Economics of Department, NCU Macroeconomics Theory

Dynamic Analysis
Blanchard, O., and Fischer, S., 1989. Lectures on Macroeconomics, Ch. 10. The MIT Press.
Samuelson, P., 1947. Foundations of Economic Analysis, Ch. 9. Harvard University Press.

The comparative statics and the dynamics complement each other:

a. The result of the comparative statics must be tested by the dynamics and then it will be
meaningful.
b. We can use the dynamic conditions to make definite the result of the comparative statics
if it is ambiguous.

Here we use a seminal correspondence principle (對應原理) in Samuelson (1947) to explore

this issue and in turn make an interesting application.

1. Correspondence principle
Example 1:

The demand function for goods: = ( ); <0

The supply function for goods: = ℎ( , ); ℎ 0

The market-clearing condition for goods: =

(1) The result of the comparative statics (the effect of the parameter α on the price )

( ) = ℎ( , )

Total differential: =ℎ +ℎ ⇒ = 0
?

(2) Dynamic analysis

The Walras’ price adjustment process:

= [ − ] = [ ( ) − ℎ( , )],

where the parameter is the adjustment speed of market (i.e., the response
degree of the price).

(3) Taylor linear approximation

( )= ( )+ ( − )+ ( − ) +⋯
2!

1
Economics of Department, NCU Macroeconomics Theory

1
( , )= ( , )+ ( − )+ ( − )+ ( − ) + ( − )
2!
Linear approximation:

( )≅ ( )+ ( − )

( , )≅ ( , )+ ( − )+ ( − )

Using Taylor linear approximation, we linearize the demand function and the supply
function around the new equilibrium to yield:

( )= ( )+ ( − )

ℎ( , ) = ( , )+ℎ ( − )+ℎ ( − )

Substituting these equations into the Walras’ price adjustment equation, we obtain:

= ( )+ ( − ) − ℎ( , )−ℎ ( − )−ℎ ( − )

We know that the new equilibrium is ( ) = ℎ( , ) and also know =

during the adjustment process. So, the Walras’ price adjustment equation can be
rewritten as follows.

= ( − ℎ )( − )

(4) First-order linear differential equation

We set the first-order differential equation in the variable , with two parameters
and :

= = −

Let and are the particular solution (特定解) and homogenous solution (齊
次解) of the differential equation respectively. The general solution (一般解) of the
variable is：

= +

First, we solve the particular solution, which satisfies the condition = 0.

= − =0→ =

The homogeneous solution is regardless of the coefficient (parameter ) and then

2
Economics of Department, NCU Macroeconomics Theory

satisfies the first-order differential equation of the variable . That is,

=−

Integrating the above equation yields:

= − → = → = = ,

where the parameter is the characteristic root ( 特 性 根 ). Combining the

particular solution and the homogeneous solution, we can obtain the general
solution:

= +

(5) Result

Using this method, we derive the general solution of the price:

= ( − ℎ )( − ),

( )
= + .

The variable converges to the level if − ℎ < 0. We also make use of this
condition to make definite the result of the comparative statics. That is,

= < 0.

Example 2:

We consider the IS-LM model with the acceleration principle (加速原理). To be

specific, investment is the function of the interest rate and income.

The market-clearing condition for goods and services:

= ( ) + ( , ) + ; 0 < < 1, < 0, >0 (1)

The market-clearing condition for money:

= ( , ); > 0, <0 (2)

a. The result of the comparative statics (the effect of monetary policy on the
income )
3
Economics of Department, NCU Macroeconomics Theory

Total differential:

IS: = + + + (3)

LM: = + (4)

As a result, the slopes of the IS and LM curves are respectively:

= 0, if + 1 (5)

=− > 0 (6)

Here we can’t derive the explicit solution of the interest rate and income
because equations (1) and (2) are the implicit functions. So, we only explore the
effect of policies in this model. To obtain this effect, we use the matrix algebra:

+ −1 −
=

Using the Cramer’s rule yields:

