A Macroeconomic Model: The Neoclassical Growth

Model

Advanced Macroeconomics

September 2024

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 1 / 48
Labor-Leisure and Consumption-Savings Decisions

There are two main decisions in Macroeconomics: Labor-Leisure

Choice and Savings-Consumption Decision.
We are going to construct a model including these two decisions in
order to study different Macroeconomic problems.
Besides, we will study investment in physical capital, this is very
important for Macro analysis.
We will analyze efficient allocations and compare them with the
competitive equilibrium.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 2 / 48
 Efficient Allocations

A Two-Period Model: Pareto Optima

A representative individual has one unit of time that can devote to

work l or to leisure 1 − l.
How much time he devotes to each activity depends on his utility
function u (c, 1 − l ). This function is C 2 with u10 , u20 > 0 and
u100 , u200 < 0.
She owns a stock of physical capital K which allows her to transfer
wealth between periods. (There are no bonds)
At period 0, the capital stock is K0 > 0.
With this capital and labor she can produce the consumption good
with the technology F (K , l ). This function is C 2 with FK0 , Fl0 > 0 and
FK00 , Fl00 < 0.
F (·, ·) also satisfies the INADA conditions and is CRS

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 3 / 48
 Efficient Allocations

Period 0

Production is Y0 = F (K0 , l0 ). The individual has to choose between

two margins
More consumption C0 and less leisure 1 − l0 (or less consumption and
more leisure).
More savings/investment today (and less C0 ), but higher capital K1
and production Y1 next period, which implies higher C1 .
The individual is subject to two restrictions
Resource constraint: Y0 = C0 + I0
Capital law of motion: K1 = I0 + (1 − δ)K0
We can simplify
K1 = Y0 − C0 + (1 − δ)K0

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 4 / 48
 Efficient Allocations

Period 1 and Utility Function

Analogously for period 1

K2 = Y1 − C1 + (1 − δ)K1

Which comes from the corresponding

Resource constraint (RC) Y1 = F (K1 , l1 ) = C1 + I1 .
Law of motion (LM), K2 = I1 + (1 − δ)K1 .
The aggregate utility function is

u (C0 , 1 − l0 ) + βu (C1 , 1 − l1 )

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 5 / 48
 Efficient Allocations

Social Planner’s Problem

The Planner’s problem (or efficient allocations) is

max u (C0 , 1 − l0 ) + βu (C1 , 1 − l1 )

C0 ,C1 ,l0 ,l1 ,K1

subject to

K1 = F (K0 , l0 ) − C0 + (1 − δ)K0
K2 = F (K1 , l1 ) − C1 + (1 − δ)K1

given K0 > 0 and the transversality condition K2 = 0

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 6 / 48
 Efficient Allocations

Transversality Condition

The transversality condition imposes K2 = 0.

Why? What’s its meaning?
It must be noticed that in the two-period model the TC implies
I1 = 0 and that, if it is to hold, we have to assume δ = 1. Why?
The TC is an optimality condition, if it is not satisfied the solution to
our problem is not optimal. Why?

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 7 / 48
 Efficient Allocations

Planner’s Problem: Solution

The solution to the Planner’s problem is given by a system of

equations that determines the value of the variables: K0 , K1 , K2 , C0 ,
C1 , l0 y l1 .
We have 7 unknowns but only 5 equations from the Planner’s
problem.
We thus need two other equations to have a determinate system of
equations. This two equations are
Initial condition K0 .
Transvesality condition K2 = 0.
The transversality condition is not only an optimality condition but it
is also necessary for the system to be determinate!!!

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 8 / 48
 Efficient Allocations

Exercise I

Solve the Planner’s problem.

