The Neoclassical Growth Model

Alessandro Spiganti
Universitˋa Ca’ Foscari di Venezia

September 2021

Macroeconomics I
The Neoclassical Growth Model

- The exogenous saving rate plays a key role in

determining the properties of the balanced
growth path in the Solow model

- The Ramsey (or Ramsey-Cass-Koopmans)

model differs from the Solow model because it
explicitly models the consumer side and
endogenises savings

- By specifying the preference orderings of

individuals, it enables better understanding of
the factors affecting savings and allows for
welfare comparisons and efficiency analyses

2 / 66
The Neoclassical Growth Model

- Not only a basic growth model, but also a workhorse model for many areas of
macroeconomics (growth, fluctuations, monetary policy, climate change, asset price,
distributional issues...)

- References: Romer (2019, Chapter 2A), Barro and Sala-I-Martin (1995, Chapter 2),
Acemoglu (2008, Chapter 5-8, advanced)

3 / 66
Framework

4 / 66
Production I

- The production side of the economy mirrors that of the Solow model
- Factors and product markets are competitive
- There is a large number of identical firm, allowing us to consider a representative firm
- Each firm produces according to

Y (t ) = F (K (t ), A(t )L(t ))

- Assumptions
- the production function exhibits diminishing marginal products in each factor and
constant returns to scale, and satisfies the Inada conditions
- A(t ) grows at rate g and L(t ) grows at rate n

5 / 66
Production II

- Given the constant returns to scale assumption, we can write the production function
in intensive form  
K (t )
y (t ) = F , 1 ≡ f (k (t )) (1)
A(t )L(t )

- y (t ) ≡ Y (t )/(A(t )L(t )) and k (t ) ≡ K (t )/(A(t )L(t )) are output and capital per
efficiency unit of labour

6 / 66
Production III

- The representative firm employs labour and capital, pays them their marginal
products, and sells the resulting output

R (t ) = f 0 (k (t )) (2a)
ω (t ) = A(t ) f (k (t )) − k (t )f 0 (k (t ))
 
(2b)

7 / 66
Production III

- The representative firm employs labour and capital, pays them their marginal
products, and sells the resulting output

R (t ) = f 0 (k (t )) (2a)
ω (t ) = A(t ) f (k (t )) − k (t )f 0 (k (t ))
 
(2b)

7 / 66
Production III

- The representative firm employs labour and capital, pays them their marginal
products, and sells the resulting output

R (t ) = f 0 (k (t )) (2a)
ω (t ) = A(t ) f (k (t )) − k (t )f 0 (k (t ))
 
(2b)

- Since capital depreciates at rate δ, the interest rate is

r (t ) = R (t ) − δ = f 0 (k (t )) − δ (3)

7 / 66
Households I

8 / 66
Households I

- Consider an economy consisting of a unit measure of infinitely-lived households

- There is an uncountable number of households e.g. the set of households is H = [0, 1]
- Each agent is infinitesimal and will have no effects on aggregate
- The advantage of the unit measure is that average and aggregate are the same

8 / 66
Households II

- Households are identical

- Each household has L(t ) family members with

L(t ) = exp (nt ) i.e. L(0) = 1 (4)

- C (t ) is total (household) consumption at time t

- ς(t ) ≡ C (t )/L(t ) is consumption per person (or member of the household)
- Let c (t ) be consumption per efficiency unit of labour i.e.

C (t ) ς (t )
c (t ) ≡ = (5)
A(t )L(t ) A(t )

9 / 66
Households’ Utility I

- Each person has the instantaneous utility function u (ς), such that
- u 0 (ς) > 0
- u 00 (ς) < 0, i.e. agents prefer to smooth consumption
- it satisfies Inada-type conditions

10 / 66
Households’ Utility I

10 / 66
Households’ Utility II

- Each household wishes to maximize lifetime utility, U

Z ∞
U= u (ς(t )) exp (nt )exp (−ρt )dt (6)
t =0

11 / 66
Households’ Utility II

- Each household wishes to maximize lifetime utility, U

Z ∞
U= u (ς(t )) exp (nt )exp (−ρt )dt (6)
t =0

- The household receives an utility of u (ς(t )) per household member at t

11 / 66
Households’ Utility II

- Each household wishes to maximize lifetime utility, U

Z ∞
U= u (ς(t )) exp (nt )exp (−ρt )dt (6)
t =0

- The household receives an utility of u (ς(t )) per household member at t

- This corresponds to a total utility of u (ς(t )) L(t ) = u (ς(t )) exp(nt )

