Honors - Economic Analysis III

Lecture 3: Neoclassical model

Simon Mongey

Winter, 2021

1
 This lecture

Issues with Solow model

Neoclassical model

- Ramsey (1928), Cass (1965), Koopmans (1965)

Optimality conditions

Steady state + Transition dynamics

Next: Decentralized economy, Welfare Theorems, Government

2
 Recap of Solow model
• Equilibrium conditions (without growth!)
yt = Af (kt ) = Aktα
yt = ct + it
kt+1 = (1 − δ)kt + ξf (kt )
ct = (1 − ξ)yt
• Solution
kt+1 = g(kt )
• Steady state
∗
 
k = g(k) , c = f (k) − δk −→ Golden rule −→ k , c∗
• Transition dynamics
∂g(k) k b ∂f (k) k b
kt+1 =
b kt , ct =
b kt = αk̂t
∂k ∂k
3
 Issues with Solow model

• Positive (descriptive)

• Within country over-time differences in ybt

• Cross country differences in ybt as functions of parameters {ξ, δ, α, A}

• Normative

• Households don’t optimize → Bad for policy

• E.g. If A increases for a few periods, shouldn’t ξ respond? Maybe households would like
to save some of a temporary increase in their income?

4
 Issues with Solow model
• Neoclassical model
• Allow households to make savings decisions
• Control the division of output into consumption and investment
ct + it = f (kt )
kt+1 = (1 − δ)kt + it
• Economy of optimizing individuals → Good for policy
• Time series of endogenous variables {ct , yt , kt , it }∞
t=0 determined by parameters and
initial conditions.
• Keep other features of Solow → Still match growth facts!

• Work horse model in modern macroeconomics

X + Welfare - Centralized vs. Decentralized economy
X + Labor supply
X + Stochastic processes for At . Real Business Cycle model
X + Heterogeneity. Study inequality.
× + Fiscal policy - ‘Ramsey problem’
× + Monetary policy - ‘New Keynesian models’. ‘Nominal’ Business Cycle model
5
 Environment - Centralized
• Time - Discrete t = 0, 1, 2 . . .
• Agents - Representative household with N identical workers
• Goods - One good can either be used for consumption or investment

Ct + It = Yt
• Preferences - Utility of the household at date 0 is
∞
X
β t U (Ct , Nt ) , β ∈ (0, 1) , U (Ct , Nt ) = u(Ct /Nt )
t=0
• Technology - Constant returns to scale production technology

Yt = AF (Kt , Nt ) = AKtα Nt1−α .

Capital depreciates at rate δ
Kt+1 = (1 − δ)Kt + It , δ ∈ [0, 1] , K0 > 0
6
 Environment - Centralized
• Time - Discrete t = 0, 1, 2 . . .
• Agents - Representative household
• Goods - One good can either be used for consumption or investment
ct + it = yt
• Preferences - Utility of the household at date 0 is
∞
X
β t u(ct ) β ∈ (0, 1)
| {z }
t=0
Rate of time preference

• Technology - Constant returns to scale production technology

yt = Af (kt ) = Aktα .
Capital depreciates at rate δ
kt+1 = (1 − δ)kt + it , δ ∈ [0, 1] , k0 > 0
7
 Environment - Centralized
• Time - Discrete t = 0, 1, 2 . . .
• Agents - Representative household
• Goods - One good can either be used for consumption or investment

ct + it = yt
• Preferences - Utility of the household at date 0 is
∞
X
β t u(ct ) = u(c0 ) + βu(c1 ) + β 2 u(c2 ) · · · + β t u(ct ) + . . .
t=0
• Technology - Constant returns to scale production technology

yt = Af (kt ) = Aktα .
Capital depreciates at rate δ
kt+1 = (1 − δ)kt + it , δ ∈ [0, 1] , k0 > 0
8
 Problem

Household chooses sequences of {ct , kt+1 }∞

t=0

∞
X
max β t u(ct )
t=0

subject to the series of constraints

ct + kt+1 ≤ f (kt ) + (1 − δ)kt , t = 0, 1, 2, . . .

and initial conditions

k0 > 0

9
 Problem - Lagrangean
• Constrained optimization problem
∞
X ∞
X
L= β t u(ct ) + λt [f (kt ) + (1 − δ)kt − ct − kt+1 ]
t=0 t=0