Build the Lagrangean and compute the first order conditions for Ct ,
lt , Kt +1 , λt .
Show that one obtains the following five foc:
1 The consumption-leisure margin in period 0

u20 (C0 , 1 − l0 ) = Fl0 (K0 , l0 )u10 (C0 , 1 − l0 )

2 The consumption margin between the two periods

u10 (C0 , 1 − l0 ) = (1 + FK0 (K1 , l1 ) − δ) βu10 (C1 , 1 − l1 )

3 The consumption-leisure margin in period 1

u20 (C1 , 1 − l1 ) = Fl0 (K1 , l1 )u10 (C1 , 1 − l1 )

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 9 / 48
 Efficient Allocations

Exercise II

4 The RC and the LM for period 0

K1 = F (K0 , 1 − l0 ) − C0 + (1 − δ)K0

5 and for period 1

K2 = F (K1 , 1 − l1 ) − C1 + (1 − δ)K1

Together with the initial condition K0 and the TC K2 = 0 provide a

system of 7 equations for 7 unkonwns.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 10 / 48
 Efficient Allocations

Net Real Interest Rate

What’s (1 + FK0 (Kt , 1 − lt ) − δ)?

We usually denote it as 1 + Rt , the net real interest rate.
It gives the return of reducing one unit of consumption today and
increasing investment by one unit.
One forgone unit of consumption today, yields tomorrow 1 + the
marginal product of capital (FK0 (Kt , 1 − lt )) minus the depreciation of
capital (δ).
Thus Rt = FK0 (Kt , 1 − lt ) − δ.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 11 / 48
 Efficient Allocations

Efficient Allocations

The first two foc state the properties that efficient allocations must
satisfy:
u20
1 = Fl0 ⇒ MRS2,1 = MgPl
u10
u10 (t )
2 = 1 + FK0 (Kt +1 , 1 − lt +1 ) − δ ⇒ MRStC,t +1 = 1 + Rt +1
βu10 (t + 1)
Task: Explain the economic meaning of these two equations.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 12 / 48
 Efficient Allocations

Infinite Horizon Problem

The Planner’s problem can be generalized to ∞ periods

∞
max ∑ βt u (Ct , 1 − lt )
{Ct ,lt ,Kt +1 }t∞=0 t =0

subject to

Kt +1 = F (Kt , lt ) − Ct + (1 − δ)Kt , ∀ t = 0...∞

 
λt
given K0 > 0 and the TC: limt →∞ Kt + 1 = 0
λ0
We construct the Lagrangean and compute the foc in each period for
Ct , lt , Kt +1 and λt .

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 13 / 48
 Efficient Allocations

First Order Conditions

The Lagrangean,

∞
L= ∑ { βt u (Ct , 1 − lt ) + λt [F (Kt , lt ) − Ct + (1 − δ)Kt − Kt +1 ]}.
t =0

The first order conditions of the problem are,

(Ct ) → βt u10 (Ct , 1 − lt ) = λt
(lt ) → βt u20 (Ct , 1 − lt ) = λt Fl0 (Kt , lt )
(Kt +1 ) → λt = λt +1 [FK0 (Kt +1 , lt +1 ) + (1 − δ)]
(λt ) → F (Kt , lt ) = Ct + Kt +1 − (1 − δ)Kt .

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 14 / 48
 Efficient Allocations

Lagrangean

L = u (C0 , 1 − l0 ) + λ0 [F (K0 , l0 ) − C0 + (1 − δK )K0 − K1 ] +

βu (C1 , 1 − l1 ) + λ1 [F (K1 , l1 ) − C1 + (1 − δK )K1 − K2 ] +

....

βt −1 u (Ct −1 , 1 − lt −1 ) + λt −1 [F (Kt −1 , lt −1 ) − Ct −1 + (1 − δK )Kt −1 − Kt ] +

βt u (Ct , 1 − lt ) + λt [F (Kt , lt ) − Ct + (1 − δK )Kt − Kt +1 ] +
βt +1 u (Ct +1 , 1 − lt +1 ) + λt +1 [F (Kt +1 , lt +1 ) − Ct +1 + (1 − δK )Kt +1 − Kt +2 ] +

....

