9.

The Stochastic Growth Model

Yvan Becard
PUC-Rio

Macroeconomics I, 2023
 Production Is Back

▶ In previous lectures, we analyzed optimal behavior under uncertainty in the

context of a pure exchange economy with stochastic endowments
▶ Today, we consider a simple production economy with stochastic technology
▶ We study the stochastic neoclassical growth model, also referred to as the
real business cycle (RBC) model
▶ This is the last of four lectures on complete markets

2
 Lecture Outline

1. Model Setup 7. Financing the Firms

2. Central Planner 8. Recursive Formulation
3. Time Zero Trading: Arrow-Debreu 9. Recursive: Central Planner
4. Implied Wealth Dynamics 10. Recursive: Sequential Trading
5. Sequential Trading: Arrow 11. Recursive Competitive Equilibrium
6. Equivalence of Allocations 12. Exercise

Main Reference: Ljungqvist and Sargent, 2018, Recursive Macroeconomic

Theory, Fourth Edition, Chapter 12

3
1. Model Setup
 The Stochastic Growth Model

▶ The environment resembles that of the standard neoclassical growth model

▶ The key difference is that technology is now stochastic
▶ We also make two minor changes
1. The labor supply is not inelastic anymore, leading to labor supply decisions
2. We introduce a second type of firm that builds capital, so as to induce more
trades among agents and price more items, in particular the capital stock

5
 Stochastic Event

▶ In each period t ≥ 0, there is a realization of a stochastic event st ∈ S

▶ The stochastic event st is an aggregate, or economy-wide, state variable
▶ Let the history of events until t be st = [s0 , s1 , . . . , st ]
▶ The history st is publicly observable

6
 Probabilities

▶ As usual, the unconditional probability of observing a particular sequence

of events st is given by probability measure πt (st )
▶ For t > τ , the probability of observing st conditional on the realization of
history sτ is πt (st |sτ )
▶ Again, the initial state s0 in period 0 is nonstochastic, ie π0 (s0 ) = 1

7
 Households

▶ A representative household has preferences over streams of consumption

ct (st ) and leisure ℓt (st )
∞ X
X
β t u[ct (st ), ℓt (st )]πt (st ), β ∈ (0, 1)
t=0 st

▶ u satisfies the usual Inada conditions

u(0, ℓ) = u(c, 0) = 0 lim uc (c, ℓ) = lim uℓ (c, ℓ) = ∞

c→0 ℓ→0
uc , uℓ > 0, ucc , uℓℓ < 0 lim uc (c, ℓ) = lim uℓ (c, ℓ) = 0
c→∞ ℓ→1

8
 Work or Chill

▶ In each period, the household is endowed with one unit of time that can be
either devoted to leisure ℓt (st ) or labor nt (st )

1 = ℓt (st ) + nt (st )

▶ If the utility function did not depend on leisure, u[c(st )]πt (st ), the household
would choose ℓt (st ) = 0 and nt (st ) = 1 for all t, ie an inelastic labor supply
▶ Here, because households value leisure, they end up choosing nt (st ) < 1

9
 Why Do We Have a Representative Household?

▶ As soon as we assume complete markets, which is the case in this model, we

can rationalize the representative household construct as follows
▶ Assume there are I consumers named i = 1, 2, . . . , I, and all consumers have
the same utility function ui [cit (st ), ℓit (st )] = u[cit (st ), ℓit (st )]
▶ Each consumer receives idiosyncratic labor productivity shocks eit (st )nit (st ),
but optimally trades state-price securities to insure the risk away
▶ In sum, there is no idiosyncratic risk in this economy, only aggregate risk

10
 Production Function

▶ Output is produced according to the production function

At (st )F [kt (st−1 ), nt (st )]

▶ Notice capital in period t depends on the state in period t − 1

▶ At (st ) is a stochastic process of Harrod-neutral technology shocks
▶ F satisfies the standard assumptions