11 / 66
Households’ Utility II

- Each household wishes to maximize lifetime utility, U

Z ∞
U= u (ς(t )) exp (nt )exp (−ρt )dt (6)
t =0

- The household receives an utility of u (ς(t )) per household member at t

- This corresponds to a total utility of u (ς(t )) L(t ) = u (ς(t )) exp(nt )
- exp (−ρt ) represents discounting future utils at the subjective discount rate ρ (or rate of
time preference or rate of impatience)
- Assume ρ − n > 0
- One unit of consumption today is valued more than one unit of consumption tomorrow

11 / 66
Households’ Supply of Labour

- Each adult supplies inelastically one unit of labour service per unit of time
- The wage income per adult person is ω (t )
- Total labour income of the household is ω (t )L(t )

12 / 66
Households’ Assets I

- Households hold assets in the form of physical capital K (t ) and loans (bonds) B (t )
- The economy is closed, so households can only lend and borrow from other households
- In the absence of asymmetry and uncertainty, a no-arbitrage condition ensures that each
asset pays the same return r (t )
- Total asset income is
r (t ) (K (t ) + B (t )) ≡ r (t )A(t ) (7)
- Bonds will not actually be used in equilibrium, they are only added for pedagogical
reasons

13 / 66
Households’ Assets II

- A household faces the following flow budget constraint at all dates

ω (t )L(t ) + r (t )A(t ) = C (t ) + Ȧ(t ) (8)

| {z } | {z } |{z} | {z }
labour income asset income consumption net asset accumulation

14 / 66
Households’ Assets III

- Let the household’s net assets per person be a(t ) ≡ A(t )/L(t ). This evolves
according to  
A(t )
d L(t ) Ȧ(t )
ȧ(t ) = = − na(t ) (9)
dt L(t )

15 / 66
Households’ Assets III

- Let the household’s net assets per person be a(t ) ≡ A(t )/L(t ). This evolves
according to  
A(t )
d L(t ) Ȧ(t )
ȧ(t ) = = − na(t ) (9)
dt L(t )

- Combine (8) and (9) to get the budget constraint in per capita terms

ȧ(t ) = ω (t ) + (r (t ) − n) a(t ) − ς(t ) (10)

15 / 66
Households’ Assets IV

- This is not a proper budget constraint on the individual, as it does not rule out Ponzi
games
- The household can borrow to finance current consumption and then use future
borrowings to roll over the principal and pay all the interest
- Since no principal ever gets repaid, today’s consumption is free
- The household could borrow to finance an arbitrary high level of consumption in
perpetuity

16 / 66
Households’ Assets V

- The infinite-horizon no Ponzi game condition is that the present value of assets must be
asymptotically non-negative
 Z t 
lim a(t ) exp − (r (s ) − n) ds ≥ 0 (11)
t →∞ 0

- In the long-run, a household’s debt per person cannot grow as fast as r (t ) − n, so that the
level of debt cannot grow as fast as r (t )
- However, asymptotically no individual would ever want to have positive wealth (they
would consume that instead), so that (11) holds with equality

17 / 66
Definition of Equilibrium I

- Formally, the household should choose how many bonds to hold, B (t )

- But all households are alike, so in general equilibrium there cannot be borrowers and
lenders
- The bond market will not be used in equilibrium, and A(t ) = K (t )
- The household divides its income (from the labour and capital it supplies) at each point
in time between consumption and saving in capital so as to maximize its lifetime utility
- Output is either consumed or invested
- The firms hire workers and rent capital in competitive factor markets as to maximize
their profits

18 / 66
Definition of Equilibrium II

- A competitive equilibrium of the Ramsey economy consists of paths

[c (t ), k (t ), ω (t ), R (t )]t∞=0 , such that the representative household maximizes lifetime
utility (6) subject to the flow constraint (10) and the no-Ponzi game condition (11),
given an initial capital per efficiency unit k (0), factor prices evolving according to (2),
an interest rate given by (3), and a given sequence of {L(t ), A(t )}t∞=0

19 / 66
The Household’s Maximization Problem

20 / 66
Household Maximization

- The representative household wishes to maximize lifetime utility in (6)

- choosing a stream of consumption [ς(t )]t∞=0
- subject to the flow constraint (10) and the no-Ponzi game condition (11)
- This is dynamic problem set in continuous time
- The methodology used to solve these problems is the maximum principle of optimal
control
- The following slides outline the procedure to find the first-order conditions