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 15 / 48
 Efficient Allocations

First Order Conditions

For every two periods t and t + 1, the foc provide the following three
equations:
1 Consumption-leisure margin (in period t)

u20 (Ct , 1 − lt ) = Fl0 (Kt , lt )u10 (Ct , 1 − lt )

2 Euler Equation (saving-consumption margin)

u10 (Ct , 1 − lt ) = (1 + FK0 (Kt +1 , lt +1 ) − δ) βu10 (Ct +1 , 1 − lt +1 )

3 The RC and the LM for period t

Kt +1 = F (Kt , lt ) − Ct + (1 − δ)Kt

Together with the initial condition K0 and the transversality condition

these equations provide a determinate system of equations.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 16 / 48
 Efficient Allocations

Transversality Condition I

In the infinite horizon problem the TC states that the Net Present
Value (NPV) of future K at “t + 1”, goes to zero as t grows.

λt λ t λ t −1 λ 2 λ 1
Kt + 1 = ... Kt + 1
λ0 λ t −1 λ t −2 λ 1 λ 0

But from the foc (Kt +1 ) we know that

λt
= 1 + FK0 (Kt +1 , lt +1 ) − δ = 1 + Rt
λ t +1

and therefore
λt 1 1 1
Kt + 1 = ... Kt +1 = NPV (Kt +1 )
λ0 1 + Rt 1 + Rt −1 1 + R1

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 17 / 48
 Efficient Allocations

Transversality Condition II

The TC imposes that K cannot grow “too fast”. The growth rate of
K has to be lower that the net interest rate Rt .
The TC precludes the existence of bubbles and is also known as the
Non-Ponzi game condition.
Any recent example?...and more to come...

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 18 / 48
 Competitive Equilibrium

Competitive Equilibrium: Households

Households own physical capital and firms (profits) and rent capital
and labor to firms at rate rt and wage wt .
They also buy the consumption good, that yields utility, and
investment. They decide the labor supply lt .
Thus, the household budget constraint (BC) for period t is

Ct + It = wt lt + rt Kt + πt

Since they own the capital stock, are also subject to the physical
capital law of motion (LM)

Kt +1 = It + (1 − δ)Kt

Substituting It in the equations above we obtain

Kt +1 = wt lt + rt Kt + πt − Ct + (1 − δ)Kt

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 19 / 48
 Competitive Equilibrium

Competitive Equilibrium: Firms

Firms produce and sell the only good of the economy and hire labor
and capital from households. They maximize profits given by equation

πt = F (Kt , lt ) − wt lt − rt Kt
Both households and firms are price-takers (wt and rt ).
Without loss of generality we normalize prices so that the price of the
consumption good is 1 Pt = 1.
Utility and production functions are strictly increasing and strictly
concave.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 20 / 48
 Competitive Equilibrium

Competitive Equilibrium I

For a given utility function u (Ct , 1 − lt ) and production technology

F (Kt , lt ), a Competitive Equilibrium is set of sequences for the quantities
{Ct , lt , It , Yt , Kt +1 }t∞=0 such that given the prices {wt , rt }t∞=0 :
a) Households maximize utility
∞
max ∑ βt u (Ct , 1 − lt )
{Ct ,lt ,It ,Kt +1 }t∞=0 t =0

subject to

= wt lts + rt Kts + πt
Ct + It
Kts+1 = It + (1 − δ)Kts , ∀ t = 0...∞
 
λt
given K0 > 0 and the TC limt →∞ Kt +1 = 0
λ0
Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 21 / 48
 Competitive Equilibrium

Competitive Equilibrium II

b) Firms maximize profits

πt = Yt − wt ltd − rt Ktd

c) Markets are in equilibrium

Yt = Ct + It

lts = ltd ; Kts = Ktd

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 22 / 48
 Competitive Equilibrium

FOC: Firms

Firms choose the amount of labor and capital following the two foc:
∂Yt
1 = Fl0 (Kt , lt ) = wt ⇔ PMgl = wt
∂lt
∂Yt
2 = FK0 (Kt , lt ) = rt ⇔ PMgK = rt
∂Kt
For every period t.
Recall that if the production function displays Constant Returns to
Scale (CRS)... profits are zero.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 23 / 48
 Competitive Equilibrium

