Macroeconomics 1b: Fall 2024.

Homework 2
Professor: Michal Kejak. Teaching assistant: Jose Gomez Castro

Student: Kai Angel-Augusto Sanchez Pajuelo

CERGE-EI, 1st year PhD studies

2024/2025

1. Negishi Algorithm: Problem Statement

The social planner seeks to maximize the weighted sum of utilities of two agents over time, subject to a
feasibility constraint. The problem can be written as:

∞ X
X
max β t πt (st |s0 ) [λu(c1,t (st )) + (1 − λ)u(c2,t (st ))]
{c1,t (st ),c2,t (st )}
t=0 st

subject to the feasibility constraint:

c1,t (st ) + c2,t (st ) = et (st ), ∀t, st ,

where:

• c1,t (st ) is the consumption of agent 1 at time t and state st ,

• c2,t (st ) is the consumption of agent 2 at time t and state st ,
• β is the discount factor (0 < β < 1),

• πt (st |s0 ) is the probability of state st given the initial state s0 ,

• λ is the Pareto weight for agent 1,
• 1 − λ is the Pareto weight for agent 2,
• u(c) is the utility function for both agents,

• et (st ) = e1,t (st ) + e2,t (st ) is the total endowment available at t, st .

The utility function is assumed to be CRRA (Constant Relative Risk Aversion):

c1−γ
u(c) = , γ > 0.
1−γ

1
A. First Order Conditions, Allocations, and Shadow Prices

The Lagrangian for this problem is:

∞ X
X h i X∞ X h i
L= β t πt (st |s0 ) λu(c1,t (st )) + (1 − λ)u(c2,t (st )) + µt (st ) et (st ) − c1,t (st ) − c2,t (st ) .
t=0 st t=0 st

Expanding:

∞ X
X h i
L= β t πt (st |s0 ) λu(c1,t (st )) + (1 − λ)u(c2,t (st )) − µt (st )c1,t (st ) − µt (st )c2,t (st ) + µt (st )et (st ) .
t=0 st

1. FOC for c1,t (st )

Differentiating L with respect to c1,t (st ), we get:

∂L h i
= β t πt (st |s0 ) λu′ (c1,t (st )) − µt (st ) = 0.
∂c1,t (st )

Rearranging:

λu′ (c1,t (st )) = µt (st ).

Substituting u′ (c) = c−γ , we have:

µt (st ) = λc1,t (st )−γ .

2. FOC for c2,t (st )

Similarly, differentiating L with respect to c2,t (st ), we get:

∂L h i
= β t πt (st |s0 ) (1 − λ)u′ (c2,t (st )) − µt (st ) = 0.
∂c2,t (st )

Rearranging:

(1 − λ)u′ (c2,t (st )) = µt (st ).

Substituting u′ (c) = c−γ , we have:

µt (st ) = (1 − λ)c2,t (st )−γ .

2
3. Combining the FOCs

From the two FOCs, we have:

λc1,t (st )−γ = (1 − λ)c2,t (st )−γ .

Rearranging to find the ratio of c1,t (st ) to c2,t (st ):

1/γ
c1,t (st )

λ
= .
c2,t (st ) 1−λ
 1/γ
Define α = λ
1−λ . Then:

c1,t (st ) = αc2,t (st ).

4. Using the Feasibility Constraint

From the feasibility constraint:

c1,t (st ) + c2,t (st ) = et (st ).

Substitute c1,t (st ) = αc2,t (st ):

αc2,t (st ) + c2,t (st ) = et (st ).

Factor out c2,t (st ):

c2,t (st )(1 + α) = et (st ).

Solve for c2,t (st ):

et (st )
c2,t (st ) = .
1+α

Substitute back to find c1,t (st ):

αet (st )
c1,t (st ) = αc2,t (st ) = .
1+α

5. Shadow Prices and Competitive Prices

From the FOCs, the shadow prices are:

µt (st ) = λc1,t (st )−γ .

Competitive equilibrium prices for the (t, st )-goods are:

q0,t (st ) = β t πt (st |s0 )u′ (c1,t (st )).

3
Substituting u′ (c1,t (st )) = c1,t (st )−γ , we have:

q0,t (st ) = β t πt (st |s0 )µt (st )/λ.

Thus, the shadow prices µt (st ) align with the competitive equilibrium prices q0,t (st ).