21 / 66
Household Maximization: Optimal Control I

- Step 0: Construct the Hamiltonian function

H (a(t ), ς(t ), t, λ(t )) ≡ e (n−ρ)t u (ς(t ))+λ(t )ȧ(t )

= e(n−ρ)t u (ς(t )) + λ(t )[ω (t ) + (r (t ) − n)a(t ) − ς(t )] (12)
(n − ρ )t
=e {u (ς(t )) + µ(t ) [ω (t ) + (r (t ) − n)a(t ) − ς(t )]}

22 / 66
Household Maximization: Optimal Control I

- Step 0: Construct the Hamiltonian function

H (a(t ), ς(t ), t, λ(t )) ≡ e (n−ρ)t u (ς(t ))+λ(t )ȧ(t )

= e(n−ρ)t u (ς(t )) + λ(t )[ω (t ) + (r (t ) − n)a(t ) − ς(t )] (12)
(n − ρ )t
=e {u (ς(t )) + µ(t ) [ω (t ) + (r (t ) − n)a(t ) − ς(t )]}

- Start with the utility function in t

22 / 66
Household Maximization: Optimal Control I

- Step 0: Construct the Hamiltonian function

H (a(t ), ς(t ), t, λ(t )) ≡ e (n−ρ)t u (ς(t ))+λ(t )ȧ(t )

= e(n−ρ)t u (ς(t )) + λ(t )[ω (t ) + (r (t ) − n)a(t ) − ς(t )] (12)
(n − ρ )t
=e {u (ς(t )) + µ(t ) [ω (t ) + (r (t ) − n)a(t ) − ς(t )]}

- Start with the utility function in t

- Add a Lagrange multiplier times the change in assets

22 / 66
Household Maximization: Optimal Control I

- Step 0: Construct the Hamiltonian function

H (a(t ), ς(t ), t, λ(t )) ≡ e (n−ρ)t u (ς(t ))+λ(t )ȧ(t )

= e(n−ρ)t u (ς(t )) + λ(t )[ω (t ) + (r (t ) − n)a(t ) − ς(t )] (12)
(n − ρ )t
=e {u (ς(t )) + µ(t ) [ω (t ) + (r (t ) − n)a(t ) − ς(t )]}

- Start with the utility function in t

- Add a Lagrange multiplier times the change in assets

22 / 66
Household Maximization: Optimal Control I

- Step 0: Construct the Hamiltonian function

H (a(t ), ς(t ), t, λ(t )) ≡ e (n−ρ)t u (ς(t ))+λ(t )ȧ(t )

= e(n−ρ)t u (ς(t )) + λ(t )[ω (t ) + (r (t ) − n)a(t ) − ς(t )] (12)
(n − ρ )t
=e {u (ς(t )) + µ(t ) [ω (t ) + (r (t ) − n)a(t ) − ς(t )]}

- Start with the utility function in t

- Add a Lagrange multiplier times the change in assets
- Rewrite in terms of µ(t ) = e −(n−ρ)t λ(t )
- The costate variable λ(t ) represents the value of the asset (an increment in income) at t in
units of time-zero utils
- The current-state costate variable µ(t ) = e −(n−ρ)t λ(t ) represents the value of the asset at t in
terms of current-value prices

22 / 66
Household Maximization: Optimal Control II

- The Hamiltonian captures the maximization problem of the household at time t

H (a(t ), ς(t ), t, µ(t )) = e (n−ρ)t {u (ς(t )) + µ(t ) [ω (t ) + (r (t ) − n)a(t ) − ς(t )]} (13)

- The first term captures the direct contribution of consumption on utility

- The second term captures the change in assets due to the choice of consumption

23 / 66
Household Maximization: Optimal Control III

- Step 1: it will be convenient to restructure the problem in terms of current-value

Hamiltonian

Ĥ (a(t ), ς(t ), µ(t )) ≡ u (ς(t )) + µ(t ) [ω (t ) + (r (t ) − n)a(t ) − ς(t )] (14)

- Now consider an infinitesimal change in the paths of the control variable with a
corresponding change in the paths of the state variable

24 / 66
Household Maximization: Optimal Control IV

- Step 2: Take the derivative of the Hamiltonian with respect to the control variable and
set it to zero

