1 The dynamic behaviour assumptions made in Sect. 20.7 plus the assumption
on the dynamics of money supply, give rise to the following system of differential
equations:
in
dY 1 [I(Y,R) S(Y,R) M(Y)+Xo],
=
dt
dR
= ^2 [ L ( Y, R ) L
s) ,
dt
dL s
= W3[XoM(Y)+K(R)J
dt
from which, by linearizing at the equilibrium point and denoting with an overbar
the deviations from equilibrium, we have
dY e dI as M -. alI os 1
dt = 1 aY c^Y a y +
R R
dR _ c2 c^L Y + c2 -L R
c2L5,
dt DY R
dL s 3M - K -
c3
dt = Y+c3l^R.
Stability depends on the roots of the cllaracteristic equation
C1 C1 OH a 1 o
[ Y as aY ] ^
ar. ar,
c2 ^Y
C3
ay
C 3 dK 0
_ + [-cI (aY as
ay
all
aY
C2 ar, 2
n
r3 Rj
C2C3 al y y
+{ c31< + CIC2 ay a a ) aR \&R aH) ay
+CI C2C3 as mc (al as
[ ( JY - ay ^aY) DI a H) a
=0.
Assurning, for simplicity, c i = 1, the stability conditions are
al aM < ar,
aY c^Y
as aY CM
a ar _ as a r,
aI _ as _ aAf aR (Y,-? 1^ dY
-
ay aY ] < DL
[ J7
DR
c)I _ as am
a a s _ a ll ( ale d,^ aY
W HY aY j < ^l K
u^
Answers to Exercises 651
and
(O
[SOR dY Y a y ) ( ^R an) aY) > 0'
Note that the first three inequalities are necessary and sufficient for each real root
of the characteristic equation to be negative and therefore exclude monotonic (but
not oscillatory) instability; the fourth inequality, on the other hand, is necessary
and sufficient, together with the previous ones, for the complex roots to have a
negative real part, and therefore to exclude oscillatory instability.
Before performing tlie comparativa static exercise we have to check that
(j ay ) ay \ OH
- a?) 0
J= aL L 0.
ay al? 1
aM
y aK 0
a aR
Y = (Xo, a 1 , G) ,
R = (Xo, a l , G) ,
Ls = (Xo,al,G),
al es aM a y al _ as aR = _
( 8Y 8Y aY) axo + ( OR aR) axo
ar, aY ar, aR al, s 0,
axo + dR axo axo
aM aY aK aR _ 1
aY exo 01? exo
The solution of this system gives the values for the unknowns aaY aR
xo axo axo which
turn out to be
f axo
OH [( .
/J,
o j-y
f
aLs 1a ar as aL al a
axo l aR (a y a y + [ aOR ( aR R ), } l J'
11:17 VV C1 a UU 1.1;Ltt t;1Je5
wllere J )S\1
aY 1 ^Y ^3Y ) cal?
^ t
-
Given the third stability condition, J is positive. Once this has been ascer-
tained, we fiiid that 3Y/OX 0 > 0, while the signs of DR/OX 0 , OL s /OXO remain
uncertain. Note, in fact, that ( jY ^k) < 0 is a necessary stability condition in
a closed econoiny, but need no \t llold in an
/
open econorny.
20.1.b) Differentiating the equilibrium relationships with respect to eq we
root
obtairl a system whose solution is
(but
sary aY - aM ax as /J
a 1 aal
- OR
C C OR aR)
ve
DR DM (Dl as /l J
aa Da l (DDY aY
DL. amaL 31 as aK _ DL aI _ as j
sia l cxl {D)Y (1{ DRC) al? ^3R ()Y )r }')}
Given J > 0, OM/Da l > 0, it can be checked that OY/a i < 0 while the signs of
OR/3X 0 , OL s /OX Q rernain uncertain.
20.1.c) To study the consequences of an exogenous variation of government
expenditure we differentiate with respect to C and obtain:
aY _ DKIj
ea l DR !
_
aR _ DM / T
Yl`I^
aL, _ ax DL DM
( al y
3a 2 DY 3J + al? a ^`l
` n .
T
Given J
G J ,
.
we cal iili^it^ui^liety see tl ^{ 3Y /3
Ullat, v1 / (-J a2
^-..1
:anCl n /.`a.c,^
v^c^v^ aro po iti^.r,
tJ ^,. ub
m
tlie siga of 0L s /3a 2 is undeflned.
E (p, a l , a 2) = D (p, a
1) S (p, a 2 ) .