FOC: Households I

∞
L= ∑ { βt u (Ct , 1 − lt ) + λt [wt lt + rt Kt + πt − Ct + (1 − δK )Kt − Kt +1 ]}.
t =0

which implies the following foc for period t

( Ct ) → βt u10 (Ct , 1 − lt ) = λt
(lt ) → βt u20 (Ct , 1 − lt ) = λt wt
( Kt + 1 ) → λt = λt +1 [rt +1 + (1 − δ)]
( λt ) → Kt +1 = wt lt + rt Kt + πt − Ct + (1 − δ)Kt , (BC)+(LM).
The household and firm foc yield the two margins and the resource
constraint.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 24 / 48
 Competitive Equilibrium

FOC: Households II

1 Consumption-leisure margin (period t)

u20 (Ct , 1 − lt ) = F20 (Kt , lt )u10 (Ct , 1 − lt )

2 Euler Equation (savings-consumption margin)

u10 (Ct , 1 − lt ) = (1 + F10 (Kt +1 , lt +1 ) − δ) βu10 (Ct +1 , 1 − lt +1 )

3 From the BC, the LM and the zero profit condition we obtain

Kt +1 = F (Kt , lt ) − Ct + (1 − δ)Kt

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 25 / 48
 Competitive Equilibrium

Competitive Equilibrium and Pareto Optima

It is now clear that the equations that determine the CE are the same
than those for the efficient allocations.
So we have shown the First Welfare Theorem for our economy: “If
there are no distortions (distotionary taxes, asymetry of information,
market power...) any CE is efficient”.
But this is not always the case. We will study how the decissions
taken by agents are affected once distortions are introduced in our
setting.
Before we study now the properties of the equilibrium solution.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 26 / 48
 Competitive Equilibrium

Equilibrium Dynamics

The equlibrium solution of the model is a set of sequences

{Ct , lt , Kt +1 }t∞=0 :
C0 , C1 , C2 , ...
l0 , l1 , l2 , ...
K0 , K1 , , K2 , ...
For every two periods t and t + 1 we have 3 equations (from the foc).
Thus between period 0 and 1 we have
1 6 unknowns: C0 , C1 , l0 , l1 , K0 , K1
2 and 4 equations.
Between period 1 and 2
1 3 more unknowns: C2 , l2 , K2
2 and 3 more equations
and so on ... thus we need two equations for the system to be
determinate: The initial condition and the transversality condition.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 27 / 48
 Competitive Equilibrium

Transitional Dynamics

Different initial choices for consumption (C0 ), lead to different equilibrium

paths for a same initial condition K0 (Assume l constant).

( K ee ,C ee )

2
C0

2
C0
1
C0

K0 K
Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 28 / 48
 Competitive Equilibrium

Transitional Dynamics

Only one value for C0 satisfies the transversality condition for a given
initial condition K0 .
It can be shown that the equilibrium path that satisfies the TC is the
one that converges to the steady state (K ∗ , C ∗ ).

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 29 / 48
 Competitive Equilibrium

Steady State I

A dynamical system is in a steady state if the variables that define the

behavior of the system are unchanging in time, that is, ∀ t = 0...∞
it must be that
Ct = Ct +1 = C ∗
lt = lt +1 = l ∗
Kt = Kt +1 = K ∗
To obtain the steady state values (K ∗ , C ∗ , l ∗ ) one must impose these
three conditions in the 3 equations obtained from the foc:
1 Consumption-leisure choice

u20 (C ∗ , 1 − l ∗ ) = Fl0 (K ∗ , l ∗ )u10 (C ∗ , l ∗ )

2 Euler equation

u10 (C ∗ , 1 − l ∗ ) = (1 + FK0 (K ∗ , l ∗ ) − δ) βu10 (C ∗ , 1 − l ∗ )

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 30 / 48
 Competitive Equilibrium

Steady State II

3 Resource constraint

K ∗ = F (K ∗ , l ∗ ) − C ∗ + (1 − δ )K ∗

From the Euler equation we obtain the capital-labor ratio (with a

CRS production function)

FK0 (K ∗ , l ∗ ) = 1/β − 1 + δ

From the RC
C ∗ = F (K ∗ , l ∗ ) − δK ∗

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 31 / 48
 Competitive Equilibrium

Exercise

In a model without leisure (l = 1).

Compute K ∗ y C ∗ , for a Cobb-Douglas production function
1− σ
Yt = AKtα L1−α (and utility function u (Ct ) = C1t−σ ).