Ĥc (a, ς, µ) = u 0 (ς(t )) − µ(t ) = 0

(15)
i.e. u 0 (ς(t )) = µ(t )

25 / 66
Household Maximization: Optimal Control V

- Step 3: Take the derivative of the Hamiltonian with respect to the state variable and
set it equal to −λ̇

Ĥa (a, ς, µ) = µ(t )(r (t ) − n) = −λ̇

(16)
= −µ̇(t ) + (ρ − n)µ(t )

- Rearrange (16),
µ̇(t )
= −(r (t ) − ρ) (17)
µ (t )

26 / 66
Household Maximization: Optimal Control VI

- Step 4: Add a transversality condition

- Equation (17) and the transition equation for ȧ in (9) form a system of two differential
equations in the variables µ̇ and ȧ
- For the system to be determined, we need two boundary conditions
- One initial condition is given by the initial value of the state variable, i.e. k (0)
- One terminal condition is given by the transversality condition

27 / 66
Household Maximization: Optimal Control VII

- The transversality condition in the infinite-horizon case with discounting is

lim e (n−p)t µ(t )a(t ) = 0 (18)

t →∞

- The intuitive explanation is that the value of the asset must be asymptotically 0
- Otherwise, something valuable would be left over

28 / 66
Household Maximization: The Transversality Condition

- Integrate (17) and combine it with (15) evaluated at 0

- Substitute the obtained expression for µ(t ) into the transversality condition (18), you
get
 Z t 
i.e. lim a(t ) exp − (r (s ) − n) ds = 0 (19)
t →∞ 0

- The quantity of assets per person does not grow asymptotically at a rate as high as r − n
- It would be suboptimal for households to accumulate positive assets forever, because
utility would increase if these assets were instead consumed in finite time

29 / 66
Household Maximization: Recap

- The current-value Hamiltonian

Ĥ (a(t ), ς(t ), µ(t )) ≡ u (ς(t )) + µ(t ) [ω (t ) + (r (t ) − n)a(t ) − ς(t )] (20)

- The first-order conditions

Ĥc (a, ς, µ) = u 0 (ς(t )) − µ(t ) = 0 (21a)

Ĥa (a, ς, µ) = µ(t )(r (t ) − n) = −µ̇(t ) + (ρ − n)µ(t ) (21b)

- The transversality condition

lim e (n−p)t µ(t )a(t ) = 0 (22)

t →∞

30 / 66
Household Maximization: First-Order Conditions

- Condition (21a) implies

u 0 (ς(t )) = µ(t ) (23)

u 00 (ς(t )) ς(t ) ς̇(t ) µ̇(t )

= = −(r (t ) − ρ)
u 0 (ς(t )) ς(t ) µ (t )

31 / 66
Household Maximization: First-Order Conditions

- Condition (21a) implies

u 0 (ς(t )) = µ(t ) (23)

- Differentiate with respect to time, divide by µ(t ), and multiply LHS by ς(t )/ς(t )

u 00 (ς(t )) ς(t ) ς̇(t ) µ̇(t )

= = −(r (t ) − ρ)
u 0 (ς(t )) ς(t ) µ (t )

31 / 66
Household Maximization: First-Order Conditions

- Condition (21a) implies

u 0 (ς(t )) = µ(t ) (23)

- Differentiate with respect to time, divide by µ(t ), and multiply LHS by ς(t )/ς(t )

u 00 (ς(t )) ς(t ) ς̇(t ) µ̇(t )

= = −(r (t ) − ρ)
u 0 (ς(t )) ς(t ) µ (t )

31 / 66
Household Maximization: First-Order Conditions

- Condition (21a) implies

u 0 (ς(t )) = µ(t ) (23)

- Differentiate with respect to time, divide by µ(t ), and multiply LHS by ς(t )/ς(t )

u 00 (ς(t )) ς(t ) ς̇(t ) µ̇(t )

= = −(r (t ) − ρ)
u 0 (ς(t )) ς(t ) µ (t )

31 / 66
Household Maximization: First-Order Conditions

- Condition (21a) implies

u 0 (ς(t )) = µ(t ) (23)

- Differentiate with respect to time, divide by µ(t ), and multiply LHS by ς(t )/ς(t )
- The RHS is equal to −(r (t ) − ρ) by equation (17)

u 00 (ς(t )) ς(t ) ς̇(t ) µ̇(t )

= = −(r (t ) − ρ)
u 0 (ς(t )) ς(t ) µ (t )