6. Allocations with Cyclical Endowments

The allocations become state-dependent:

1. If t is even:
α 1
c1,t (st ) = (e1,h + e2,l ), c2,t (st ) = (e1,h + e2,l ).
1+α 1+α
2. If t is odd:
α 1
c1,t (st ) = (e1,l + e2,h ), c2,t (st ) = (e1,l + e2,h ).
1+α 1+α

B. Deriving the Price System q0,t (st )

We now derive the competitive equilibrium price system q0,t (st ).

1. Substituting the Marginal Utility

The price system is related to the marginal utility of consumption for Agent 1. From FOCs, we have:

q0,t (st ) = β t πt (st |s0 )u′ (c1,t (st )).

Substitute u′ (c) = c−γ :

q0,t (st ) = β t πt (st |s0 )c1,t (st )−γ .

2. Using the Allocations for c1,t (st )

From the feasibility condition and FOCs, we know:

 1/γ
α λ
c1,t (st ) = et (st ), where α = .
1+α 1−λ

Substitute this expression into the price equation:

 −γ
α
q0,t (st ) = β t πt (st |s0 ) et (st ) .
1+α

3. Simplifying the Price System

Simplify the price equation: γ

1+α

q0,t (st ) = β πt (st |s0 )
t
et (st )−γ .
α

Substitute the cyclical pattern for et (st ):

4
 1. If t is even: γ
1+α

q0,t (st ) = β πt (st |s0 )
t
(e1,h + e2,l )−γ .
α

2. If t is odd: γ
1+α

q0,t (st ) = β πt (st |s0 )
t
(e1,l + e2,h )−γ .
α

C. Solving for λ∗

We solve for the Pareto weight λ∗ that ensures the budget constraint for Agent 1 is satisfied:

∞
X ∞
X
β t πt (st |s0 )q0,t (st )c1,t (st ) = β t πt (st |s0 )q0,t (st )e1,t (st ).
t=0 t=0

Substitute the expressions for q0,t (st ) and c1,t (st ):

∞ γ ∞ γ
1+α 1+α
  
X α X
β πt (st |s0 )
t
et (st )1−γ
= β πt (st |s0 )
t
et (st )−γ e1,t (st ).
t=0
1+α α t=0
α

Splitting into Even and Odd Periods

1. For even t:
et (st ) = e1,h + e2,l , e1,t (st ) = e1,h .

2. For odd t:
et (st ) = e1,l + e2,h , e1,t (st ) = e1,l .

The equilibrium condition is split into even and odd sums:

X X
Even: β t . . . , Odd: βt . . . .
even t odd t

Proving Agent 2’s Budget Constraint Holds Automatically

From the feasibility condition, we know:

c1,t (st ) + c2,t (st ) = et (st ), ∀t, st ,

which implies that the total consumption equals the total endowment.
Using this, the total value of the endowments can be written as:
∞ X
X ∞ X
X
q0,t (st )et (st ) = q0,t (st ) c1,t (st ) + c2,t (st ) .


t=0 st t=0 st

This equation separates into the budget constraints of the two agents:
∞ X
X ∞ X
X ∞ X
X ∞ X
X
q0,t (st )c1,t (st ) + q0,t (st )c2,t (st ) = q0,t (st )e1,t (st ) + q0,t (st )e2,t (st ).
t=0 st t=0 st t=0 st t=0 st

5
If the budget constraint for Agent 1 is satisfied:
∞ X
X ∞ X
X
q0,t (st )c1,t (st ) = q0,t (st )e1,t (st ),
t=0 st t=0 st

then the budget constraint for Agent 2 must hold automatically:

∞ X
X ∞ X
X
q0,t (st )c2,t (st ) = q0,t (st )e2,t (st ).
t=0 st t=0 st

Thus, solving for λ∗ using Agent 1’s budget constraint ensures that Agent 2’s budget constraint is also
satisfied.

D. Alternative Price System

The problem asks whether the alternative price system:
q0,t (st ) = β t et (st )−γ ,
can be used in place of the price system derived in 1.b), and whether it would produce the same results as
in 1.c).

1. Comparing the Price Systems

The Price System from 1.b): The price system derived in 1.b) is:
γ
1+α

q0,t (st ) = β πt (st |s0 )
t
et (st )−γ ,
α
where:
 1/γ
• α= λ
1−λ ,

• πt (st |s0 ) accounts for the probabilities of states.