Provided that
3D OS
ap ap$o,
there exists a function
p =p(al,a2)
of wllich e vant to know tlie partial derivatives, that give us the reaction of the
equilibrium price to a shift in the pararneters. Let us first examine the effect of
a shift in the twstes pararneter, i.e. a l . Totally differentiating the excess demand
function with respect to a l we have
OD Op OS Op 0D
cap Sa l d p 3c 1 Da i'
frorrl wllicll
^^ f
Answers to Exercises 653
Given that the siga of (c3D/Op OS/ p ) (see Chap. 13) is negative and that
D/a l is positive, we cari coriclude that a shift in tastes has the effect of increasing
the equilibrium price.
We can similarly calculate
p _ OS / aD OS
0() 2 Oaz cap p
Given that DDS/0a 2 > 0, we can concliide that a reduction in production costs has
the effect of reducing price.
To study the effects of sliifts in tlie two 1>arainet.ers iri tlie Ivla shallian case,
we start from the excess demand price equation
E' (q, a
l, a 2) _ Pd (q, a l) Ps (q, a2)
Provided tliat
pd ps
(9q Oq 0
of which we want to know the partial derivatives, that give us the reaction of the
equilibrium quantity to a shift in the parameters. Let us first examine the effect of
a shift in the tastes parameter, i.e. a l . Totally differentiating the excess demand
price furiction with respect to a l we have
3 O O 0
Pd q p5 q __ 31),1
Oq 3a l Oq aa l 3a1'
whence
Oq OPd /O Pd 0Ps
3a 1 0a 1 3q Oq
Knowing that p d /a l > 0 and that (ap d /aq 0p 5 /Oq) < 0 (see Chap. 13), we
can say that a shift in a l causes an increase in q.
We similarly calculate
O
q aps / Pd aps
0a 2 0a 2 q Oq
The effect of a shift in a 2 is an increase in q, given that 3p 5 /3a 2 is negative.
20.3. To determine the effects of a shift in S k and S h on the steady state values of
k*, h* and y*, we need to totally differentiate the solution functions (Eqs. 19.27)
with respect
p tos^ and s,l,account being
g taken that ds
h > 0 dsk+ds jz = 0, namely
d S /1 C,S, d s 1, <
Thus h* will iiicrease, according to economic intuition; it is easy to see that dk*/dsh
lvill also lee positive if tlie inicial valles of s !L , s k satisfy tlie inequality
Si, <
sk
Finally, dy*/ds i , > 0 if both dk*/ds j , and dli*/ds, L are positive, wliile its sigri
rernairis uncertaiii when h* increases but k* decreases. A shift in investment from
physical to human capital is prohably (but not certainly) favoiirable to growth.
he
of
ad Chapter 21
21.1. The Jacobians of tlie two diferential equation systems are
3yi + y1 1 + 2y1Y2
2y
-1 + 1 J2 y1 +3/
and
le -3y1 - J2 1 -
2
J1 J2
2y
-1 - 1Y2 -Ji -3y2
At the equilibrium point Y 1 = Y 2 = 0, both become
0 1
-1 0
1 , 2 = +i.
y
Yii/ + Y 2 Y2 = ( + J2)2
Answers to Exercises 655
dr
dt - 2r 2 = 0
or
-r-2dr + 2dt = 0.
This is a differential equation with separated variables (see Chap. 24, Eqs. (24.8)
and (24.9)) whose solution is
- r -2 dr + 2f dt = r -1 + 2t = A,
1 c
r _ A - 2t 1 - 2ct'
Ji y + J2J2 = -
( J1 + J2)2
from which
dr+ 2r 2 =0,
dt
that can be solved as abo y e, yielding
c
r=
1+2ct
This fraction tends to zero as t -> oo, hence r(t) -+ 0, and the equilibrium point
is asymptotically stable.
21.3. We first observe that this non-linear system is autonomous (the function
h(y) does not contain t explicitly), hence its linear part is a uniformly good ap-
proximation to the wliole system. The characteristic egtiatiori of the matrix A
is
hence the roots are -1, -2, -6. The system is locally asyrnptotically stable.
UJJ 0
r t 1
rp sin2 Oo 1
( ; 'o, o, to) = t 1
9( o, o) sin 0p + (1 t sin 2 Oo )
2 + 1 + t`'
It is inirnediately evident that this expressioll tends to zero as t --> oo. Consider
now a t = t l = sin --2 0o arad calculate the value of the expression. We have
8)
r t ro r sin2 Op 1
( l; r o, 0 0, to) t 4 e 2
g(o, o) Lsin 0p + (1 t l sin 0 0 ) 1 +t
ro sine 0o 1
g( o, o) t o
sin Op + (1 sin -2
Oo sin' 0)2 + 1 + sin ` 0 0
_ ro 1
t 2 -
g(0o, o) sin 0p + 1 + sin - Oo
Now, for 0 0 -- * 7r, we know that sin 0 0 --> 0, and hence sin e 0 0 > 0, sin -4 Op > oo.
h It follows that r(t l ; ro, Op, to) --> oo, which means that the condition of Definition
21.1 is not satisfied and the system is not stable. This is due to the fact that the
t function T(1, yo, to) is not coritinuous alorig the ray Oo = 7r.