31 / 66
Household Maximization: The Euler Equation I

- Rearranging, we obtain an Euler equation

ς̇(t ) −u 0 (ς(t ))
= (r (t ) − ρ) 00 (24)
ς (t ) u (ς(t )) ς(t )

- Euler equations characterize optimising behaviour in problems with intertemporal

trade-offs
- Here, it determines the growth rate of consumption when the household is maximizing
utility

32 / 66
Household Maximization: The Euler Equation II

−u 0 (ς(t ))
 
ς̇(t )
= (r (t ) − ρ )
ς (t ) u 00 (ς(t )) ς(t )
| {z }
positive

- Consumption rises if the real return exceeds the rate at which the household
discounts future consumption
- It changes more the larger is the magnitude of the elasticity of intertemporal substitution

33 / 66
Household Maximization: The Euler Equation III

−u 0 (ς(t ))
 
ς̇(t ) 1
=
ς (t ) r (t ) − ρ u 00 (ς(t )) ς(t )

- Intuitively, imagine the household reducing consumption by an infinitesimal amount,

investing this for an infinitesimal amount of time, and then consuming the proceeds
- If the household is optimizing, the marginal impact of this change on lifetime utility must
be zero
- If the impact is strictly positive, the household can marginally raise its lifetime utility by
making the change
- If the impact is strictly negative, the household can raise its lifetime utility by making the
opposite change

34 / 66
Household Maximization: Utility Function I

ς̇ −u 0 (ς(t ))
= (r − ρ) 00
ς u (ς(t )) ς(t )

- To find a steady state in which r (t ) and ς̇(t )/ς(t ) are constant, we need the
intertemporal elasticity of substitution to be constant
- A common functional form with this property is the constant intertemporal elasticity of
substitution (CIES) utility function

ς (t )1− θ − 1
u (ς(t )) = (25)
1−θ
- The elasticity of substitution for this utility function is the constant 1/θ, with θ > 0

35 / 66
Household Maximization: Utility Function II

- The Euler equation then simplifies to

ς̇(t ) 1
= (r (t ) − ρ ) (26)
ς (t ) θ

36 / 66
Get To Know Your Utility Function I

- We will often use this particular functional form for the utility function, known as
constant intertemporal elasticity of substitution (CIES) or constant relative risk aversion
(CRRA) ( 1− θ
c −1
1− θ if θ 6= 1 and θ ≥ 0
u (c ) = (27)
ln(c ) if θ = 1

37 / 66
Get To Know Your Utility Function I

- We will often use this particular functional form for the utility function, known as
constant intertemporal elasticity of substitution (CIES) or constant relative risk aversion
(CRRA) ( 1− θ
c −1
1− θ if θ 6= 1 and θ ≥ 0
u (c ) = (27)
ln(c ) if θ = 1

- The first derivative is u 0 (c ) = c −θ > 0

- The second derivative is u 00 (c ) = −θc −θ −1 < 0

37 / 66
Get To Know Your Utility Function I

- We will often use this particular functional form for the utility function, known as
constant intertemporal elasticity of substitution (CIES) or constant relative risk aversion
(CRRA) ( 1− θ
c −1
1− θ if θ 6= 1 and θ ≥ 0
u (c ) = (27)
ln(c ) if θ = 1

- The first derivative is u 0 (c ) = c −θ > 0

- The second derivative is u 00 (c ) = −θc −θ −1 < 0
- The Arrow-Pratt coefficient of relative risk aversion is constant,

u 00 (c )c
− =θ
u 0 (c )

37 / 66
Get To Know Your Utility Function II

- Suppose the utility function is separable, so that if we have two goods, c1 and c2 , total
utility is given by
U = u ( c1 ) + u ( c2 )

38 / 66
Get To Know Your Utility Function II

- Suppose the utility function is separable, so that if we have two goods, c1 and c2 , total
utility is given by
U = u ( c1 ) + u ( c2 )

- The marginal rate of substitution between these two goods is

−θ
c1−θ
1/θ
u 0 (c1 ) u 0 ( c1 )
 
c1 c1
= = ↔ =
u 0 (c2 ) c2−θ c2 c2 u 0 ( c2 )

38 / 66
Get To Know Your Utility Function II

- Suppose the utility function is separable, so that if we have two goods, c1 and c2 , total
utility is given by
U = u ( c1 ) + u ( c2 )