2. Can the Alternative Price System Be Used? No, the alternative price system cannot be used
because:

1. No Dependence on λ:
• The alternative system ignores the Pareto weight λ, which determines the planner’s trade-off
between agents’ utilities.
• As a result, it cannot adjust prices to balance both agents’ budget constraints simultaneously.
2. Budget Constraint Inconsistency:
• In the correct price system, individual consumption prices are proportional to marginal utilities,
reflecting the trade-off in consumption between agents.
• The alternative system treats the total endowment as the determinant of prices, which does not
ensure individual budget constraints hold.
3. No State Probabilities (πt (st |s0 )):
• The alternative system ignores state probabilities, which are crucial for state-contingent pricing
in competitive equilibrium.
• This omission breaks the alignment between the planner’s solution and the equilibrium.

6
3. Would It Produce the Same Results as in 1.c)? No, the alternative price system would not
produce the same results as in 1.c):

• Failure to Reflect λ∗ :
The equilibrium value λ∗ derived in 1.c) relies on the correct price system, which incorporates α =
 1/γ
λ
1−λ . Without α, the alternative system cannot recover the same equilibrium.

• Budget Constraint Violations:

The alternative system is based only on et (st ) and does not distinguish between agents. As a result, it
cannot guarantee that both agents’ budget constraints are satisfied.
• Allocations Would Differ:
Allocations derived using the alternative system would not align with the planner’s solution or the
competitive equilibrium, as the prices fail to reflect individual marginal utilities.

4. Why the Price System in 1.b) is Correct The price system in 1.b) ensures:

1. Marginal Rate of Substitution:

u′ (c1,t (st )) λ
= ,
u (c2,t (st ))
′ 1−λ
aligning with the planner’s trade-off between agents.

2. Competitive Equilibrium Conditions: Prices reflect shadow prices from the planner’s solution,
ensuring feasibility and budget constraint satisfaction.

Note on Exceptions for the Alternative Price System While the alternative price system:

q0,t (st ) = β t et (st )−γ ,

cannot generally replace the price system derived in 1.b), there are a few specific exceptions where it
could work:

1. Homogeneous Agents:
If both agents have identical preferences (λ = 0.5, γ) and symmetric endowments (e1,t (st ) = e2,t (st ) =
et (st )
2 ), the planner’s trade-off becomes irrelevant. In such cases, the system alternative approximates
the correct marginal prices.
2. Representative Agent Model:
In a model with a single representative agent, the planner’s and the market’s optimization problems
are equivalent, and the marginal utility of the total endowment (u′ (et (st ))) determines prices directly:

q0,t (st ) = β t et (st )−γ .

3. Constant Endowments and Probabilities:

If the total endowment et (st ) and the state probabilities πt (st |s0 ) are constant across time and states,
the prices become proportional to et (st )−γ , and the alternative price system becomes a reasonable
approximation.

These exceptions, however, are highly restrictive and rarely apply in more general macroeconomic models,
where heterogeneity in agents and state-dependent pricing are critical.

7
E. Compute λ∗ , Allocations, and Compare Utilities

1. Given Parameters

• β = 0.9,

• γ = 2,
• Endowments:
e1,h = 1.5, e1,l = 1, e2,h = 2, e2,l = 0.5.

The total endowments alternate cyclically:

(
e1,h + e2,l = 1.5 + 0.5 = 2, if t is even,
et =
e1,l + e2,h = 1 + 2 = 3, if t is odd.

Agent 1’s endowment alternates as:

(
e1,h = 1.5, if t is even,
e1,t =
e1,l = 1, if t is odd.

2. Budget Constraint for Agent 1

Agent 1’s budget constraint is:

∞
X ∞
X
β t q0,t (st )c1,t (st ) = β t q0,t (st )e1,t (st ),
t=0 t=0

where:

 1/γ
• Allocations: c1,t (st ) = 1+α et (st ),
α
with α = λ
1−λ ,

1+α γ
• Prices: q0,t (st ) = β t et (st )−γ .


α

Substituting these into the budget constraint:

∞ γ ∞ γ
1+α 1+α
  
X α X
βt et (st )1−γ = βt et (st )−γ e1,t (st ).
t=0
1+α α t=0
α

3. Separate into Even and Odd Periods

The cyclical nature of the endowments allows us to separate the sum into even and odd periods.
Even t:

• Total endowment: et = 2,
• Agent 1’s endowment: e1,t = 1.5.