Chapter 22
22.1. Before solving the exercise we define the variables. Following Ramsey, we
denote by x (t) and a (t) the total rate of consumption and labour in the economy,
and by c (t) its capital. The production function is y = f (a, c) . Knowing that
U (x) and V (a) are respectively the total rate of utility of consurnption and the
total rate of disutility of labour, U (x) V (a) denotes the net enjoyment to be
maximized. So, the optimal control problem is
[
B (x) V (a)]
max W U de,
! = f de
hence
In this way we have simplified the problem by incorporating the constraint in the
objective functional, and the T-Iamiltonian coincides with the intermdiate func-
tion of the objective functional. Hence considering x as the control variable the
maximum principie gives
OH I u (x) [ f (a, c) x] + [B U (x) + V (a)] 0^
u
de
(x)
dt
13 [ U ( x ) V(a)1,
that is the Ramsey ralle.
In the original Rarnsey rnodel the discounting factor is not present, so that
current and future consumption have the same weight. Differently, in the model
with discounting, consumption increases, remains constant or decreases, depending
on whether the marginal product of capital exceeds, is equal to or is less than the
rate of preference. This is very important because the higher is the marginal
product of capital relative to tlie rate of time preference, the more it pays to
depress the current level of consumption in order to enjoy higher consumption
later.
22.2.a) The optimal growth problem with final terminal time is the same as Eqs.
(22.18), except that the upper limit of the welfare integral is t I , a given finite
parameter, and tliere is an additional condition that the terminal stock of capital
per worker, i.e. k (t 1 ) , must be no less than some given level, k 1 . The problem
is solved as in the infinite horizon case and the canonical equations (22.25) are
still applicable. In this case a terminal condition on the costate variable has to be
added. Applying the transversality condition (22.6), where x(t i ) = k(t 1 ) - k 1 , we
get the relation stated in the exercise.
y = f (r, 1).
Considering the deviation from the equilibrium point, we can write
y
=f( r 1
, ) f( r
e, 1 ) = [f ( r o, 1 ) f ( r e, l ) ] eOt,
where 3 is the stable root of Eq. (22.29).
Dividing it by f (r, 1) we have
f ( r, 1 ) - f ( r ea 1 ) ,.., ln f (r, 1)
= ln
f
(
ro,
1)1c-Qc
f( r 1
e, ) f( r
e, 1
) Lf(e,1)j
Assuming the difference f (r, 1) - f (r,;1) very small, the approximation is quite
good. So, we have
-(P- n
) t
+ 1 -
(A - 5
- p) t < 0,
p+5> 1 -0 (A-6-p)+6+n.
0
To show that this i nequality enseres cp > 0, we can rewrite it as follows
p-^ 1 -0
0
^> > 1-
0
(A- 6 ) +^z.
P>1-(A- 6 ) +n,
0
that iinplies
Hence
co > 0.
22.5. From the definition of u, p we get u' = m' - p', c' = e' - p'. Substituting
m', p', e' from the dynamic equations of the model, we liave
c
' =r-r^-0(y-f)-'ir.
Answers to Exercises 659 66
We now substitute y = y + 6c + bRo from the IS into the LM schedule to obtain hei
r in terms of u, c and the exogenous variables
U B zI .
mar
22.6. Before the oil discovery, the equilibrium point is at c e = 0, u e = k7 ,\(7r +r*). men
After the oil discovery, the new equilibrium point is as found in the previous
exercise, namely c e = )R,, /, ue = k(y + R) .\( 7r + r*) . The pre-oil equilibrium It is
point can be taken as the initial point of the post-oil system. Given the saddle- thes,
point stability of the model, the real exchange rate will jump on the stable arm of equa
the saddle and move along it toward the new equilibrium point. cona
To determine the nature of this jump we have to determine the slope of the
stable arm. This can be done along the lines treated in Chap. 21, Sect. 21.3.2.4,
Eq. (21.17) . Letting l < 0 be the stable root, the slope of the stable arm referred
to the new equilibrium point will be whic
c _ l
u 06 > 0.
The real exchange rate, to go from the initial value of zero to the new (negative) 7
equilibrium value will have to drop to a value on the stable arm which is below obta:
the new equilibrium value, after which it will start increasing towards equilibrium.