• First order necessary conditions

ct : 0 = β t u0 (ct ) − λt
kt+1 : 0 = −λt + λt+1 f 0 (kt+1 ) + (1 − δ)
 

• Combining conditions
0 = −β t u0 (ct ) + β t+1 u0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)
 

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

10
 Euler equation
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

• Interpretation - Perturbational method

• Suppose we have the optimal {c∗t , kt+1
∗
}∞
t=0
• Reduce consumption by ε units only today, i.e. at date t
• Loss in utility today is approx. u0 (c∗t )ε u(ct ) ≈ u(c∗t ) + u0 (c∗t )(ct − c∗t )
• Plan: Take ε, invest it in capital, consume proceeds tomorrow
• Two effects tomorrow (i) increase output by f 0 (kt+1
∗
)ε, (ii) increase capital by (1 − δ)ε
• Gain in utility tomorrow approx. u0 (c∗t+1 )[f 0 (kt+1
∗
) + (1 − δ)]ε
• If {c∗t , kt+1
∗
} is optimal, then should be no change in total utility:

0 = −u0 (c∗t )ε + βu0 (c∗t+1 ) f 0 (kt+1

∗
 
) + (1 − δ) ε

11
 Partial solution

• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

• Resource constraint

ct + kt+1 = f (kt ) + (1 − δ)kt

• Initial conditions

k0 > 0

12
 Transversality condition

Household chooses sequences of {ct , kt+1 }Tt=0 Truncated at T < ∞

T
X
max β t u(ct )
t=0

subject to the series of constraints

c0 + k1 ≤ f (k0 ) + (1 − δ)k0 , k0 > 0
c1 + k2 ≤ f (k1 ) + (1 − δ)k1
...
cT + kT +1 ≤ f (kT ) + (1 − δ)kT

Issue - Household wants kT +1 infinitely negative

13
 Problem
Household chooses sequences of {ct , kt+1 }Tt=0
T
X
max β t u(ct )
t=0

subject to the series of constraints

ct + kt+1 = f (kt ) + (1 − δ)kt , t = 0, 1, 2, . . . , T

initial conditions

k0 > 0

and a non-negativity constraint on capital

kt+1 ≥ 0

14
 Transversality condition
• Constrained optimization problem
T
X T
X h i XT
t
L= β u(ct ) + λt f (kt ) + (1 − δ)kt − ct − kt+1 + µt kt+1
t=0 t=0 t=0
• First order necessary conditions
ct : 0 = β t u0 (ct ) − λt
kt+1 : 0 = −λt + λt+1 f 0 (kt+1 ) + (1 − δ) +µt
 

kT +1 : 0 = −λT +µT
• Multipliers and constraints
λt ≥ 0 , 0 = f (kt ) + (1 − δ)kt − ct − kt+1
µt ≥ 0 , 0 ≤ kt+1
• Complementary
h slackness conditions i
0 = λt f (kt ) + (1 − δ)kt − ct − kt+1
h i
0 = µt kt+1
15
 Transversality condition
• Optimal cT

λT = β T u0 (cT )

• Optimal kT +1

µT = λT = β T u0 (cT )

• Complementary slackness

β T u0 (cT )kT +1 = 0

• The present discounted utility value of ‘left over’ resources must be equal to zero
• We can generalize this to the infinite horizon case

lim β T u0 (cT )kT +1 = 0

T →∞

16
 Example - A ‘Cake-eating’ problem
• Constrained optimization problem
∞
X
max = β t ct
{ct }∞
t=0 t=0

subject to

Wt+1 = R(Wt − ct ) ∀t , W0 > 0 , Wt+1 ≥ 0

• Suppose βR < 1, then eat the whole cake today. This is fine.
• Suppose βR > 1, then defer consumption every period. This seems weird.
... as t → ∞ we never eat the cake?
• Transversality condition - PDV of future cake has to be zero

lim β T WT +1 = 0
T →∞

17
 Full solution
• Initial conditions

k0 > 0

• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

• Resource constraint

ct + kt+1 = f (kt ) + (1 − δ)kt

• Transversality condition

lim β T u0 (cT )kT +1 = 0

T →∞

18
 Characterizing the solution
1. Steady state

- Solve (c, k) as a function of parameters of the model

- If we change a parameter in the economy, how does the steady state change?