Odd t:

8
 • Total endowment: et = 3,
• Agent 1’s endowment: e1,t = 1.

The budget constraint becomes:

β2 β 1.5 β2 1
   
β α α
2 1−γ
+ 31−γ
= + .
1 − β2 1+α 1 − β2 1+α 1 − β 2 2γ 1 − β 2 3γ
| {z } | {z }
Even terms Odd terms

4. Substitute Parameters

With γ = 2:

• 21−γ = 2−1 = 0.5, 31−γ = 3−1 = 13 ,

• 2γ = 4, 3γ = 9,
• β = 0.9.

Substitute into the equation:

1 1.5 1
  
α
0.5 + 0.9 · = + 0.9 · .
1+α 3 4 9

Simplify coefficients:
1
0.5 + 0.9 · = 0.5 + 0.3 = 0.8,
3
1.5 1
+ 0.9 · = 0.375 + 0.1 = 0.475.
4 9
Thus:  
α
· 0.8 = 0.475.
1+α

Solve for 1+α :

α

α 0.475
= = 0.59375.
1+α 0.8

5. Solve for α

Using α
1+α = 0.59375, solve for α:

0.59375 0.59375
α= = ≈ 1.4625.
1 − 0.59375 0.40625

9
6. Compute λ∗
 1/γ
From α = λ
1−λ , and γ = 2:
λ
α2 = .
1−λ

Substitute α = 1.4625:
λ
1.46252 = .
1−λ

Solve for λ:
1.46252 2.14 2.14
λ= = = ≈ 0.681.
1 + 1.46252 1 + 2.14 3.14

Thus, λ∗ ≈ 0.681.

7. Compute Allocations

For t even (et = 2):

α 1.4625
c1,t (st ) = ·2= · 2 ≈ 1.19, c2,t (st ) = 2 − 1.19 ≈ 0.81.
1+α 1 + 1.4625

For t odd (et = 3):

α 1.4625
c1,t (st ) = ·3= · 3 ≈ 1.79, c2,t (st ) = 3 − 1.79 ≈ 1.21.
1+α 1 + 1.4625

8. Compare Utilities

Utility for each agent:

∞
X ∞
X
U1 = β t u(c1,t (st )), U2 = β t u(c2,t (st )).
t=0 t=0

With λ∗ = 0.681, Agent 1 has higher weight in the planner’s problem and thus enjoys higher utility in
equilibrium.

Final Results:

1. Pareto Weight:
λ∗ ≈ 0.681.

2. Allocations:

• t even: c1,t ≈ 1.19, c2,t ≈ 0.81,

• t odd: c1,t ≈ 1.79, c2,t ≈ 1.21.

3. Higher Utility: Agent 1 enjoys higher utility due to their higher Pareto weight (λ∗ ).

10
2. Static Trade: Problem Statement
We consider a static two-country economy with two goods: apples a and bananas b. Each country has the
same constant elasticity utility function:
1/ρ
U (a, b) = [(1 − ω)aρ + ωbρ ] ,

where:

• 0 < ω < 1,
• ρ < 1, and
• The elasticity of substitution is σ = 1−ρ .
1

Country 1 is endowed with y1 apples, and Country 2 is endowed with y2 bananas. We solve for:

1. A Pareto optimal allocation and the implicit relative price q,

2. The competitive equilibrium associated with initial endowments,

3. A method to calibrate ω using an observed import share.

A. Solve for a Pareto Optimal Allocation

1. Social Planner’s Problem

The planner maximizes a weighted sum of utilities:

max λU1 (a1 , b1 ) + (1 − λ)U2 (a2 , b2 ),

a1 ,b1 ,a2 ,b2

subject to:
a1 + a2 = y1 , b1 + b2 = y2 .

2. Lagrangian

The Lagrangian is:

1/ρ 1/ρ
L = λ [(1 − ω)aρ1 + ωbρ1 ] + (1 − λ) [(1 − ω)aρ2 + ωbρ2 ] + µ1 (y1 − a1 − a2 ) + µ2 (y2 − b1 − b2 ).