This is a case of undershooting.
kjtju
UUU 4`c i':; l::) 1. L eI e
= by
Chapter 23
23.1. From tlie suggested Liaptiiiov fiiriction V = -L (z 1 - 1) 2 + 2c (z2 - 1) 2 we
llave
1 z -r z +
dV= ( l ) i 1(z2
- 1)z2
dt
amn , hencesee Chap. 19, Eqs. (19.30)-
.
dV az0 ap
` = (zl
1
) (zl 2 zl
) + (
2
2
1 ) ( z l z 2 2 2)
i I dt
It is now easy to show that this expression is always negative (except of course at
lnd tlic cquilil)rium point z i = z = 1, where it is zero) by uising Fig. 19.1. Consider,
for example, points like A ira that figure. Iri Sect. 19.2 we have shown that
1
at poiiit A, where \ve also have z l < 1, z 2 < 1 or z 1 - 1 < 0, z 2 - 1 < 0. It follows
that
(z1
1
) ( z z1
6
z
1)
<0
, ( 2
z -1
) ( z z z
2) <0,
23.2 Civen the mathematical structure of the model, in which both y lt and Y2t
y
depend en y it-1, it is suflicient to examine the difference equation lt = h(ylt-1)
In fact, if y l (the leader's oiitput) tends to a stationary point, so will Y2 (the
th follower's output). This is consistent with the economics of the leader-follower
model.
m od
Let us first compute the derivative of thc function h with respect to its argu-
^), ment, i.e.
us ii ti' - dy1t/dy1t-1
It is easier to compute dy lt /dy lt _ 1 as d 1t/ dy lt-1 = (d y 2t/ dy lt-1) ( dy lt/ dy 2t), since
m y
e these two derivatives can be obtained by applying the implicit function rule to the
-
of equations defining the first-order conditions. Totally differentiating the first-order
condition for the follower we get
me 1
2
dy2t + df 1 + dy
2t d C2 d
y2 t
dy2t 0
) d2f 1 + 22t + d f dY y
4 dY2 dylt-1 dY dylt-1 dyzt-i dy d 1t-1 '
which gives
2
dy 2t _ y2t + ^^
dy
1t-1 ( y2t d +) + ( - -2
Totaily differentiating the first-order conditions for the leader we similarly
:4 ohtain
d^^2t
_
c2
yit Y ^- Y
_ d2r
(^1t^i + dV) (dY , )
Ariswers to Exercises 661
everywhere.
We can now pass to the study of global stability. Using the mean value theorem
( Okuguchi, 1976) we have
1?,'
I T1t ylt-1) = j I j Ylt-1 Ylt-21 ,
and y
where h' is computed at an intermediate value between yit-1 lt-2 Since we
have just shown that Jh' < 1, it follows from the Wu and Bro w n theoremsee Eq.
(23.9), where we are using the absolute value normthat global stability obtains.
Chapter 24
1
24.1. With p= 2 , the given production function becomes
Y = a 2 K+L+2a KL .
be rewritten as
n
T' =s a 2 - r+2af+1 .
s
24.1.a) If we define
n 2
A
s
n 2
B
s
r' =s (A\+1)(B^/r+1).
24.1.b) To solve it, we simply change the notation of the time derivative from
r' to dr/dt and rewrite the equation as-
1
+1)dr-dt.=0,
s(A +1)(B
mand 1 which has separated variables. ` o sim plif tlie integration, let us muloP 1
1^ 1 Y g Y through
g
0 (tlie '
by (ns) , obtaining
c ut ) _
rative
s (A ^ + 1) (B 3r + 1) d7 . - (ns)
2
dt = 0,
^olute
wliose solutiori is
_f( ns ) 2 dt+
f (A +l)(B +l)
dr=C
j s
oreen
The second integral can also be explicitly solved, so that we obtain
- (ns) 2 t + 1
ln (A^ + 1) - 1
ln ( B^ + 1) =,
A B
re we from which
Eq . (A^ + 1) (B ..^ 1^ - Fj = D exp [(ns) 2
t] ,
^ins.
-C
where D e is the new arbitrary constant. Assuming r = ro known for t = 0,
from this last equation we have
and so,
1 1 2
A^+ B^+ B
= exp ( (ns) tl
A \/ 1
1) B r +1 L
mes
which is the final form of the solution. From this equation it is not possible to make
r explicit as a function of t, so that the solution is not very useful_ for ohtain.i.n.g in.
E
a simple way a clear picture of the behaviour of r over time. This, however, can
can be obtained easily by means of the phase diagram (see the next exercse).
24.2. We first determine the egtiilibrium point(s) by setting r' = 0, which gives
s (a \/ + 1) 2 - nr = 0. Rearranging terms we have
n
a 2 -- r+2a^+1=0.
s
Let x _ +\/> 0 (remember that the square root has to be taken with the positive
sign) and solee the quadratic
1 I (a2_)x2+2ax+l=O.
re = [('Z ) 2
-a
5
Answers to Exercises 663
n
f(r)=s a 2 -- r+2af-}-1 .
s
We have
f(0) = s > 0,
df
s [( a 2 n/s) + a/ Vn ,
dr
d2f
_ as/2v.
dr2
The second derivative is always negative for r > 0. As regards the first derivative,
we can distinguish two cases:
2
(a) it is always positive for r > 0 if a n/s > 0. This means that the function
is monotonically increasing from s, hence it never intersects the horizontal axis
(raxis) at positive values of r. This confirms what we already know, that in this
case there is no real positive root.