=− 0, if ( + − 1) − 0
( + − 1) −
b. Dynamic analysis

Samuelson (1947) develops an auto-adjustment mechanism of the endogenous

variables:

= [ ( )+ ( , )+ − ]; >0 (7)

= [ ( , )− ]; >0 (8)

Using Taylor linear approximation, we linearize equation (7) and equation (8)
around the new equilibrium ( , , and ) to yields :

= [ ( )+ ( − )+ ( , )+ ( − )+ ( − )

+( − )+ −( − )− ]

= [ ( , )+ ( − )+ ( − )−( − )− ]

We know that the market-clearing condition for goods and services and the

4
Economics of Department, NCU Macroeconomics Theory

market-clearing condition for money hold. That is ( ) + ( , )+ =

and ( , ) = . We make use of these conditions to obtain:

= [( + − 1)( − )+ ( − )] (9)

= [ ( − )+ ( − )] (10)

Based on equations (9) and (10), we obtain the dynamic system:

= + , (11)

= + , (12)

where is the particular solution of and is the particular solution of .

As for and , they are the solve-for parameters. Differentiating equations (11)
and (12) with respect to time, we derive:

= = ( − ) (13)

= = ( − ) (14)

Substituting (13) and (14) into (9) and (10) obtain:

( + − 1) − − 0
=
− − 0
Given this result, we can derive the characteristic equation:
( + − 1) −
=0
−

Solving:

−[ ( + − 1) + ] + [( + − 1) − ]=0 (15)

Let and are respectively two characteristic roots of the dynamic system.
Here we set < 0 and < 0 to satisfy the dynamic stability condition. The
trace and determinant are respectively:

+ = ( + − 1) + <0 (16)

= [( + − 1) − ] >0 (17)

We obtain the dynamic stability condition: [( + − 1) − ] > 0.

This condition implies that the slope is larger in the LM curve than in the IS curve.
5
Economics of Department, NCU Macroeconomics Theory

Application: an extension of the IS-LM model

Blanchard and Fischer (1989, p. 532-536) modifies the IS-LM model by introducing a
bond market.

The goods market adjustment equation:

= [ ( )+ ( )+ − ]. (1)

The market-clearing condition for money:

( , )= , (2)

where is the interest rate of the long-term bond and is the interest rate of the
short-term bond. Let and are the interest income of the long-term bond and the
price of the long-term bond respectively to derive = / . Moreover, we know that
− ( − / ) is the relative return of the long-term bond and the short-term bond. The
market-clearing condition for bond is

− + = (3)

We assume a perfect substitution between the long-term bond and the short-term bond.
Equation (3) can be rearranged as follows.

= − (4)

Based on equations (1), (2) and (4), we obtain a macroeconomic model with goods
market, money market and bond market and then make use of these equations to solve ,
and . First, taking a total differential of (1) to obtain:

= [ + + − ]

This equation can be written a reduce form:

= , , , (5)

where = [ − 1] < 0, = < 0 and = > 0. Equation (5) is the goods

market adjustment equation and the adjustment speed is slow. Moreover, substituting
equation (4) into equation (2) and differentiating, we derive

−
+ − =

6
Economics of Department, NCU Macroeconomics Theory

We can rearrange this equation by assuming = 0 in the initial economy to obtain the
reduce form:

= , , , (6)

where = / < 0, = > 0 and =− / > 0. Based on equations (5) and

(6), we obtain the dynamic system.

We use the Gandolfo (1980)’s method to solve equations (5) and (6). The Gandolfo
(1980)’s method:

= + + (g1)

= + + (g2)

Based on equations (g1) and (g2), we derive the general solutions of and :

= + , (g3)

= + , (g4)

where and are the particular solutions and and are the homogenous
solutions. First, we substitute = 0 and = 0 into equations (g1) and (g2) to solve the
particular solutions. We obtain:

−
= −

Using the Cramer’s rule yields:

= (g5)

= (g6)

Moreover, we are regardless of the exogenous parameters and and then equations
(g1) and (g2) can be expressed as follows.