F (0, n) = F (k, 0) = 0 lim Fk (k, n) = lim Fn (k, n) = ∞

k→0 n→0
Fk , Fn > 0, Fkk , Fnn < 0 lim Fk (k, n) = lim Fn (k, n) = 0
k→∞ n→∞

▶ Write the function in intensive form: F (k, n) ≡ nf (k̂) where k̂ ≡ k

n

11
 Constraints

▶ Output goods are consumption and investment goods

ct (st ) + it (st ) ≤ At (st )F [kt (st−1 ), nt (st )]

▶ Capital accumulates according to

kt+1 (st ) = (1 − δ)kt (st−1 ) + it (st )

▶ Capital kt+1 (st ), to be used in production in t + 1, is built in advance in t

12
 Consolidated Constraint

▶ The investment good it (st ) can take negative values, meaning the capital
stock is reversible, ie it can be freely reconverted into consumption
▶ Consumption, however, cannot be negative
▶ Plug the law of motion of capital into the resource constraint

ct (st ) + kt+1 (st ) − (1 − δ)kt (st−1 ) ≤ At (st )F [kt (st−1 ), nt (st )]

13
2. Central Planner
 Problem of the Central Planner

▶ The planner chooses an allocation {ct (st ), ℓt (st ), it (st ), nt (st ), kt+1 (st )}∞
t=0 to
maximize the utility function, subject to
▶ The time constraint
▶ The resource constraint
▶ The initial capital stock k0
▶ The stochastic process for the level of technology At (st )

15
 Lagrangian

▶ Write a Lagrangian, in which we directly plug the time constraint

∞ X
X 
L= β t πt (st ) u[ct (st ), 1 − nt (st )]
t=0 st

t
 t t−1 t t−1 t t

+ µt (s ) At (s )F [kt (s ), nt (s )] + (1 − δ)kt (s ) − ct (s ) − kt+1 (s )

▶ µt (st ) is a process of Lagrange multipliers on the resource constraint

16
 First-Order Conditions

▶ For each t and st , the first-order conditions are

ct : uc (st ) = µt (st )
nt : uℓ (st ) = uc (st )At (st )Fn (st )
X
kt+1 : πt (st )uc (st ) = β uc (st+1 )πt+1 (st+1 )[At+1 (st+1 )Fk (st+1 ) + 1 − δ]
st+1 |st

17
 Notation

▶ In the FOC for capital, note that

X X X
=
st+1 st st+1 |st

where the summation over st+1 |st means we sum over all possible histories
s̃t+1 such that s̃t = st
▶ Summations over histories and events are different
X X
̸=
st+1 |st st+1

18
 Summing Over Histories

t−2

t−1

t+1
(st+1 = 0 | st ) (st+1 = 1 | st )
P
st+1 |st

19
 Summing Over Events

t−2

t−1

t
trading period

t+1
(st+1 = 0 | st ) (st+1 = 1 | st )
P
st+1

20
3. Time Zero Trading: Arrow-Debreu Securities
 Three Types of Agents

▶ Let’s solve the competitive equilibrium with time 0 trading of a complete set
of dated- and history-contingent Arrow-Debreu securities
▶ Trades occur among three representative agents
1. The representative household
2. A representative goods producer, which we call type I firm
3. A representative capital producer, which we call type II firm

22
 Model Diagram

Labor Rent labor

Goods
market Producers
or
lab
out ods
nt l go
Re
Sel d
den
vi
Goods di
Households Buy goods
p ay
market ta
l,
pi
ca
Sel
l in nt Bu
itia Re yg
ood
l ca s
pit
al

Capital Buy initial capital,

rent out capital,
Capital
market pay dividend Producers

23
 Actions

▶ Households own the initial capital stock k0 , sell it to capital producers at

date 0, rent out labor services to and buy goods from goods producers
▶ Goods producers rent labor from households, rent capital from capital
producers, produce, sell output goods to households and capital producers
▶ Capital producers buy initial k0 from households, buy investment goods
from goods producers, produce capital, rent out capital to goods producers