2
(b) In the econornically rneaningful case a n/s < 0, the first derivative is
2 2
positive, zero, negative, respectively, for r 2
a / (n/s a ) , r > 0. The critical
2
2
value a / (n/s a^2
is smaller than r e 1, so that we have a diagram in which the
function f (r) starts at the ordinate s for r = 0, increases up to its maximum for
r = a 2 / (n/s a 2 ) 2 , then starts decre as ing and cuts the raxis from abo y e at
r = r e , after which it decreases toward oo.
Fr om this phase diagram it can easily be seen that in case (a) r increases
beyond all bounds, whereas in case (b) it tends monotonically from both cides to
the equilibrium point r , which is therefore monotonically stable.
Let us note that this exercise, together with exercise 1, provide an example {
of those cases in which, although the differential equation is explicitly integrable,
qualitative methods yield better results. This is due to the fact that the explicit
function r = r (t) cannot be obtained from the integral , as ave remarked aboye
(see also Chap. 21, Sect. 21.3.1 remark (2)).
24.3. From the model we obtain the equations
1 a
Y= K' +
1 b 1 b ,
K* = k K, ka
1 b 1 b'
1
Consider the inequalit;y
a2 1
C
2 l2'
(a2
m/s) [(n/s) 2 - aJ
which gives a
l
[(n/s)
2
a
2
< a 2 n/s 2 . This can be writt.en as a
2
( n/s)
z
a2
Y= 1
K + a
1 b 1 b ,
l
k ka
K=1 b K 1 _}_ 1 b
The actual capital stock increases and passes through K e , but cannot stop there
because we still nave K' = K 1 > 0 (point E). Thus the capital stock goes on
increasing up to the point in which it reaches its desired level (corresponding
to point B), where K' falls sharply to zero given the assumptions. But then
income falls to a/(1 b) and the desired capital stock to ka/(1 b), which is
now smaller than the existing capital stock. Therefore K' falls to K 2 (point C);
consequently, income falls to (a + K2 ) /(1 b) and the desired capital stock to
k (a + K2 ) / (1 b).
The desired capital stock goes on decreasing: it passes through Ke , but it
cannot remain there since K' is still negative ( K' = K2 < 0 : point F), so that
it decreases more and more cantil it reaches its desired value k (a + 1(2) / (1 b),
corresponding to point D. Here K' falls to zero, Y increases to a/ (1 b) and the
desired capital stock is ka/ (1 b), so that K' rises immediately to K 1 , and we are
again at point A.
Tllis s hows that in the phase diagram we have a continuous movement in the
segments AB and CD, with a discontinuous jump from B to C and from D to A.
24.4
24.4.a) The equilibriurn point is obtained letting p' = = 0, whence
i E 2
E_1 ^1 bq
^l(E-1
bq ) =0.
1+E 1
1 1 -3q2
\ 2
E+E = 0,
Ans we rs to Exercises 665
with roots
Ei E E -4y
= 2
These are complex conjugate with positive real part, hence the origin is an unstable
focus (see Table 21.1) .
24.4.b) Applying remark (ii) of the Poincar-Bendixson positive criterion, it
only remains to find C 2 . Setting it as siiggested in tlie hint, we then consider the
Liapunov function V = + q 2 ), from which
dV d
_ p p + q^lq = p 1 (1 p2
[( - + E)p q] + q [p ( 8q3 q 2
)] = q bq4
dt dt dt E)
1q 1
4b
2
2b
2
+ ( 1 E) p 2
.
We begin by observing that both terms in square brackets are non-negative, hence
2
if either of therri is greater than 1/46, thcn dV/dt < O. Now, b ( q 2 e^ > 1/4 6
if q 2 > 1/b, and (1 E)p 2 > 145 if p 2 > 1/46(1 E).
It follows that a sufficient condition for dV/dt < 0 is p 2 + q 2 > 1/46(1 E) +1/b
or
+q 2
pl > 46(1 E) Pc
Hence, if we take C2 as the circle given in the exercise, it follows from the Liapunov
function that all trajectories coming from outside C 2 pass through the points of
C2 for t increasing and penetrate iii the circular region D included between C2
and the (arbitrarily small) circle C l drawn around the origin. Since there is no
singular point in D (the only singular point is the origin), at least one limit cycle
exists in D.