- The marginal rate of substitution between these two goods is

−θ
c1−θ
1/θ
u 0 (c1 ) u 0 ( c1 )
 
c1 c1
= = ↔ =
u 0 (c2 ) c2−θ c2 c2 u 0 ( c2 )

- By definition, the elasticity of substitution is also constant

38 / 66
Equilibrium

39 / 66
Equilibrium I

- The equilibrium prices are given by

R (t ) = f 0 (k (t ))
ω (t ) = A(t ) f (k (t )) − k (t )f 0 (k (t ))
 

- The interest rate by

r (t ) = R (t ) − δ = f 0 (k (t )) − δ

40 / 66
Equilibrium II

- We need a law of motion for k (t )

41 / 66
Equilibrium II

- We need a law of motion for k (t )

- We have a law of motion for a(t )

ȧ(t ) = ω (t ) + (r (t ) − n) a(t ) − ς(t )

41 / 66
Equilibrium II

- We need a law of motion for k (t )

- We have a law of motion for a(t )

ȧ(t ) = ω (t ) + (r (t ) − n) a(t ) − ς(t )

- Remember that bonds are not used in equilibrium, so

A(t ) K (t )
a (t ) ≡ = = k (t )A(t )
L(t ) L(t )

41 / 66
Equilibrium II

- We need a law of motion for k (t )

- We have a law of motion for a(t )

ȧ(t ) = ω (t ) + (r (t ) − n) a(t ) − ς(t )

- Remember that bonds are not used in equilibrium, so

A(t ) K (t )
a (t ) ≡ = = k (t )A(t )
L(t ) L(t )

- Assuming Ȧ(t )/A(t ) = g,

ȧ(t ) = k̇ (t )A(t ) + k (t )gA(t )

41 / 66
Equilibrium III

- Thus,

ω (t ) + (r (t ) − n) a(t ) − ς(t ) = k̇ (t )A(t ) + k (t )gA(t )

→ A(t )[f (k (t ))−k (t )f 0 (k (t ))]+(f 0 (k (t ))−δ−n)k (t )A(t )−c (t )A(t )=k̇ (t )A(t )+k (t )gA(t )

42 / 66
Equilibrium III

- Thus,

ω (t ) + (r (t ) − n) a(t ) − ς(t ) = k̇ (t )A(t ) + k (t )gA(t )

→ A(t )[f (k (t ))−k (t )f 0 (k (t ))]+(f 0 (k (t ))−δ−n)k (t )A(t )−c (t )A(t )=k̇ (t )A(t )+k (t )gA(t )

- Dividing throughout by A(t ),

k̇ (t ) = f (k (t )) − (δ + n + g ) k (t ) − c (t ) (29)

- This is the flow constraint for the overall economy

- It is the key relation that determines the evolution of k (t ) and y (t ) = f (k (t ))

42 / 66
Equilibrium IV
- In the Solow model, the evolution of c (t ) was determined by the constant savings rate
- Here, we know that ς(t ) grows in accordance with the Euler equation,

ς̇(t ) 1
= (r (t ) − ρ )
ς (t ) θ

43 / 66
Equilibrium IV
- In the Solow model, the evolution of c (t ) was determined by the constant savings rate
- Here, we know that ς(t ) grows in accordance with the Euler equation,

ς̇(t ) 1
= (r (t ) − ρ )
ς (t ) θ

- Since ς(t ) = c (t )A(t ),

ċ (t ) ς̇(t )
= −g
c (t ) ς (t )

43 / 66
Equilibrium IV
- In the Solow model, the evolution of c (t ) was determined by the constant savings rate
- Here, we know that ς(t ) grows in accordance with the Euler equation,

ς̇(t ) 1
= (r (t ) − ρ )
ς (t ) θ

- Since ς(t ) = c (t )A(t ),

ċ (t ) ς̇(t )
= −g
c (t ) ς (t )

- Therefore,
ċ (t )
1
= f 0 (k (t )) − δ − ρ − gθ (30)
c (t ) θ

- This is the other key relationship, determining the evolution of c (t )

43 / 66
Equilibrium IV

- The two key relationships in the Ramsey model are

k̇ (t ) = f (k (t )) − (δ + n + g ) k (t ) − c (t )
ċ (t )
1
= f 0 (k (t )) − δ − ρ − gθ
c (t ) θ

- These form a system of two differential equations in k (t ) and c (t )

- Together with the initial condition k (0) and the transversality condition, this system
determines the time path of k (t ) and c (t )

44 / 66
Equilibrium: The Transversality Condition

- We can also re-write the transversality condition in (18) in terms of

k (t ) = a(t ) exp(−gt )
 Z t 
0


lim exp − f (k (s )) − δ − n − g ds k (t ) = 0 (31)
t →∞ 0

- k (t ) will be constant in a steady-state, k ?