24
 Problem of the Household

▶ The household maximizes

∞ X
X
β t u[ct (st ), 1 − nt (st )]πt (st )
t=0 st
∞
XX ∞ X
X
subject to qt0 (st )ct (st ) ≤ wt0 (st )nt (st ) + pk0 k0
t=0 st t=0 st

▶ qt0 (st ) is the price of one unit of output/consumption contingent on history st

▶ wt0 (st ) is the price of one unit of labor contingent on history st
▶ pk0 is the price of one unit of the initial capital stock

25
 Interpretation

▶ We are at time 0, markets open, and all trading takes place

▶ At time 0, the consumer sells her entire lifetime income stream, made of
P P
labor income ∞ t=0 st wt (s )nt (s ) and a one-off capital sale pk0 k0
0 t t

▶ The consumer sells her labor to firm I and her capital to firm II
▶ With the proceeds, the consumer buys an infinite sequence of consumption
P P 0 t
claims ∞ t=0 st qt (s )ct (s ) from firm I, the goods producer
t

▶ At the end of period 0, when trading is complete, markets close forever

26
 Lagrangian

▶ Write a Lagrangian
∞ X
X
L= β t u[ct (st ), 1 − nt (st )]πt (st )
t=0 st
"∞ ∞ X
#
XX X
+η wt0 (st )nt (st ) + pk0 k0 − qt0 (st )ct (st )
t=0 st t=0 st

▶ η is the unique Lagrange multiplier on the time 0 budget constraint

27
 First-Order Conditions

▶ The first-order conditions are

ct (st ) : β t uc (st )πt (st ) = ηqt0 (st )

nt (st ) : β t uℓ (st )πt (st ) = ηwt0 (st )

28
 Goods Producer – Firm of Type I

▶ The goods producer, or firm I, operates the production technology

▶ At time 0, the firm enters into state-contingent contracts for each t and each
st to rent capital ktI (st ) and labor services nt (st ) and sell output yt (st )
▶ It trades with households over labor services nt (st ) and goods ct (st ) and it
trades with capital producers over capital services ktI (st ) and goods it (st )

29
 Problem of the Goods Producer

▶ The goods producer maximizes

∞ X
X 
qt0 (st )[ct (st ) + it (st )] − rt0 (st )ktI (st ) − wt0 (st )nt (st )
t=0 st
subject to ct (st ) + it (st ) ≤ At (st )F [ktI (st ), nt (st )]

▶ rt0 (st ) is the price for renting capital, ie the rental rate
▶ Note that all variables in this problem are conditioned on the history st

30
 First-Order Conditions

▶ Plug the constraint into the objective

∞ X
X  0 t
max qt (s )At (st )F [ktI (st ), nt (st )] − rt0 (st )ktI (st ) − wt0 (st )nt (st )
t=0 st

▶ The first-order conditions are

ktI (st ) : qt0 (st )At (st )Fk (st ) = rt0 (st )
nt (st ) : qt0 (st )At (st )Fn (st ) = wt0 (st )

31
 Capital Producer – Firm of Type II

▶ The capital producer operates a technology to transform output goods (ie

investment goods) into capital goods
▶ At time 0, it enters into state-contingent contracts for each t and each st
▶ It purchases initial capital k0 from households and builds new capital
II (st ) by purchasing investment goods i (st ) from goods producers
kt+1 t
▶ It earns revenues by renting out capital rt+1
0 (st+1 )k II (st ) to goods producers
t+1

32
 Problem of the Capital Producer

▶ The capital producer maximizes profit

∞ X
X  0 t II t−1
−pk0 k0II + rt (s )kt (s ) − qt0 (st )it (st )
t=0 st
subject to II
kt+1 (st ) = (1 − δ)ktII (st−1 ) + it (st )

