L ECTURE 0: F OUNDATIONS

Heterogeneous Agents in Macroeconomics

Rustam Jamilov

University of Oxford
Fall, 2023
 G ORMAN (1961)

Finite set N with cardinality N of agents, indexed by i. Denote yi (p, ωi ) as preference for a
homogenous good of agent i, given price p ∈ R1 , and wealth ωi . Then, aggregate demand:
N
X N
X
Y(p, ω1 , . . . , ωN) = yi (p, ωi ) ≡ Y(p, ωi )
i i
| {z }
?

1/15
 G ORMAN (1961)

Aggregate-wealth-preserving re-distribution. Total differentiation of Y:

N
d Y(p, N
P
i ωi )
X ∂ yi (p, ωi )
=0→ d ωi = 0
d ωi ∂ ωi
i

True iff:
∂ yi (p, ωi ) ∂ yj (p, ωj )
= ∀i, j ∈ N
ωi ωj
Strong static aggregation if linear Engel curves = MPC homogeneity.

Argument in Gorman (1961) is through indirect utility functions.

2/15
 R UBINSTEIN (1974)

Agent maxes consumption ct given HARA utility, discount factor β, wealth ωt , risky asset
f
allocation αt , risky return Rat , risk-free return Rt , and linear absolute risk aversion
′′ ′ 1
RA(c) ≡ −U(c) /U(c) = A+Bc :

T
!
X
max E β t U(ct )
{ct ,αt }
t
 
f
s.t. ωt+1 = (ωt − ct ) αt Rt + (1 − αt )Rat

3/15
 R UBINSTEIN (1974)
T
!
X
t
max E β U(ct )
{ct ,αt }
t
 
f
s.t. ωt+1 = (ωt − ct ) αt Rt + (1 − αt )Rat

Strong dynamic aggregation if:

1. Homogeneous curvature B (necessity).
2. Homogeneous ω0 , β, and U (sufficiency).
3. Homogeneous β and B ̸= 0 (sufficiency).
4. Complete markets & B = 0 (sufficiency).
5. Complete markets & ω0j = ω0 & B = 1 & A = 0 (sufficiency).
Need condition (1) plus any of (2)-(5).
4/15
 C ONSTANTINIDES (1982)

Debreu (1959) model: m consumers indexed by i, n firms indexed P by j, l goods indexed by h,

wealth endowments ωi ≡ (ωi,1 , . . . , ωil ), shares in firms sin with m sij = 1, consumption
xi ≡ (xi1 , . . . , xil ) ∈ Xi , production yj ≡ (yj1 , . . . , yjl ) ∈ Yj , concave utility U over consumption
of goods, price vector p ≡ (p1 , . . . , pl ).

Equilibrium is the price vector p∗ such that (m + n)-tuple [(x∗i )m ∗ n

i=1 , (yj )j=1 ] achieves: (1)
consumers max utility subject to budget constraint and X , (2) firms max profit subject to Y,
(3) markets clear. Under standard Arrow-Debreu assumptions, the equilibrium exists and is
Pareto optimal.

5/15
 C ONSTANTINIDES (1982)

There exists a vector of positive numbers λ ≡ (λ1 , . . . , λm ) such that the solution to (1) below
is (xi ) = (x∗i ) and (yj ) = (y∗j ):
Xm
max λi Ui (xi )
x,y
i=1

subject to

yj ∈ Yj , ∀j
xi ∈ Xi , ∀i
X X X (1)
xih = yjh + ωih , ∀h
i j i

6/15
 C ONSTANTINIDES (1982)

Problem (1) is equivalent to (2) below:

m
" #
X
max max λi Ui (xi )
y x
i=1

subject to

yj ∈ Yj , ∀j
xi ∈ Xi , ∀i
X X X (2)
xih = yjh + ωih , ∀h
i j i

7/15
 C ONSTANTINIDES (1982)

Paggregate consumption z ≡ (z1 , . . . , zl ) and good-specific total consumption

Define
zh ≡ i xih . Now solve (3) below:
m
X
U(z) ≡ max λi Ui (xi )
x
i=1

subject to

xi ∈ Xi , ∀i
(3)
X
xih = zh , ∀h
i

8/15
 C ONSTANTINIDES (1982)
Finally, solve (4) below:
max U(z)
z,y

subject to

yj ∈ Yj , ∀j
(4)
X
yjh + ωh = zh , ∀h
j

Given λ ≡ (λ1 , . . . , λm ), if m consumers are replaced by one representative agent with utility
over aggregate consumption U(z), endowment equal to the sum of m consumers’
endowments, and shares P the sum of the m consumers’ shares, then price vector p∗ and the
associated (1 + n)-tuple ( i x∗i , y∗j ) is an equilibrium of Problem (4).

Weak static aggregation under complete markets.

9/15
 D EATON AND PAXSON (1994)
1
Discount factor is δ = (1+r) , C∗ is the bliss level, ω0 ≥ 0 is initial endowment, quadratic utility.
Permanent and transitory shocks. Agent solves:
T
" #
1X t ∗ 2
max E0 − δ (C − Ct )
C 2
t=1

subject to:
T
X
δ (yt − Ct ) + ωo = 0
t=1
p
(5)
yt = yt + ϵt
p p
yt = yt−1 + ηt

10/15
 D EATON AND PAXSON (1994)
First-differencing the solution:

1
∆Ct = ηt + PT−t ϵt ≈ ηt (6)
τ
τ =0 δ

Assume covi Cit−1 , ηti = 0. Then:

   
vari Cit − vari Cit−1 ≈ var (ηt ) (7)

Within-cohort rise in consumption inequality is the variance of permanent shocks. Complete

markets benchmark: consumption inequality should not rise within cohorts.
The permanent income model:
1. Consumption inequality rises linearly within cohorts.
2. Substantial, uninsured idiosyncratic risk.
11/15
D EATON AND PAXSON (1994)

12/15
 B LUNDELL , P ISTAFERRI , AND P RESTON (2008)
General empirical model:
1
∆ct = αηt + β PT−t ϵt
τ
τ =0 δ

1. α = β = 1: permanent income model.

2. α = β = 0: complete markets.

BPP (2008) estimate: α̂ = 32 .

Three general identification problems:
Issue 1: accumulation of wealth and the precautionary savings motive lower α.
Issue 2: permanent shocks are anticipated, i.e. news about future shocks.
Issue 3: shocks are less persistent and AR(1) process decays exponentially.

13/15
 TAKEAWAY

A representative agent can be constructed with complete markets.

Moreover, with linear Engel curves, equilibrium does not depend on (re-distribution).

Existence of a representative agent by itself does not imply irrelevance of (re-)distribution.

The choice of λ and the distribution of ω can influence the aggregate outcome.

14/15
 TAKEAWAY

Complete markets are unrealistic.

The permanent income model may be a better “standard model”.

Substantial idiosyncratic risk and a single risk-free asset to insure.

May be too extreme. Reality is more flexible.

Either agents have access to devices to partially insure against shocks or . . .

. . . idiosyncratic risk is persistent but perhaps not permament.

15/15