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 32 / 48
 Competitive Equilibrium

Dynamical Systems I

For dynamical systems we define two types of variables:

State (or predetermined) variables: Kt . The value of these variables
is fixed at time t.
Control variables: Ct and lt . These are the choice variables at t.
As mentioned above the solution to the optimization problem is a set
of sequences {Ct , lt , Kt +1 }t∞=0 .
But we may also want to express the equilibrium solution as a set of
recursive policy functions (whenever it is possible...?)
1 For capital
Kt +1 = gk (Kt )
2 For consumption
Ct = gc (Kt )
3 And for leisure
lt = gl (Kt )

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 33 / 48
 Competitive Equilibrium

Dynamical Systems II

Thus for a given value of the state variable we can compute the
contemporaneous choice for Ct and lt and the next period’s value for
the state Kt +1 .
The bad news are that for most models we are not able to compute
neither the equilibrium sequences nor the policy functions. (see
problem set)
Thus we have to resort to numerical methods to approximate the
equilibrium solution: (with the aid of computers!!!)
1 Dynamic Programming
Value function iteration
Policy function iteration
2 Euler equation
Log-linearization
Perturbation methods
Projection methods

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 34 / 48
 Competitive Equilibrium

Taxes

Now we are going to consider an economy with distotionary taxes.

There are taxes to capital and labor income and consumption.
This taxes affect the consumer budget constraint.

(1 + τc )Ct + It = (1 − τl )wt lts + (1 − τk )rt Kts + πt + Tt

∀ t = 0...∞

We can solve the competitive equilibrium and find the foc that solve
the maximization problem.
How do taxes change the optimality conditions? (see problem set)
What about taxes on income and/or taxes on investment? (see
problem set)

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 35 / 48
 Competitive Equilibrium

Phase Diagram (Cobb-Douglas, no Leisure)

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 36 / 48
 Competitive Equilibrium

Dynamics: Cobb-Douglas Example

In a model without leisure (l = 1), with production function

1− σ
Yt = AKtα L1−α and utility function u (Ct ) = C1t−σ , the dynamics is
given by:
Euler equation
Ct +1 1/σ
= (1 + αA[AKtα − Ct + (1 − δ)Kt ]α−1 − δ) β

Ct
Resource constraint

Kt +1 − Kt = AKtα − Ct − δKt

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 37 / 48
 Competitive Equilibrium

Shooting

A basic method for solving two-point boundary value problems is

shooting.
If we knew the optimal initial choice C0 besides the initial condition
K0 we could solve all the dynamic equations of our model.
However, this is not the case. But in a simple one sector growth
model we know that the optimal initial choice C0 is such that the TC
is satisfied and we have seen (not proved) that this implies that the
optimal solution converges to the stationary solution
Thus, we can make a guess on C0 and “hope” that it is the one that
hits the ss. However, this is not usually going to happen (in fact the
probability that it happens is zero, why?)
Instead, we can make a guess C0a and check when the solution is
going “too far” from the ss, then we can update the initial guess to
C0b so that we get closer to the ss
Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 38 / 48
 Competitive Equilibrium

Shooting (Dynamics)

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 39 / 48
 Competitive Equilibrium

Shooting: Algorithm (No leisure in Util.)

Step 1: Given a K0 < K ∗ choose an initial

0(= Cmin ) ≤ C0a ≤ F (K0 , L)(= Cmax ). Choose a stopping criterion
e>0
Step 2: Compute the model, that is, the sequences {Ct } and {Kt }
until Ct +1 < Ct or Kt +1 < Kt , and denote this period as T
Step 3: If | C (T ) − C ∗ |< e, STOP, otherwise
1 if C
t +1 < Ct then set Cmin = C0a and a new C0b = (Cmin + Cmax ) /2,
2 if K
t +1 < Kt set Cmax = C0a and a new C0b = (Cmin + Cmax ) /2
go to step 2

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 40 / 48
 Competitive Equilibrium

Reverse Shooting

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 41 / 48
 Competitive Equilibrium