▶ Note that some variables are conditioned on st−1 , others on st

33
 Who Bears the Risk?

▶ Capital in period 0 k0II is bought without any uncertainty about the rental
price r00 (s0 ) in that period
▶ But for any future period t, investment in capital ktII (st−1 ), conditioned on
st−1 , is made without knowing the rental price rt0 (st ), conditioned on st
▶ So firm II must deal with the risk associated with capital being assembled
one period prior to becoming an input for production and yielding a return
▶ Firm I, on the other hand, can choose how much capital ktI (st ) to rent in
period t conditioned on history st , ie it faces no risk at all

34
 Problem of the Capital Producer

▶ Plug the capital accumulation constraint into the objective and rearrange
n o
k0II − pk0 + r00 (s0 ) + q00 (s0 )(1 − δ)
 
∞ X
X  X 
II
+ kt+1 (st ) −qt0 (st ) + 0
[rt+1 (st+1 ) + qt+1
0
(st+1 )(1 − δ)]
 
t=0 s
t s t+1 t
|s

▶ Profit is a linear function of investments in capital k0II and kt+1

II (st )

▶ What must happen to the two terms in curly brackets {. . . }?

35
 Details of the Computation
∞ Xn
X o
− pk0 k0II + rt0 (st )ktII (st−1 ) − qt0 (st )kt+1
II
(st ) + qt0 (st )(1 − δ)ktII (st−1 )
t=0 st
∞ X
X
= − pk0 k0II + r00 (s0 )k0II + q00 (s0 )(1 − δ)k0II + −qt0 (st )kt+1
II
(st )
t=0 st
∞ Xn
X o
+ rt0 (st )ktII (st−1 ) + qt0 (st )(1 − δ)ktII (st−1 )
t=1 st
∞ X
X
= k0II −pk0 + r00 (s0 ) + q00 (s0 )(1 − δ) + −qt0 (st )kt+1
II
(st )

t=0 st
∞ X
X X n o
0
+ rt+1 (st+1 )kt+1
II
(st ) + qt+1
0
(st+1 )(1 − δ)kt+1
II
(st )
t=0 st st+1 |st

= k0II−pk0 + r00 (s0 ) + q00 (s0 )(1 − δ)


 
∞ X
X  X 
II
+ kt+1 (st ) −qt0 (st ) + 0
[rt+1 (st+1 ) + qt+1
0
(st+1 )(1 − δ)]
t
 
t=0 s t+1 t s |s

36
 Perfect Competition Means Zero Profit

▶ If the terms in curly brackets are positive, the firm wants infinite k0II , kt+1
II

▶ If the terms in curly brackets are negative, the firm wants zero k0II , kt+1
II

▶ In equilibrium: 1) perfect competition implies zero profits; 2) supply equals

demand, meaning capital cannot be zero or infinite, 0 < k0II , kt+1
II < ∞

▶ We conclude that the two terms in curly brackets from the firm’s profit
equation must be equal to zero: this is the zero-profit condition

37
 First-Order Conditions

▶ Based on the preceding, equilibrium prices satisfy

pk0 = r00 (s0 ) + q00 (s0 )(1 − δ)

X
qt0 (st ) = 0
[rt+1 (st+1 ) + qt+1
0
(st+1 )(1 − δ)]
st+1 |st

38
 Summary of Necessary Conditions
▶ Households

β t uc (st )πt (st ) = ηqt0 (st )

β t uℓ (st )πt (st ) = ηwt0 (st )

▶ Goods producer or firm I

qt0 (st )At (st )Fk (st ) = rt0 (st )

qt0 (st )At (st )Fn (st ) = wt0 (st )

▶ Capital producer or firm II

pk0 = r00 (s0 ) + q00 (s0 )(1 − δ)