- For (31) to be satisfied in the limit, we need the following assumption

f 0 (k ? ) − δ > n + g (32)

45 / 66
The Dynamics of c (t )

ċ (t )
1
= f 0 (k (t )) − δ − ρ − gθ
c (t ) θ

- In a steady-state ċ (t ) = 0, i.e.

f 0 (k ? ) = δ + ρ + gθ (33)

- This is a vertical line at k ? , the value that makes ċ = 0

- f 0 (k (t )) is monotonically decreasing in k
- ċ (t ) > 0 when k (t ) < k ?
- ċ (t ) < 0 when k (t ) > k ?

46 / 66
The Dynamics of c (t ): Graph

47 / 66
The Dynamics of c (t ): Graph

47 / 66
The Dynamics of k (t )

k̇ (t ) = f (k (t )) − (δ + n + g ) k (t ) − c (t )

- In a steady-state k̇ = 0, i.e.

c (t ) = f (k (t )) − (δ + n + g ) k (t ) (34)

- If c (t ) < f (k (t )) − (δ + n + g ) k (t ), then k̇ (t ) > 0

- If c (t ) > f (k (t )) − (δ + n + g ) k (t ), then k̇ (t ) < 0

48 / 66
The Dynamics of k (t ): Graph

49 / 66
The Dynamics of k (t ): Graph

49 / 66
The Dynamics of k (t ): Golden Rule

c (t ) = f (k (t )) − (δ + n + g ) k (t )

- c (t ) has a maximum at
f 0 (kgold ) = δ + n + g (35)

- Note that (32), (33), and (35) imply that k ? < kgold (modified golden rule)
- Inefficient oversaving cannot occur in the Ramsey world because the agents are
optimizing
- The optimizing agent saves less than to attain the golden rule value, because impatience
makes it not worthwhile to sacrifice more current consumption

50 / 66
The Phase Diagram

51 / 66
The Phase Diagram

51 / 66
The Phase Diagram

51 / 66
The Phase Diagram

51 / 66
The Phase Diagram: Initial Condition and Saddle Path

52 / 66
The Phase Diagram: Initial Condition and Saddle Path

52 / 66
The Phase Diagram: Initial Condition and Saddle Path

52 / 66
The Phase Diagram: Initial Condition and Saddle Path

52 / 66
The Phase Diagram: Initial Condition and Saddle Path

52 / 66
The Phase Diagram: Initial Condition and Saddle Path

52 / 66
The Saddle Path: Existence and Uniqueness
- The dynamic equilibrium follows the stable saddle path towards the steady-state pair
(k ? , c ? )
- There is some critical value c (0), for a given k (0), for which the economy converges to
the steady state
- This path is the only one that satisfies all the first-order conditions, including the
transversality condition
- Other paths are not equilibria

53 / 66
The Saddle Path: Existence and Uniqueness
- The dynamic equilibrium follows the stable saddle path towards the steady-state pair
(k ? , c ? )
- There is some critical value c (0), for a given k (0), for which the economy converges to
the steady state
- This path is the only one that satisfies all the first-order conditions, including the
transversality condition
- Other paths are not equilibria
- All trajectories satisfy the equations of motion for c and k
- But if the initial level of consumption is too high, c (t ) increases and eventually the capital
stock would reach zero: this violates feasibility
- And if the initial level of consumption is too low, c (t ) would eventually reach zero and
capital would accumulate, eventually violating the transversality condition

53 / 66
The Saddle Path: Existence and Uniqueness
- The dynamic equilibrium follows the stable saddle path towards the steady-state pair
(k ? , c ? )
- There is some critical value c (0), for a given k (0), for which the economy converges to
the steady state
- This path is the only one that satisfies all the first-order conditions, including the
transversality condition
- Other paths are not equilibria
- All trajectories satisfy the equations of motion for c and k
- But if the initial level of consumption is too high, c (t ) increases and eventually the capital
stock would reach zero: this violates feasibility
- And if the initial level of consumption is too low, c (t ) would eventually reach zero and
capital would accumulate, eventually violating the transversality condition
- Whereas k (0) is given, the initial consumption per capita must jump to c (0) on the
stable arm, and then (k, c ) monotonically travels along this arm towards the steady
state