Shooting: Evaluation

In favor
We can approximate the true solution as closely as desired
We only (?) need to have sufficient knowledge of the phase diagram
and the stationary solution
Against
We need a final condition (in this case a steady state)
If the dynamics is very complex (or unknown) it can be useless:
Cycles, unstable ss,...
If there are many control variables the computational cost can be high
It is not useful if there are stochastic processes

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 42 / 48
 Competitive Equilibrium

Linearization I
Another way to study the dynamics of a non-linear system is to
linearize it around the steady state. Suppose we have the following
non-linear system

Ct + 1 = g ( Ct , K t )
Kt +1 = h (Ct , Kt )

We can construct a Taylor expansion around the steady state

Ct +1 = c ∗ + gC0 (C ∗ , K ∗ )(Ct − C ∗ ) + gK0 (C ∗ , K ∗ )(Kt − K ∗ )

Kt +1 = K ∗ + hC0 (C ∗ , K ∗ )(Ct − C ∗ ) + hK
0
(C ∗ , K ∗ )(Kt − K ∗ )

and defining steady state deviations C̃t = Ct − C ∗ , K̃t = Kt − K ∗

C̃t +1 = gC0 (C ∗ , K ∗ )C̃t + gK0 (C ∗ , K ∗ )K̃t

K̃t +1 = hC0 (C ∗ , K ∗ )C̃t + hK
0
(C ∗ , K ∗ )K̃t
Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 43 / 48
 Competitive Equilibrium

Linearization II

or in matrix notation
Ct − C ∗ gC0 gK0
   
xt +1 = Axt , where xt = and A =
Kt − K ∗ hC0 0
hK

The stability (convergence to the s.s.) of linear systems depends on

the eigenvalues of matrix A.
a) If the number of negative eigenvalues is equal to the
number of state variables the system is determinate. In
our case it means that the stable manifold has
dimension 1.
b) If the number of negative eigenvalues is greater than the
number of state variables the system is indeterminate.
In our case it means that the dimension of the stable
manifold is greater than 1.

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 44 / 48
 Competitive Equilibrium

Linearization III

c) If the number of negative eigenvalues is zero the s.s. is

a source (unstable).

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 45 / 48
 Competitive Equilibrium

Implicit Function Theorem I

Let H (x, y ) = 0, where y = h (x ) is an implicit function

H : Rn × Rm → Rm
h : Rn → Rm , but h (·) is unknown and is defined by H (x, h (x )) = 0
Assume that we only know h (x0 ) = y0
Differentiating leads to Dx H (x, h (x )) = 0 or

Hx (x, h (x )) + Hy (x, h (x ))hx (x ) = 0 (1)

This is useless yet since h (x ) is unknown, but we know that at x0 h (·)

takes the value h (x0 ), and therefore we can compute hx (x0 ) from (1)
At x0 , if Hy (x0 , y0 ) is nonsingular we have

hx (x0 ) = −Hy (x0 , y0 )−1 Hx (x0 , y0 ) (2)

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 46 / 48
 Competitive Equilibrium

Implicit Function Theorem II

Thus, we can construct a linear approximation of h (x )

hL (x ) = h (x0 ) + hx (x0 )(x − x0 )

We can check whether (unit free) residuals H (x, hL (x )) = 0 are close

to zero. If they are not small enough we can increase the degree of
the approximation
We have computed Dx H (x, h (x )) = 0 in (1), we can differentiate
again to obtain Dxx H (x, h (x )) = 0, which yields

Hxx + 2Hxy hx + Hyy hx hx + Hy hxx = 0

We do not know h (·) but we know y0 = h (x0 ) and we have computed

hx (x0 ), we can then find
0
hxx (x0 ) = −(Hy0 )−1 (Hxx
0 0 0
+ 2Hxy 0 0 0
hx + Hyy hx hx )
Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 47 / 48
 Competitive Equilibrium

Implicit Function Theorem III

We construct the quadratic approximation

1
hQ (x ) = h (x0 ) + hx0 (x − x0 ) + (x − x0 )T hxx
0
( x − x0 )
2
and we can check the quality of the approximation by looking at the
residuals H (x, hQ (x ))

Advanced Macroeconomics A Macroeconomic Model: The Neoclassical Growth Model September 2024 48 / 48