→ Steady state comparative statics

2. Dynamics

- Is the steady state globally stable ?

- What determines the local dynamics around steady state?

- From steady state, if we change a parameter, how does the economy evolve?

→ Dynamic comparative statics

• PS2 - Solution for u(c) = log(c), δ = 1, f (k) = Ak α

19
 1. Steady state
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

• Resource constraint

ct + kt+1 = f (kt ) + (1 − δ)kt

Golden-rule? Now redundant

There are no left over ‘policy variables’ like the savings rate for the planner to choose.

Monetary economics (Prof. Uhlig) - Policy parameters return!

Central bank rule for nominal interest rates: it = φy ybt + φπ πt

20
 1. Steady state
• Euler equation

u0 (c) = βu0 (c) f 0 (k) + (1 − δ)

 

• Resource constraint

c + k = f (k) + (1 − δ)k

21
 1. Steady state
• Euler equation

u0 (c) = βu0 (c) f 0 (k) + (1 − δ)

 

• Resource constraint

c + k = f (k) + (1 − δ)k

Golden-rule? Now redundant

• There are no left over ‘policy variables’ like the savings rate for the planner to choose.

22
 1. Steady state
• Euler equation

u0 (c) = βu0 (c) f 0 (k) + (1 − δ)

 

• Resource constraint

c + k = f (k) + (1 − δ)k

Golden-rule? Now redundant

• There are no left over ‘policy variables’ like the savings rate for the planner to choose.

• Monetary economics (Prof. Uhlig) - Policy parameters return!

• Central bank rule for nominal interest rates: ît = φy ybt + φπ πt

23
 1. Steady state
• Euler equation
1−β ∗ ∗
f 0 (k) = +δ Solow Golden Rule: f 0 (k ) = δ =⇒ k < k
β
• Resource constraint

c = f (k) − δk

Comparative statics

↓ β Value future output relatively less so ↓ k.

Also lower ↓ c.

↑ δ Wasted savings, requires higher ↑ f 0 (k), so ↓ k.

Also lower ↓ c

24
 1. Steady state
• Euler equation
1−β ∗ ∗
f 0 (k) = +δ Solow Golden Rule: f 0 (k ) = δ =⇒ k < k
β
• Resource constraint

c = f (k) − δk

Comparative statics

↓ β Value future output relatively less so ↓ k.

Also lower ↓ c.

↑ δ Wasted savings, requires higher ↑ f 0 (k), so ↓ k.

Also lower ↓ c
Why is c increasing in k around (c, k)?
25
 2. Dynamics - Phase diagram
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

• Resource constraint

ct + kt+1 = f (kt ) + (1 − δ)kt

• No change in consumption: ct+1 = ct

1 = β f 0 (kt ) + (1 − δ)
 
→ kt = k

• No change in capital: kt+1 = kt

ct = f (kt ) − δkt

26
 2. Dynamics - Phase diagram
∆ct = 0 : kt = k
∆kt = 0 : ct = f (kt ) − δkt

27
 2. Dynamics - Phase diagram
u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)
 

ct + kt+1 = f (kt ) + (1 − δ)kt

28
 2. Dynamics - Phase diagram
u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)
 

If kt+1 > k, then low M P Kt+1 , so ↑ ct , ↓ ct+1 so consumption is falling.

29
 2. Dynamics - Phase diagram
ct + kt+1 = f (kt ) + (1 − δ)kt
If ct > c(kt ), then consuming more than f (kt ) − δkt , so capital is falling.

30
 2. Dynamics - Phase diagram
• To the N.W. (-) we violate the resource constraint
• Increasing marginal product of capital, increasing consumption, at some point
ct > f (kt )

31
 2. Dynamics - Phase diagram
• To the S.E. (&) we violate the transversality consition
• Decreasing marginal product of capital, falling consumption, u0 (ct ) → ∞ despite
accumulating capital

32
 2. Dynamics - Saddle path
• On the saddle-path all equilibrium conditions hold
• Economy converges to steady-state

33
 2. Dynamics - Comparative statics
• If initially in steady-state and we permanently change a parameter
(i) What changes the ∆ct = 0 and ∆kt = 0 lines?
(ii) Is there a new saddle path?
(iii) Consumption jumps to new saddle path and economy converges (at a decreasing rate) toward steady-state.
- Should be able to describe this behaviour in terms of the location of the new steady state (why has k
increased? decreased?), and optimal capital accumulation decision of the household along the transition path.