X
qt0 (st ) = 0
[rt+1 (st+1 ) + qt+1
0
(st+1 )(1 − δ)]
st+1 |st

39
 Equilibrium

▶ In equilibrium, markets clear, ie supply equals demand

▶ Let’s compute the equilibrium price and quantities in the three markets
1. Labor market
2. Capital market
3. Goods market

40
 Labor Market Equilibrium

▶ Labor supply is set by the household’s FOC for labor

▶ Labor demand comes from the goods producer’s (firm I) FOC for labor
▶ Combine the two

β t uℓ (st )πt (st ) = ηqt0 (st )At (st )Fn (st )

41
 Capital Market Equilibrium

▶ Capital supply comes from the capital producer’s (firm II) FOC
▶ Capital demand comes from the goods producer’s (firm I) FOC for capital
▶ Combine the two
X
qt0 (st ) = 0
qt+1 (st+1 )[At+1 (st+1 )Fk (st+1 ) + 1 − δ]
st+1 |st

42
 Goods Market Equilibrium

▶ Supply of goods comes from the goods producer (firm I)

▶ Demand for goods comes from households and capital producers (firm II)
▶ By Walras’ law, or by virtue of the resource constraint at equality, the goods
market is in equilibrium

At (st )F [kt (st−1 ), nt (st )] = ct (st ) + it (st )

43
 Consumption–Labor Choice

▶ Plug the household’s consumption FOC into the labor market equilibrium
equation
uℓ (st )
= At (st )Fn (st ) = wt (st )
uc (st )
▶ This is the intratemporal labor supply–consumption decision
▶ The marginal rate of substitution (MRS) between leisure and consumption
uℓ (st )
uc (st ) equals the relative price of leisure, ie the wage wt (s )
t

44
 Euler Equation

▶ Plug the household’s consumption FOC into the capital market equilibrium
equation
X
πt (st )uc (st ) = β uc (st+1 )πt+1 (st+1 )[At+1 (st+1 )Fk (st+1 ) + 1 − δ]
st+1 |st

▶ This is the intertemporal consumption–saving decision, the Euler equation

t
▶ The MRS between consumption today and tomorrow βE uuc (s(st+1 )
)
equals the
0 c
relative price of consumption, ie the (expected) interest rate

45
 Equivalence

▶ The previous two expressions are identical to the central planner’s

first-order conditions
▶ The allocation in the competitive equilibrium with time 0 trading is the
same as the Pareto efficient allocation

46
4. Implied Wealth Dynamics
 Change the Numeraire

▶ In the Arrow-Debreu world, trades are only executed at time 0

▶ We can still compute how the household’s wealth evolves over time
▶ For this we need to express all prices, wages, and rental rates in terms of
time t, history st consumption goods
▶ In other words, we change the numeraire

48
 Deflating

▶ We obtain

qτ0 (sτ ) uc (sτ )

qτt (sτ ) ≡ 0 = β τ −t πτ (sτ |st )
t
qt (s ) uc (st )
w0 (sτ )
wτt (sτ ) ≡ 0τ t
qt (s )
r0 (sτ )
rτt (sτ ) ≡ τ0 t
qt (s )
▶ Notice that
qt0 (st )
qtt (st ) = =1
qt0 (st )

49
 Wealth

▶ In lecture 7, we computed households’ financial wealth as total wealth

minus the present value of current and future endowment streams
▶ Here, we subtract the present value of current and future labor income
▶ Household wealth, or the value of all current and future net claims, in time
t, history st consumption goods is
∞ X
X 
t
Υt (s ) ≡ qτt (sτ )cτ (sτ ) − wτt (sτ )nτ (sτ )
τ =t sτ |st