24.4.c) Application of Nemitzky's theorem gives rise to the expression
3
N
= p[( 1 + E)p - q] + q [.P
( 9 q)].
It is immediately obvious that this expression coincides with dV/dt examined in
4.b). Hence N < 0 for p 2 + q 2 = P 2 > Pc Similarly it can be shown that N > 0
1
for p 2 +q 2 = P1 where Pi is taken sufficiently
Y smaller than P^^ for example
P 2P c
2
It is sufficient to observe that if we take q < 1/2b and p 2 < 1/86(1 E), then
b(q2 zb 2
< (1/2)(1/46) and (1 E)p 2 < (1/2)(1/46), hence
)
2
b(q2 2b 1
+ ( 1 E ) p2 < 4b
1
which means that N is positive. Thus a sufficient condition for N> 0 is p 2 + q 2 <
(1/2)[1/46(1 E) + 1/6] or
1 5 - 46 1
p2
+g2< 246(1E)
-2PC
Since there is no singular point in the circular ring D between the two circles
of radii pi and P 2 , there is a stable limit cycle in D.
24.5. The Jacobian of the Lotka-Volterra systeni is
al b1u b1v
b2u (a 2 b2v)
0
bla2/b2
b 2 a 1 /b l 0
.\ 2 + ala2 = 0.
The roots are pare imaginary, hence the singular point is locally a centre.
Chaptear 25
t,
25.1. Let v[D(r, a) S(r)] = f (r, a), and assume that the demand and supply
curves become tangent for the value ao of the parameter (with a corresponding
value re of the equilibrium exchange rate), hence OD(re , o, ao)/dr = aS(re,o) /r.
Then we compute:
(1.a) f (re , 0, ao) /CUT = v{D(re , o, CYo) /r DS(re , o)/CUr] = 0,
(1.b) 3 2 f ( r e,0, a o)/ 3r2 = v { 3 2D ( r e,0, a o)/ 0r2 02S(re,0)/0r2]
= VD 2 S(1'e 0)/ 3r2 0,
(1.C) Df (re,0, ao)/a > 0,
where we have used the data of tlie problem. It follows that the conditions of
definition 1 in Sect. 25.2.1 are satisfied.
,\ 2 +a l + a2 =0,
wliere
a l - [ a ( IY S Y) + IK } ,
a2= a[(IYSY)(IK-6)IY(IKSK)].
A necessary condition for complex roots is a 2 > 0. which gives
Since (Iy Sy) > 0, a 2 > 0 requires b < 1. This is the critical value of b (note
that it is not obvious that S < 1, as this partly depends on the time unit chosen).
The rest of the analysis proceeds as in the text.
Answers to Exercises 667
25.3. If we consider the characteristic equation (25.40), the condition for the roots
to be complex is
ay
C9J1 - ( .(9Y2 - a - l DY2
0 = a23 1 - Q + d Y2 - 2 - 4
<0
Ok, (9k2 kl (Ok2 ak2 akl
am ay2
Q
2
+ 4 ay' 3712 < o
ak l ak2 ak2 akl
whence
4 3 Ji 3Y2 < _ay 1 3L12
ak 2 ak l Ok1 3k2
25.4. Performing a linear approximation at the equilibrium point (where I =
S, L = L 5 ) we obtain the Jacobian
,QLy LR
which means that the IS schedule is not only positively sloped (due to the Kaldor
assumption) but has a slope greater than the slope of the LM schedule.
As regards a l , we have two degrees of freedom (the parameters a,,3). Let us
for example take -y = a/,8 as a parameter. Then
LR
= /3['y(Iy Sy) + L R ] = 0 according as 7 < 'yo, 'Yo = R > 0
< > 1ySy
Since a l = 0 for the critical value 'yo of the parameter, the Jacobian has a pair
of pure imaginary roots there, \ 1,2 = i a 2 . The roots will remain complex con-
.
jugate for a l sufficiently small, namely for -y sufficiently near to ' yo This satisfies
assumption (H.1) of the Hopf bifurcation theorem.
V.JV
d (_ai) 1
d7 = 2^ (Iy Sy)
> 0'
7=70
x 2 x
7 'Y x 1 2 + 7q5
x2 2x l x2 1
A2 + al,\ + a2 = 0,
\vhere
a
l = 1-
a 2 7(1 -
Given the definition of 0 = 13/b = )8 /(s + (6) < 1, it follows that a 2 > 0, hence the
necessary condition for the roots to be complex is satisfied. As regards a l , it is
easy to see that
al 0 according as -y ^ 'yo, -yo - 1.
1 1
A 1,2 = 2 a l 2 al 4a2.