34
 2. Dynamics - Saddle path

Parameters - α = 0.50, δ = 0.20, β = 0.90, f (k) = Ak 0.50 , u(c) = log c

35
 2. Dynamics - Saddle path

Parameters - α = 0.50, δ = 0.20, β = 0.90, f (k) = Ak 0.50 , u(c) = log c

36
 2. Dynamics - Saddle path

Parameters - α = 0.50, δ = 0.20, β = 0.90, f (k) = Ak 0.50 , u(c) = log c

Can we say something about what governs how quickly the economy converges?

37
 2. Dynamics - Local dynamics
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)kt+1

 

• Resource constraint

kt+1 = f (kt ) + (1 − δ)kt − ct

• Local approximation of equilibrium conditions around steady-state
• Euler equation - δ = 1

u00 (c)c βu00 (c)f 0 (k)c βu0 (c)f 00 (k)k

     
ct = ct+1 + kt+1
b
u0 (c) u0 (c) u0 (c)
b b

• Resource constraint
f 0 (k)k
     
1k c
kt+1 =
b kt −
b ct
b
k k k

38
 2. Dynamics - Local dynamics
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)kt+1

 

• Resource constraint

kt+1 = f (kt ) + (1 − δ)kt − ct

• Local approximation of equilibrium conditions around steady-state

• Euler equation
u00 (c)c u00 (c)c f 00 (k)k
     
ct = ct+1 + βf 0 (k)b
kt+1
u0 (c) u0 (c)
b b
f 0 (k)
• Resource constraint
     
k b k b c
kt+1 = kt − ct
b
k βk k

39
 2. Dynamics - Local dynamics
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)kt+1

 

• Resource constraint

kt+1 = f (kt ) + (1 − δ)kt − ct

• Local approximation of equilibrium conditions around steady-state

• Euler equation
u00 (c)c u00 (c)c f 00 (k)k
     
ct = ct+1 + kt+1
b
u0 (c) u0 (c)
b b
f 0 (k)
• Resource constraint
     
k b k b c
kt+1 = kt − ct
b
k βk k

40
 2. Dynamics - Local dynamics
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

• Resource constraint

kt+1 = f (kt ) + (1 − δ)kt − ct

• Local approximation of equilibrium conditions around steady-state

• Euler equation - u0 (c) ∝ c−σ , f 0 (k) ∝ k α−1

−σb
ct = −σb
ct+1 + (α − 1)b
kt+1

• Resource constraint
1 c
kt+1 = b
b kt − bct
β k
41
 2. Dynamics - Local dynamics
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

• Resource constraint

kt+1 = f (kt ) + (1 − δ)kt − ct

• Local approximation of equilibrium conditions around steady-state

• Euler equation - u0 (c) ∝ c−σ , f 0 (k) ∝ k α−1
 
1−α b
ct+1 − b
b ct = − kt+1
σ

1. ↑ σ more curvature in utility function

• Smoother consumption → Smaller rate of change in b kt
ct for any b
42
 2. Dynamics - Local dynamics
• Euler equation

u0 (ct ) = βu0 (ct+1 ) f 0 (kt+1 ) + (1 − δ)

 

• Resource constraint

kt+1 = f (kt ) + (1 − δ)kt − ct

• Local approximation of equilibrium conditions around steady-state

• Euler equation - u0 (c) ∝ c−σ , f 0 (k) ∝ k α−1
 
1−α b
ct+1 − b
b ct = − kt+1
σ
2. ↓ α more curvature in production
• Marginal product of capital more responsive to b
kt
• Consumption changes more quickly in response to b 6 0
kt =
43
 2. Dynamics - Local dynamics
• Rearrange these (use R.C. to sub b kt+1 in E.E.)
       