3. First Order Conditions (FOCs)

For a1 :
∂L (1 − ω)a1ρ−1
=λ 1− 1
− µ1 = 0.
∂a1 [(1 − ω)aρ1 + ωbρ1 ] ρ
Thus:
(1 − ω)aρ−1
µ1 = λ 1
1
1− ρ
.
[(1 − ω)aρ1 + ωbρ1 ]

11
For a2 :
∂L (1 − ω)aρ−1
= (1 − λ) 2
1− 1
− µ1 = 0.
∂a2 [(1 − ω)aρ + ωbρ ] ρ
2 2
Thus:
(1 − ω)a2ρ−1
µ1 = (1 − λ) 1
1− ρ
.
[(1 − ω)aρ2 + ωbρ2 ]

Equating the two expressions for µ1 :

(1 − ω)aρ−1 (1 − ω)a2ρ−1
λ 1
1
1− ρ
= (1 − λ) 1
1− ρ
.
[(1 − ω)aρ1 + ωbρ1 ] [(1 − ω)aρ2 + ωbρ2 ]

For b1 :
∂L ωb1ρ−1
=λ 1− 1
− µ2 = 0.
∂b1 [(1 − ω)aρ1 + ωbρ1 ] ρ
Thus:
ωbρ−1
µ2 = λ 1
1
1− ρ
.
[(1 − ω)aρ1 + ωbρ1 ]

For b2 :
∂L ωbρ−1
= (1 − λ) 2
1− 1
− µ2 = 0.
∂b2 ρ
[(1 − ω)a + ωbρ ] ρ
2 2
Thus:
ωb2ρ−1
µ2 = (1 − λ) 1
1− ρ
.
[(1 − ω)aρ2 + ωbρ2 ]

Equating the two expressions for µ2 :

ωb1ρ−1 ωbρ−1
λ 1
1− ρ
= (1 − λ) 2
1
1− ρ
.
[(1 − ω)aρ1 + ωbρ1 ] [(1 − ω)aρ2 + ωbρ2 ]

4. Relative Price of Apples to Bananas

The relative price q = µ1

µ2 is:
(1 − ω)a1ρ−1
q= .
ωbρ−1
1

Simplify:
1 − ω aρ−1
q= 1
· ρ−1 .
ω b1

B. Competitive Equilibrium

1. Market Clearing Conditions

In equilibrium:
a1 + a2 = y1 , b1 + b2 = y2 .

12
2. Allocations

From the Pareto optimality:

a1 λ b1 λ
= , = .
a2 1−λ b2 1−λ
Substitute market clearing conditions:
λ
a1 = y1 = λy1 , a2 = (1 − λ)y1 ,
λ + (1 − λ)
b1 = λy2 , b2 = (1 − λ)y2 .

Thus, the allocations are:

a1 = λy1 , a2 = (1 − λ)y1 ,
b1 = λy2 , b2 = (1 − λ)y2 .

C. Calibration of ω Using Import Share

1. Import Share Formula

The import share for Country 1 is defined as:

value of imports at market prices
Import Share = .
total value of consumption at market prices

Country 1 imports apples (a2 ) from Country 2, and the import share becomes:
q · a2
Import Share = ,
q · a2 + a1

2. Relative Price of Apples

From the solution to part (a), the relative price of apples to bananas is given by:
 ρ−1
1−ω a1
q= · .
ω b1

Substitute a1 = λy1 and b1 = λy2 from part (b):

ρ−1  ρ−1
1−ω 1−ω

λy1 y1
q= · = · .
ω λy2 ω y2

3. Substituting q into the Import Share Formula

The import share becomes:

 y1 ρ−1
1−ω
ω· y2 · (1 − λ)y1
Import Share =
y1 ρ−1
  .
1−ω
ω · y2 · (1 − λ)y 1 + λy 1

Factorize y1 in the numerator and denominator:

 ρ−1
· (1 − λ) · yy21
1−ω


ω
Import Share =  ρ−1 .
1−ω
(1 y1
+


ω · − λ) · y2 λ

13
4. Rearrange to Solve for ω

Rearranging the equation for ω, we have:

"  ρ−1 #   ρ−1
1−ω 1−ω
 
y1 y1
Import Share · · (1 − λ) · +λ = · (1 − λ) · .
ω y2 ω y2

Expanding:
ρ−1 ρ−1
1−ω 1−ω
     
y1 y1
Import Share · · (1 − λ) · + Import Share · λ = · (1 − λ) · .
ω y2 ω y2

Group terms involving ω :

1−ω

1−ω
 
[Import Share · (1 − λ) − (1 − λ)] · = −Import Share · λ.
ω

D. Impact of Different Preferences

We are asked to comment on how the solution changes if the countries have different preferences, such as
differing values of ω, the weight assigned to bananas in the utility function.