50
 Rewriting Wealth
∞ X
X 
Υt (st ) ≡ qτt (sτ )cτ (sτ ) − wτt (sτ )nτ (sτ )
τ =t sτ |st
∞ X
X   
= qτt (sτ ) Aτ (sτ )F [kτ (sτ −1 ), nτ (sτ )] + (1 − δ)kτ (sτ −1 ) − kτ +1 (sτ ) − wτt (sτ )nτ (sτ )
τ =t sτ |st
∞ X
X   
qτt (sτ ) Aτ (sτ ) Fk (sτ)kτ (sτ −1) + Fn (sτ)nτ (sτ) + (1−δ)kτ (sτ −1) − kτ +1 (sτ) − wτt (sτ )nτ (sτ)
 
=
τ =t sτ |st
∞ X
X 
rτt (sτ )kτ (sτ −1 ) + qτt (sτ ) (1 − δ)kτ (sτ −1 ) − kτ +1 (sτ )
 
=
τ =t sτ |st
∞
X X  X  t τ  
= rtt (st)kt (st−1)+qtt (st )(1−δ)kt (st−1)+ rτ (s ) + qτt (sτ )(1 − δ) −qτt −1 (sτ −1) kτ (sτ −1 )
τ =t+1sτ −1 |st sτ |sτ −1

rtt (st ) t−1

 
= + 1 − δ kt (s )

Second line: resource constraint; Third: Euler’s theorem; Fourth: firm’s FOCs; Fifth: rearrange;
q 0 (st )
Sixth: qtt (st ) = qt0 (st ) = 1 and zero-profit condition implying curly bracket terms equal zero
t
51
 Wealth Is Capital

▶ We find that the wealth of the representative household, excluding its labor
income, is  
Υt (st ) = rtt (st ) + 1 − δ kt (st−1 )
▶ Households invest all their wealth in the capital stock
▶ The entire capital stock is held by households

52
5. Sequential Trading: Arrow Securities
 Sequential Trading

▶ We now study the same economy with sequential trading

▶ All markets reopen in each period
1. Goods market
2. Labor market
3. Capital market

54
 Households

▶ At each date t ≥ 0 after history st , the household brings in assets ãt (st ), ie
claims to time t consumption that it bought in period t − 1
▶ In addition, the household earns new labor income w̃t (st )nt (st ) by selling
labor services to goods producers
▶ It uses these revenues to buy consumption goods c̃t (st ) and claims to time
t + 1 consumption whose payment is contingent on the realization of st+1

55
 Budget Constraint

▶ The household faces a sequence of budget constraints; the time t, history st

budget constraint is
X
c̃t (st ) + ãt+1 (st+1 , st )Q̃t (st+1 |st ) ≤ w̃t (st )ñt (st ) + ãt (st )
st+1

▶ Q̃t (st+1 |st ) is the pricing kernel: the price of one unit of consumption at time
t + 1 contingent on the realization st+1 at t + 1 when history at t is st
▶ {ãt+1 (st+1 , st )} is a vector of claims on time t + 1 consumption, ie there is
one element of the vector for each value of time t + 1 realization of st+1

56
 No Ponzi Scheme

▶ To rule out Ponzi schemes, we must impose borrowing constraints on the

household’s asset position
▶ Without these borrowing constraints, the household would find it optimal to
borrow as much as possible and roll over debt forever
▶ What is the maximal amount the household can repay?

57
 Natural Debt Limit

▶ Let’s compute the state-contingent natural debt limit

∞ X
X
w̃τt (sτ )ñmax
τ (sτ )
τ =t sτ |st

▶ The maximum the agent could earn and repay is if she promises to work
ñmax
τ = 1 for all τ and sτ if necessary
▶ But this is not credible as ℓ̃τ = 0 would ruin her utility

58
 Arbitrary Borrowing Constraint

▶ We can impose that indebtedness in any state next period −ãt+1 (st+1 , st ) is
bounded by some arbitrary constant
▶ As long as the budget constraint is bounded, equilibrium forces ensure that
the household holds the market portfolio
▶ Let’s impose ãt+1 (st+1 , st ) ≥ 0, ie wealth can never be negative