ct+1
b ct
b ct
b c0
t b
=A b which implies that =A b
kt+1
b kt kt
b k0
• Eigen-vector decomposition Av1 = λ1 v1 → AV = V Λ
     
c2
b −1 −1 k0
b 2 −1 bct
=V ΛV V ΛV =VΛ V
k2
b | {z } | {z } c0
b kt
b
A A
• In the limit as t → ∞
   t   
ct
b λ1 0 c0
−1 b
lim b = lim V t V
t→∞ kt t→∞ 0 λ2 k0
b

• Macroeconomics models will deliver 0 < λ1 < 1 < λ2

- Why? Discounting β < 1, Concavity u00 < 0, f 00 < 0
- See: PS2 Q5(c)

44
 2. Dynamics - Local dynamics
• Macro problems will deliver 0 < λ1 < 1 < λ2
• Now we can write an expression for kt and ct , using At = V Λt V −1
• Recall the simple rule for the inverse of a 2 × 2 matrix!
v11 v21 λt1 0
      
ct
b 1 v22 −v21 b c0
= t
kt
b v
det(V ) 12 v 22 0 λ 2 −v 12 v 11 k0
b

• If −v12 b k0 6= 0, then gets multiplied by the explosive eigen value λ2

c0 + v11 b
• Will continue to propogate, sending b kt and bct to either infinity or zero!
• Both cases violate equilibrium conditions!
• The saddle path consists of a value of b c0 such that given b
k0 , these conditions are not violated.
This explosive eigen value is killed off.
−v12 b c0 = 0 → b
k0 + v11 b c0 = (v11 /v12 ) × b
k0
• PS2 Q5 - Compute v11 , v12 by hand for δ = 1, u(c) = log c, f (k) = k α
• Bonus - Show b
kt = λt b
k0 . So λ1 explicitly governs rate of convergence(!!!)
1

45
 2. Dynamics - Local dynamics
• Canvas - Some simple code NCMEigensolve.m that produces these figures. Play
around with the code!
• What causes faster convergence: (↓ α, ↓ σ)? Does the RHS plot diverge in line with
c0 6= γ b
theory for b k0 ? What happens when b
k0 ≶ 0? Write some code to plot λ1 as a
function of β, σ, α, what do you learn? Have fun!

46
 3. Dynamics - Following a shock

• How does the economy respond following a shock?

• What if that shock is permanent or transitory?

• What if it is anticipated or unexpected?

• How might this behavior help us understand what generates business cycles?

• Recall: Yt , Ct and It all drop during a recession, increase during a boom

• (i) Change in k0 , (ii) Changes in β, (iii) Changes in A

47
 3. Dynamics - ↓ k0 - Unexpected
↓ c0 to reaccumulate capital stock. If ↑ σ or ↑ α, then smaller drop and slower rebuild.

48
 3. Dynamics - Changes in patience β

- Resource constraint

ct + kt+1 = Af (kt ) + (1 − δ)kt + g

- Euler equation

u0 (ct ) = βt u0 (ct+1 ) Af 0 (kt+1 ) + (1 − δ)kt+1

 

- ∆ct = 0 locus shifts

- ∆kt = 0 locus is unaffected

49
 3. Dynamics - ↑ β - Unexpected - Permanent
0
↓ c0 to accumulate capital to reach higher steady-state k and c0 .

50
 3. Dynamics - ↑ β - Expected - Permanent
↓ c0 and accumulate capital to smooth effect on consumption.

51
 3. Dynamics - Changes in productivity A

- Resource constraint

ct + kt+1 = At f (kt ) + (1 − δ)kt + g

- Euler equation

u0 (ct ) = βu0 (ct+1 ) At f 0 (kt+1 ) + (1 − δ)kt+1

 

- ∆ct = 0 locus shifts

- ∆kt = 0 locus shifts

52
 3. Dynamics - ↓ A - Unexpected - Permanent
0
↓ c0 and use up capital to smooth transition to lower c0 , k .

53
 3. Dynamics - ↓ A - Expected - Permanent
↓ c0 and accumulate capital to smooth effect on consumption.

54
 3. Dynamics - ↓ A - Unexpected - Transitory
0
↓ c0 and use up capital to smooth transition to lower c0 , k .

55
 3. Dynamics - ↓ A - Expected - Transitory
↓ c0 and accumulate capital to smooth effect on consumption.

56