1. The Current Assumption of Homogeneous Preferences

In the previous parts, we assumed both countries shared the same utility function:
1/ρ
U (a, b) = [(1 − ω)aρ + ωbρ ] ,

where ω is the common weight given to bananas. This assumption led to symmetric expressions for allocations
and prices, such as:
a1 λ 1 − ω a1ρ−1
= , q= · ρ−1 .
a2 1−λ ω b1

2. Heterogeneous Preferences: ω1 ̸= ω2

Now, suppose the countries have heterogeneous preferences, represented by:

• For Country 1: ω1 , the weight on bananas,

• For Country 2: ω2 , the weight on bananas.

The utility functions become:

1/ρ
U1 (a1 , b1 ) = [(1 − ω1 )aρ1 + ω1 bρ1 ] ,
1/ρ
U2 (a2 , b2 ) = [(1 − ω2 )aρ2 + ω2 bρ2 ] .

14
2.1 Effect on the Pareto Optimal Allocations The planner now maximizes:

max λU1 (a1 , b1 ) + (1 − λ)U2 (a2 , b2 ),

a1 ,b1 ,a2 ,b2

subject to:
a1 + a2 = y1 , b1 + b2 = y2 .

The heterogeneity in ω1 and ω2 introduces asymmetries in the marginal rates of substitution (MRS)
between apples and bananas:

(1 − ω1 )aρ−1 (1 − ω2 )aρ−1
MRS1 = 1
, MRS2 = 2
.
ω1 b1ρ−1 ω2 b2ρ−1

The Pareto optimality condition:

∂U1 ∂U2
λ = (1 − λ) ,
∂a1 ∂a2
and a similar condition for b1 and b2 , implies:
a1 b1
̸= .
a2 b2
Thus, the allocations of apples and bananas will reflect the asymmetry in how each country values the goods.

3. Impact on Relative Prices

The relative price of apples to bananas q = µ2 ,

µ1
previously symmetric, now becomes:

(1 − ω1 )a1ρ−1 (1 − ω2 )aρ−1
q= =
̸ 2
.
ω1 bρ−1
1 ω b
2 2
ρ−1

This result highlights two key changes:

1. Country-Specific Prices: If ω1 ̸= ω2 , the willingness of each country to trade apples for bananas
differs. As a result, the terms of trade (q) depend on which country dominates the trade.
2. Distorted Equilibrium Prices: Unlike the case of homogeneous preferences, the equilibrium price
reflects the weighted preferences of both countries, but the weights are no longer symmetric.

4. Economic Interpretation

4.1 Asymmetric Allocations If ω1 > ω2 :

• Country 1 values bananas more than Country 2.

• In the Pareto optimal allocation, Country 1 will consume relatively more bananas (b1 ) and fewer apples
(a1 ) compared to Country 2.

Conversely, if ω1 < ω2 :

• Country 1 values bananas less than Country 2.

• In this case, Country 1 consumes relatively fewer bananas (b1 ) and more apples (a1 ) compared to
Country 2.

15
4.2 Terms of Trade The country that values a good more (e.g., bananas for Country 1 if ω1 > ω2 ) will
be willing to pay a higher relative price for it. As a result:

• The terms of trade (relative price q) will shift in favor of the country that values bananas less, as it
has more bargaining power in trade.

E. Impact of Nonhomothetic Preferences

We analyze how the introduction of Stone-Geary utility, which incorporates nonhomothetic prefer-
ences, would affect the model. The Stone-Geary utility function is given by:
1/ρ
U (a, b) = (1 − ω)(a + ā)ρ + ω(b + b̄)ρ

,
where:

• ā < 0 and b̄ < 0 represent subsistence levels of consumption for apples (a) and bananas (b),
respectively.
• (1 − ω) and ω are weights on apples and bananas in the utility function.
• ρ < 1 determines the elasticity of substitution, with σ = 1/(1 − ρ).

This utility function introduces nonhomotheticity into the model, meaning that consumption patterns
change with income levels. The implications of this are explored below.