53 / 66
The Steady State

- The behaviour of the economy once it has reached (k ? , c ? ) is the same as in Solow
- Capital, output, and consumption per unit of effective labour are constant
- Since y and c are constant, the savings rate (y − c )/y is also constant!
- Aggregate variables grow at n + g
- Output per worker and capital per worker grow at g
- Even with endogenous savings, growth in the effectiveness of labour remains the only
source of persistent growth in output per worker

54 / 66
A Note on Efficiency

- The decentralised equilibrium is Pareto efficient

- Since all households are identical, the decentralised outcome maximises welfare among
outcomes that treat families in an identical way
- You can prove it by constructing the social planner’s optimal control problem subject to
(29)
- Alternatively, you can appeal to the First and Second Welfare Theorem for economies
with a continuum of commodities, which shows that the decentralised equilibrium and
the efficient solution are the same (but you will see those in the Micro course)

55 / 66
Comparative Statics
An Example

56 / 66
The Shape of the Saddle Path I

- The saddle path (or stable arm) expresses the equilibrium c ? as a function of k ?
- It is a policy function, expressing the optimal value of the control variable to the state
variable
- The equivalent in the Solow model was c ? = (1 − s )f (k ? )
- Its exact shape depends on the parameters of the model

57 / 66
The Shape of the Saddle Path II

ċ (t )
1
= f 0 (k (t )) − δ − ρ − gθ
c (t ) θ

- Consider, for example, the effect of θ on the shape of the stable arm
- High values of θ imply that households have a strong preference for smoothing
consumption over time (if they are poor, they will consume a lot)
- Low values of θ imply that households households are more willing to postpone
consumption in response to high rates of return

58 / 66
The Shape of the Saddle Path III

59 / 66
Comparative Dynamics I

- An unexpected shock to the θ parameter may change the loci

ċ = 0 : f 0 (k ? ) = δ + ρ + gθ
k̇ = 0 : c (t ) = f (k (t )) − (δ + n + g ) k (t )

- θ enters the Euler equation but does not enter the law of motion for k , thus only the
locus for ċ = 0 is affected
- An increase in θ lowers the steady-value of k ? (remember that f 0 (k ) is decreasing)

60 / 66
Comparative Dynamics IV

61 / 66
Comparative Dynamics V

- Before the shock, the economy was in the old steady state
? (it’s a state)
- Immediately after the shock, k cannot change since it is given at k = kold
- However, c can change (it’s a control)
- c must jump up to the level implied by the new saddle path
- The economy then adjust towards the new steady state, (knew ? , c? )
new
?
- k gradually declines towards knew < k?
?
- c initially jumps up and then declines as well towards cnew < c?

62 / 66
Comparative Dynamics VI - Zooming In

63 / 66
Comparative Dynamics VII

- The levels of consumption and capital per efficiency unit of labour vary after the
shock
- Cross-country differences in income per capita could be due to differences in the
intertemporal elasticity of substitution, the discount rate, the depreciation rate, the
population growth rate, and the form of the production function
- Once the economy converges to the steady state, variables in efficiency unit of labour
are constant, c ? , k ? , y ? = f (k ? )
- Per capita variables grow at rate g
- Aggregate variables grow at rate n + g
- As in Solow, long run growth is determined exogenously by the growth rate of
labour-augmenting technological progress

64 / 66
Conclusions

65 / 66
The Ramsey Model: Recap

- The one-sector neoclassical growth model may be the most important model in
macroeconomics
- In the Solow model, the saving rate was exogenous
- The Ramsey model opens the black box of savings and capital accumulation by specifying
preferences for households
- Because preferences are explicitly stated, we can talk about efficiency and welfare
- It provides us with fundamental mathematical and conceptual tools
- Most importantly, the Ramsey model paves the way for further analysis of capital
accumulation, human capital investments, and endogenous technological change
- Technological progress is still exogenous
- We have no new insights on the sources of cross-country income differences and
economic growth
- We need more work!

66 / 66
Bibliography I

Acemoglu, D. (2008), Introduction to Modern Economic Growth, Princeton University Press, Princeton.
Barro, R., and X. Sala-I-Martin. (1995), Economic Growth, McGraw Hill, New York.
Romer, D. (2019), Advanced Macroeconomics, McGraw Hill, New York.