EC744 Lecture Notes: Economic Dynamics: Prof. Jianjun Miao
EC744 Lecture Notes: Economic Dynamics: Prof. Jianjun Miao
Economic Dynamics
Prof. Jianjun Miao
1 Deterministic Dynamic System
State vector a
t
R
a
State transition function a
t
= j
_
a
0
, t; c
_
, a
0
= a
0
, parameter c R
j
A parametrized dynamic system: (A, j) where A R
a
and j : AR
R
j
A
Dierence equation
1 (t; a
t
, a
t+1
, ..., a
t+n
; c) = 0
Or
a
t+n
= ) (t; a
t
, a
t+1
, ..., a
t+n1
; c)
The solution a
t
is an orbit or trajectory
Linear if ) is linear. Order n
Focus on rst-order system. Why?
a
t+1
= ) (t, a
t
; c)
Autonomous system if ) does not depend on t.
The impact of t typically comes from an exogenous forcing term:
1
t
= (1 c) 1
t1
+ 1
t
If initial value is known a
0
, it is easy to obtain a solution. Boundary value
problem
General solution, particular solution
Vector eld o
a
t
= a
t+1
a
t
= ) (t, a
t
; c) a
t
= o (t, a
t
; c)
Steady state of an autonomous system
a
t+1
= ) (a
t
)
if the xed point a such that
a = ) ( a)
(Liapunov Stability) The state a is a stable xed point of the map ) if for
any c 0, there exists some c (0, c) such that
|a
c
a| < c == |a
t
a| < c
for all t c.
(Asymptotic stability) The state a is asymptotically stable if it is stable and
the constant in the last denition can be chosen so that, if |a
c
a| < c
for any c, then |a
t
a| 0 as t o.
Basin of attraction, globally asymptotically stable (Figures)
Periodic orbits
2 Scalar Linear Equations
First-order dierence equation:
a
t
= oa
t1
+ b
t
.
Superposition principle:
a
j
t
= a
c
t
+ a
j
t
where the complementary function a
c
t
is the general solution to the correspond-
ing homogeneous equation a
t
= oa
t1
and a
j
t
is a particular solution.
2.1 Homogeneous equation
a
t
= oa
t1
Phase diagram
Four cases (table): monotonic convergence, damped oscillation, monotone
divergence, explosive oscillation
General solution
a
t
= co
t
Particular solution
a
t
= a
c
o
tc
2.2 Autonomous equations
a
t
= oa
t1
+ b
By the superposition principle,
a
j
t
= co
t
+ a
j
t
.
A natural candidate for the particular solution is the steady state
a =
b
1 o
for o ,= 1.
So
a
j
t
= co
t
+ a.
Boundary value problem
a
t
= a + (a
c
a) o
tc
The solution is asymptotically stable if [o[ < 1.
2.3 Non-autonomous equations
a
t
= oa
t1
+ b
t
.
By the superposition principle,
a
j
t
= co
t
+ a
j
t
.
Backward-Looking Solution. Repeated substitution
a
t
= o
a
a
ta
+
a1
i=0
o
i
b
ti
Let a = 0 if a
0
is given
Let a o if there is no initial condition
Assume [o[ < 1 and [b
t
[ < 1
a
j
t
= co
t
+
o
i=0
o
i
b
ti
Forward-Looking Solution. Repeated substitution
a
t
=
_
1
o
_
a
a
t+a
1
o
a1
)=0
_
1
o
_
)
b
t+1+)
.
Let a o and assume [b
t
[ < 1. We must have [o[ 1.
a
j
t
= co
t
1
o
o
)=0
_
1
o
_
)
b
t+1+)
Stability
3 Stock Market Bubbles
3.1 Price Dynamics with Adaptive Expectations
(1 + v) j
t
= o
t
+ j
c
t+1
Adaptive expectations
j
c
t+1
= Aj
t
+ (1 A) j
c
t
.
Eliminating j
c
t+1
, we have
j
t+1
= oj
t
+ b
t
where
1 = 1 + v, o =
1(1 A)
1 A
,
b
t1
=
o
t
(1 A) o
t1
1 A
Backward solution
j
t
= o
t
j
0
. .
Bubble
+ j
+
t
..
Fundamental
where
j
+
t
=
A
1 A
t1
i=0
o
i
o
ti
.
Suppose o
t
= o 0 for all t. Then
j
+
t
=
_
1 o
t
_
o
v
The limit is j
+
= ov. We have
j
t
j
+
= o
t
(j
0
j
+
)
What is wrong with adaptive expectations?
Eliminating j
t
, we have
j
c
t+1
= oj
c
t
+
Ao
1 A
So
j
c
t
j
+
= o
t
(j
c
0
j
+
) .
We now have
j
t
j
c
t
= o
t
(j
0
j
c
0
)
So mistake is systematic!
3.2 Price Dynamics with Perfect Foresight
Assume
j
c
t+1
= j
t+1
.
We have
j
t+1
= 1j
t
o
Solution is
j
t
= j
+
+ c1
t
= j
+
+ (j
0
j
+
) 1
t
.
Kill bubble: c = 0!
3.3 Dividend Taxes
j
t+1
= 1j
t
(1 t) o
General solution
j
t
(t) = j
+
(t) + c1
t
= j
+
(t) + (j
0
j
+
(t)) 1
t
where
j
+
(t) =
(1 t) o
v
.
Unanticipated tax increase from t
0
to t
1
. The price immediately drops
to j
+
(t
1
) and remains this value forever.
j
+
(t
1
) =
(1 t
1
) o
v
<
(1 t
0
) o
v
= j
+
(t
0
)
Anticipated tax increase at date T. For t T, the stock price is j
t
=
j
+
(t
1
) .
For t < T, the tax rate is t
0
and the solution is
j
t
(t
0
) = j
+
(t
0
) + c1
t
= j
+
(t
0
) + (j
0
j
+
(t
0
)) 1
t
.
At time T, we must have j
T
(t
0
) = j
+
(t
1
) . So
j
+
(t
1
) = j
+
(t
0
) + (j
0
j
+
(t
0
)) 1
T
.
Solving yields
j
0
= j
+
(t
0
)
1
1
t
(j
+
(t
0
) j
+
(t
1
)) .