Since a l = 0 for the critical value -yo of the parameter, the Jacobian has a pair
of pire iinaginary roots there, )\ 1,2 = i a 2 . The roots will rernain complex con-
jugate for a l sufficiently small, namely for -y sufficiently near to 1. This satisfies
assumption (H.1) of the Hopf bifurcation theorem.
It is also easy to check that
d (-2a l ) 7
0,
d7 2 >
-Y=-Yo
Chapter 26
26.1. When the furiction f (k) is well beliaved, the function s f (k) will have the
same form as the function s f (.) in Fig. 1 3 .2 (with k in the place of r). Hence:
1.a) the intersection between this function and the straight line (1 +n)k, which
determines the equilibrium point k t + 1 = k t = 1-e, exists and is unique in the positive
quadrant.
1.b) Let us take a linear approximation at the equilibrium point, which gives
the first-order linear difference equation
s cp
k
ti +l =
1 +n k-t
where cp = (d f /dk) evaluated at the equilibrium point, and the overbar denotes
the deviations from equilibrium. F om the diagram recalled in 1.a it is easy to see
that, at the equilibrium point, the slope of the function s f (k) is smaller than the
sloP e o the
^ of
r
th straightht i
bune, hence s ^P < 1 + n i.e. [ s ^Pl (1 + n] < 1. It follows that
the solution to the linear difference equation (Chap. 3, Sect. 3.1) is monotonically
stable.
1.c) Consider now the original nonlinear difference equation
k f (lit).
t+1 = 1 _ n
The function [s/(1 + n)] f (k t ) will have the same aspect as the function s f (kt).
By using a phase diagram (see, for example, Fig. 21.14) it is easy to check global
stability.
26.2. The linear approximation to the Lorenz model (26.24) gives rise to the
following linear system
- I _ -U^ 0'J,
71 _ (Pze)!YXez,
z
^ =Yei+Xe2J^,
.
where x e , 1J e , z e are evaluated at the equilibrium point. At the origen, x e = y e =
z e = 0, which much simplifies the linear system, whose Jacobian becomes
a a 0
p 1 0
0 0 /3
/\2+(1+a),\+a(1p) =0.
Wlieri p <. i, the successioii oi the sigris of the coeflicients in this quadratic equation
is + + +, hence (see Sect. 14.1.4) both its roots will be of stable type. This shows
that the origin is locally stable when p < 1.
26.3. By using the results of the previous exercise, we see that the succession
of the signs of the coefficients in the qadratic equation is + + when p >
1, hence this equation will have one positive and one negative real root. Thus
the characteristic equation has one positive and two negative real roots, hence
the origin is a saddle point with a one-dimensional unstable manifold and a two-
dimensional stable manifold (see Chap. 18, Sect. 18.2.2.3) .
26.4. Let us first determine the two non-trivial equilibrium points., which arise
when p> 1. If we let x' _ ' = z' = 0 in system (26.24) we have
0= a(x + y),
0= pxy--zx,
0= 3z + xy.
2
From the first equation we liave x = y and consequently z = (1/ 3 )x from the
third. Substitution into the second gives
(l//.3)x3+(1p)x=0.
A root is x = 0 (the trivial solution) while the remaining two are given by
x=+ ti (p-1),
which are real and different from the trivial solution if, and only if, p > 1. Thus
wehave
1
z
(xc, ye, e) _ ( ( p 1 ), + Q(P ), p 1).
At these points the characteristic equation of the Jacobian of the system is
a\ a 0
p z e 1A x e =0,
Ye Xe ^A
whence
\3 + ( a -+- 1)A2 ( a +p),\+2a3(p 1) = 0.
Since ave are considering p > 1, all coefficients are positive, hence there can be
no non-negative real root. The crucial stability condition is (16.27) in Chap. 16,
Scct. 16.4, wliich gives
a
(rr + f3 + 1 )fl( + p) 2af3(p 1) > 0,
from which
( a - f3 1)pa(cr+0+3) > 0.
Answers to Exercises 671
a(a+,Q+3)
26.5. We know (see Sect. 16.4, p. 221) that, wlien the last inequality in the
previous exercise turns into an equality, namely when
a(a +,C3+3)
then there is certainly a pair of complex roots with zero real part. Tlie tliird root
must be real. Now, given a real root, say \ 1 , and two pure imaginary roots, say
I\ 2,3 = iw, and using the fact that the sum of the roots equals minus the coefficient
of )2 in the cliaracteristic equation, we immediately obtain )\ _ (a + j3 + 1).
Since the product of the roots equals minus the constant term, we get
2
+ 1)w = 2a,Q(p 1),
a++
from which, substitliting the value of p and solving for w, we obtain (after simple
manipulations)
As regards the Hopf bifurcation, condition (Hl) of the theorem given in Sect.