= + (g7)

= + (g8)

Based on equations (g7) and (g8), we can derive the characteristic equation:
−
=0
−
7
Economics of Department, NCU Macroeconomics Theory

Let and are respectively two characteristic roots of the dynamic system. So, can
be expressed:

= + (g9)
Differentiating equation (g9) with respect to time yields:

= + (g10)

Substituting (g9) and (g10) into (g7) yields:

= + (g11)

Combining equations (g5), (g6), (g9) and g(11), we derive the general solutions of and
are respectively:

= + + (g12)

= + + (g13)

We make use of the Gandolfo (1980)’s method to solve Blanchard and Fischer (1989)’s the
particular solution and the homogeneous solution. First, we solve the particular solution of
model. In the steady-state, equations (5) and (6) satisfy:

, , = 0, (7)

, , = 0, (8)

where and are the steady-state value (i.e, the particular solution). Taking a total
differential of (7) and (8) and using the Cramer’s rule, we obtain:

− +
=
−

−
=
−

These equations can be rearranged as follows:

= , (9)

= , (10)

8
Economics of Department, NCU Macroeconomics Theory

Based on equations (5) and (6), we derive the characteristic equation:

−
=0
−

Let and are respectively two characteristic roots of the dynamic system. The trace
and determinant are respectively:

+ = + ≶0 (11)

= − <0 (12)

Equation (12) shows that there exist a positive root and a negative root. Here we set
< 0 < . Combining equations (9) and (10), the general solutions of and are
respectively:

= + + (13)

= + + (14)

Next, we will use the graph to present the dynamic path. First, using equations (7) and
(8) yields the slopes of the = 0 locus and the = 0 locus:

=− <0 (15)

=− >0 (16)

Based on equations (13) and (14), we know that the convergence condition (stable condition)
is = 0. Here we define the locus SS is the stable arm (or the saddle path), which is all
combinations of and that satisfy = 0. The locus is described by:

− = − (17)

Equation (17) shows that the slope of the locus SS is:

= ≶0 (18)

Using equations (13) and (14) obtains:

( )
− = + ⟺( − ) 1− = (19)

9
Economics of Department, NCU Macroeconomics Theory

So, we can derive that the locus SS has a positive slope:

= >0 (20)

In addition, we use equation (19) to derive:

( − ) 1− = ⇔ 1− =−

This equation implies that the slope is larger in the locus = 0 than in the locus SS.
Moreover, equations (13) and (14) show that the divergence condition (unstable condition)
is = 0. Here we define the locus UU is the unstable arm, which is all combinations of
and that satisfy = 0. The locus is described by:

− = − (21)

Based on equation (21), we know that the locus UU has a negative slope:

= <0 (22)

Comparing equations (15) and (22), we obtain that the locus = 0 is flatter than the locus
UU. As for the other paths, we can use equations (13) and (14) to derive the slope of the
path:

Given ⟶ ∞ and ⟶ −∞, the slopes of the path are respectively:

lim = >0
⟶

lim = <0
⟶

The results imply that the other paths are the asymptote, which trend to the locus UU from
the locus SS. We also find the signs of and to response other paths by considering
equation (13). That is, there are four combinations of and in other paths. First,
given ⟶ ∞, we can evaluate the sign of according to the final value of output:

lim = + lim = + ∞
⟶ ⟶

In other words, this equation shows ⟶ ∞ if > 0 whereas ⟶ −∞ if < 0.

10
Economics of Department, NCU Macroeconomics Theory

Moreover, given ⟶ −∞, we can evaluate the sign of according to the initial value of
output:

lim = + lim = + ∞
⟶ ⟶

This equation shows that ⟶ ∞ if > 0 whereas ⟶ −∞ if < 0.

11
Economics of Department, NCU Macroeconomics Theory

R
UU ( , )
= 0 ( )

( , )
UU ( , )

SS ( , )
= 0 ( )

= 0 ( )
Y

R
UU ( , )
= 0 ( )

( , )
UU ( , )

SS ( , )

= 0 ( )

= 0 ( )
Y

12