59
 Problem of the Household

▶ The household maximizes

∞ X
X
β t u[c̃t (st ), 1 − ñt (st )]πt (st )
t=0 st
X
subject to c̃t (s ) +
t
ãt+1 (st+1 , st )Q̃t (st+1 |st ) ≤ w̃t (st )ñt (st ) + ãt (st )
st+1
and ãt+1 (st+1 , st ) ≥ 0

60
 Lagrangian

▶ Write a Lagrangian
∞ X
X
L= β t u[c̃t (st ), 1 − ñt (st )]πt (st )
t=0 st
 X 
t t t t t t t
+ ηt (s ) w̃t (s )ñt (s ) + ãt (s ) − c̃(s ) − ãt+1 (st+1 , s )Q̃t (st+1 |s )
st+1

+ νt (st , st+1 )ãt+1 (st+1 )

▶ ηt (st ) are multipliers on the flow budget constraint

▶ νt (st , st+1 ) are multipliers on the borrowing constraint

61
 First-Order Conditions

▶ The FOCs are

c̃t (st ) : β t uc [c̃t (st ), 1 − ñt (st )]πt (st ) − ηt (st ) = 0

ñt (st ) : − β t uℓ [c̃t (st ), 1 − ñt (st )]πt (st ) + ηt (st )w̃t (st ) = 0
{ãt+1 (st+1 , st )}st+1 : − ηt (st )Q̃t (st+1 |st ) + νt (st , st+1 ) + ηt+1 (st+1 , st ) = 0

▶ They hold for all st+1 , t, st

62
 Nonbinding Borrowing Constraint

▶ We conjecture that the arbitrary debt limit is not binding

▶ As a result, the Lagrange multipliers νt (st , st+1 ) are all equal to zero
▶ Let’s rewrite the FOCs with νt (st , st+1 ) = 0

63
 Rewriting the First-Order Conditions

▶ The optimal static consumption–labor choice is

uℓ [c̃t (st ), 1 − ñt (st )]

w̃t (st ) =
uc [c̃t (st ), 1 − ñt (st )]
▶ The optimal dynamic consumption–saving choice is

uc [c̃t+1 (st+1 ), 1 − ñt+1 (st+1 )]

Q̃t (st+1 |st ) = β πt (st+1 |st )
uc [c̃t (st ), 1 − ñt (st )]

64
 Goods Producer – Firm of Type I

▶ At each date t ≥ 0 after history st , the goods producer solves the usual static
problem
n o
max At (st )F [k̃ I (st ), ñt (st )] − r̃t (st )k̃tI (st ) − w̃t (st )ñt (st )
ñt (st ),k̃tI (st )

65
 First-Order Conditions

▶ The FOCs are

k̃tI (st ) : r̃t (st ) = At (st )Fk (st )

ñt (st ) : w̃t (st ) = At (st )Fn (st )

▶ The firm makes zero profit and its size is indeterminate

▶ The firm is willing to produce any quantity of output that the market
demands so long as the two FOCs are satisfied

66
 Capital Producer – Firm of Type II

▶ The capital producer’s problem is a two-period problem

1. At the end of period t after history st , the firm decides how much capital
II (st ) to produce and store; the cost of one unit of k̃ II (st ) is 1
k̃t+1 t+1
2. In the next period t + 1, the firm earns a stochastic rental revenue
II (st ) and a deterministic liquidation value (1 − δ)k̃ II (st )
r̃t+1 (st+1 )k̃t+1 t+1
▶ To finance its operations, the firm issues Arrow securities to households
▶ We use prices Q̃t (st+1 |st ) to express future income streams in today’s value

67
 Problem of the Capital Producer

▶ At each date t ≥ 0, the capital producer solves

 
 X  
II
max k̃t+1 (st ) −1 + Q̃t (st+1 |st ) r̃t+1 (st+1 ) + 1 − δ
II (st )
k̃t+1  
st+1