1. Features of Stone-Geary Utility

1.1 Nonhomothetic Preferences Unlike the homothetic preferences used in standard models, Stone-
Geary preferences capture the idea that:

• At low income levels, consumption is driven by the need to meet subsistence requirements (ā, b̄).
• As income increases beyond subsistence, consumption shifts toward luxury goods with lower subsis-
tence needs.

1.2 Income-Sensitive Marginal Utilities The marginal utility of each good now depends on how close
consumption is to its subsistence level:
∂U  1 −1
= (1 − ω)(a + ā)ρ−1 · (1 − ω)(a + ā)ρ + ω(b + b̄)ρ ρ .

∂a
Similarly for bananas:
∂U  1 −1
= ω(b + b̄)ρ−1 · (1 − ω)(a + ā)ρ + ω(b + b̄)ρ ρ .

∂b
The Stone-Geary formulation implies that marginal utilities increase rapidly as consumption approaches the
subsistence level, reflecting the priority of survival over luxury consumption at low income levels.

2. Implications for the Model

2.1 Pareto Optimal Allocations The Pareto problem maximizes:

1/ρ 1/ρ
max λ (1 − ω)(a1 + ā)ρ + ω(b1 + b̄)ρ + (1 − λ) (1 − ω)(a2 + ā)ρ + ω(b2 + b̄)ρ
 
,
a1 ,b1 ,a2 ,b2

subject to the feasibility constraints:

a1 + a2 = y1 , b1 + b2 = y2 .

16
Marginal Rate of Substitution The marginal rate of substitution (MRS) between apples and bananas
for Country 1 is:
∂U1
∂a1 (1 − ω)(a1 + ā)ρ−1
MRS1 = = .
∂U1
∂b1 ω(b1 + b̄)ρ−1

Similarly for Country 2:

(1 − ω)(a2 + ā)ρ−1
MRS2 = .
ω(b2 + b̄)ρ−1

Relative Price The relative price of apples to bananas (q) reflects the MRS and becomes:
∂U
(1 − ω)(a + ā)ρ−1
q= ∂a
= .
∂U
∂b ω(b + b̄)ρ−1

Key Implication:

• The relative price q now depends on subsistence-adjusted consumption levels (a + ā, b + b̄), making it
nonlinear and income-sensitive.
• Subsistence constraints distort the optimal allocation compared to the homothetic case, especially at
lower income levels.

2.2 Competitive Equilibrium The competitive equilibrium conditions remain:

a1 + a2 = y1 , b1 + b2 = y2 ,
but the demand functions become nonlinear due to the subsistence terms. For Country 1, demand for apples
(a1 ) is:
  1
λ(1 − ω)µ 1−ρ
a1 = − ā.
pa

Similarly, demand for bananas is:

 1
 1−ρ
λωµ
b1 = − b̄.
pb

These expressions highlight that:

1. At low income levels, consumption is dominated by subsistence needs (−ā, −b̄).

2. As income rises, demand for each good depends on its elasticity of substitution (σ = 1/(1 − ρ)).

3. Structural Transformation and Growth

3.1 Insights from Herrendorf et al. (2013) Herrendorf, Rogerson, and Valentinyi show that nonho-
mothetic preferences explain the structural transformation observed in industrialized economies:

• Goods with high subsistence levels (e.g., food) exhibit declining expenditure shares as income rises.
• Resources shift from agriculture to manufacturing and services due to changes in consumption patterns.

Stone-Geary utility captures this shift by linking consumption shares to income levels:
pi · ci
Expenditure Share of Good i = .
Total Income
For goods near their subsistence levels, expenditure shares decline with income growth.

17
3.2 Generalized Balanced Growth (GBG) from Kongsamut et al. (2001) Kongsamut, Rebelo,
and Xie extend this analysis by showing that nonhomothetic preferences generate a generalized balanced
growth path:

1. Aggregate growth remains consistent with the Kaldor facts.

2. Sectoral shares evolve dynamically, reflecting reallocation of resources toward goods with lower subsis-
tence requirements.

4. Policy Implications

1. Calibration of Subsistence Levels:

• Subsistence levels (ā, b̄) must be estimated from consumption data, particularly for low-income
economies.
• These parameters are critical for understanding poverty traps and structural transformation.

2. Welfare Analysis:

• Policies that reduce subsistence constraints (e.g., through income transfers or subsidies) can ac-
celerate shifts toward higher-value goods.

18