25.2.2 is satisfied. To check condition (H2), we have to compute the derivative of
the real part of the pair of complex roots (evaluated at the point where this real
part becomes zero, namely (a0/p)o). For this we can use sensitivity analysis,
in particular Eqs. (18.101) in Sect. 18.2.2.2. The parameter is p, hence we have
- C -2ae -0,
ap ap
+2 20 a +1 3w =
2(a + /3 + l)a0
1 a aw
^^ 1 ^l +2(a+/3+ 1) ^ i = 2af .
aP aP
739\ /3 i3
22 (a, -1) >
aP _ 0 3 1)
4(u +0+1) +8a (a + l)/(a/
i
Condition (H2) is satisfied, hence we have a Hopf bifurcation.
26.6. When p > a(o, + /3 + 3)/(a 3 - 1), the crucial stability condition (see
exercise 26.4) is not satisfied, which means that the characteristic equation will
have one negative real root and a pair of complex conjugate roots with positive
real part. The same result can be reached by observing that, since (a0/OP)e= 0 > 0
(see the previous exercise), when p increases beyond the critical point, the real
part of the cornplex roots becomes positive.
26.7. Followiiig tlie liint, to maximize the given utility function in any period
t we forrn the Lagrangian (the time subscript has been cdropped for notational
simplicity)
L = XX 2a + (1 P1 X% 1 P2X2),
c l (t) - w l = (1 +'y )[ w o - c o( t ) ] , 1
It follows that
^,a{1-[(co-ivo)/al} = ( + 1) - ?vO
co t
co(t) - w O '
hence a l c w a
co(t + 1) w 0 e { [{ o o)/ J}[ C (t) _ wo].
26.8.b) Following the hint, we get
where y t - co(t) - vio and cx - e a . rt'his map is knowvn to give risc to chaos for
a> 14.675 (or a> 2.692), see Eq. (26.10).
Chapter 27
27.1. At the steady-state equilibrium point we have r'(t) = 0 and, of course,
r(t - 1) = r(t). Thus this point is the same as in the standard neoclassical growtli
model. Linearizing the dynamic equation at the equilibrium point we have
where a superscript bar denotes t.iie deviatioiis froin equilibrium, and (d f /dr) is
evaluated at the equilibrium point. The characteristic equation is
A + n - s(d f /dr)e -A = 0,
namely
A = -n + s(d f /dr)e->',
which is in the form (27.19) . We can then apply Burger's theorem: observing
that s(d f /dr) > -1, the necessary and sufficient condition (1) requires -n <
-s(d f /dr), that is
n>
S(d f /dr).
Inspection of Fig. 13.2 shows that, at the equilibrium point, s(d f /dr) < n; hence
the stability condition is satisfied.
VVe know frorn the general treatment of rnixed equations of retarded type that
their characteristic equation has at most %wo real roots, while all the others are
complex. Hence the approach to equilibrium will be oscillatory.
IiI}sv Cls LO .LXC.l u ses
27.2.a) To avoid the possibility of prices converging to zero there must be one
zero real root, for in this case the solution will be of the type
00
t c
p( ) = I + ere"rt,
---.2
where c 1 is the arbitrary constant corresponding to the zero real root. Letting
,\ = 0 in the characteristic equation we obtain /31-t = 1 as the condition for this to
occur.
27.2.b) Let us tlien consider the equation
A+g-ge-A=0.
Since g > (), aecording to general principies this equation will have only one real
root which is a simple root; hence there will be no real root other than zero.
27.2.c) The remaining roots will be complex. Letting ai,Q be the typical pair
of complex roots and applying the procedure described in Sect. 27.4, we obtain
a = g + ge -a COS /3,
3 = ge -a sin/3.
The second equation implies 3 > 0, namely 2kir < 3 < (2k + 1)7r, where k _
0, 1, 2, ... Tlierefore, there will be a major cycle for k = 0 and (an infinite number
of) minor cycles for k = 1, 2, ...
In any of these intervals we have 0 < sin /3 < 1. It is then easy to see that,
in these intervals, a must be negative. Let us consider, for example, the interval,
2-ir < < 37r, and let us take an arbitrary value of g, say g = 1. Although we
do not know 3 (this would require the numerical solution of the system, which
we are not interested in), we know that it is greater than one (between 27r and
37r) and that 0 < sin 0 < 1. The second equation then requires e -a > 1, which in
turn implies a < 0. This proves th^.t the oscillations will be damped. For a more
general proof see Padoan and Petit (1978, Sect. 3).
/3 _ a
0g 3g
Since we know that 3 > 0 and a < 0 (exercise 2.c), it follows that 0,3/ag < 0,
which implies that the frequency of the oscillations ( /3 /27r) decreases as g increases.
Similarly, from the implicit function /3 ge -a sin /3 = 0, we obtain
Da e -a sin/3 1 )
g ge-^ sin 0 g
r,
Is
ISP