▶ The price of one unit of capital today in terms of today’s output goods is one
▶ The zero-profit condition is
X  
1= Q̃t (st+1 |st ) r̃t+1 (st+1 ) + 1 − δ
st+1

68
6. Equivalence of Allocations
 Equivalence of Allocations

▶ Time 0 trading and sequential trading are equivalent if

{ct (st ), ℓt (st ), nt (st ), it (st ), kt+1 (st )}∞ t t t t t ∞

t=0 = {c̃t (s ), ℓ̃t (s ), ñt (s ), ĩt (s ), k̃t+1 (s )}t=0

▶ To show the equivalence of allocations, we employ the guess and verify

method used in lecture 7; this is left as an exercise

70
 Guess and Verify

▶ The trick is to guess that the prices in the sequential equilibrium satisfy

Q̃t (st+1 |st ) = qt+1

t
(st+1 )
w̃t (st ) = wt (st )
r̃t (st ) = rt (st )

▶ We also guess that the household chooses the following asset portfolios

ãt+1 (st+1 , st ) = Υt+1 (st+1 ) for all st+1 and t

71
 Initial Capital

▶ We have to show that the agent can afford these asset portfolios
▶ In doing that, we will find that the required initial wealth is

ã0 = [r00 (s0 ) + 1 − δ]k0 = pk0 k0

▶ The household starts out at time 0 owning the initial capital stock
▶ This is different from lecture 7 where initial wealth ãi0 was zero for all i

72
7. Financing the Firms
 Financing the Goods Producer

▶ In each period, the goods producer must remunerate workers and capital
owners, ie capital producers
▶ The goods producer finances these expenses by selling output in the very
same period to consumers and capital producers
▶ The firm makes zero profit or loss for all t and st
▶ Thus it does not need to issue debt to finance its operations

74
 Financing the Capital Producer

▶ By contrast, the capital producer finances its purchases of capital by issuing

Arrow securities, ie one-period-ahead state-contingent claims, to households
▶ To produce k̃t+1
II (st ) units of capital today in period t, firm II issues claims

that promise to pay [r̃t+1 (st+1 ) + 1 − δ]k̃t+1

II (st ) goods tomorrow in state s
t+1
▶ Express these payouts in units of today’s time t good
X   II t
Q̃t (st+1 |st ) r̃t+1 (st+1 ) + 1 − δ k̃t+1 (s )
st+1

75
 Zero Profit, Zero Net Worth

▶ The capital producer makes zero profit, implying that it breaks even by
issuing these claims and then repaying them next period with interest
▶ It follows that the capital producer has zero net worth, or zero equity, and is
entirely financed by debt; in other words, it has infinite leverage

76
 Positive Wealth

▶ The household’s wealth is given by

ãt (st ) = Υt (st ) = [r̃t (st ) + 1 − δ]k̃t (st−1 )

▶ The wealth of the household is equal to the value of firm II, ie the value of
the capital stock
▶ The capital producer is entirely owned by its unique creditor, the household

77
 Nonbinding Constraint

▶ The household willingly holds the capital stock

▶ Equilibrium prices entice the household to enter each period with a strictly
positive net asset level
▶ We confirm the correctness of our conjecture that the zero debt limit is
never binding

78
 Unique Creditor

▶ The household is the only creditor of the capital producer

ãt (st ) = [r̃t (st ) + 1 − δ]k̃t (st−1 )

X X
ãt+1 (st+1 , st )Q̃(st+1 |st ) = [r̃t+1 (st+1 ) + (1 − δ)]Q̃t (st+1 |st ) k̃t+1 (st )
st+1 st+1
| {z }
= 1 by firm II’s FOC

▶ Thus the household budget constraint can be written as

X
c̃t (st ) + ãt+1 (st+1 , st )Q̃t (st+1 |st ) ≤ w̃t (st )ñt (st ) + ãt (st )
st+1
c̃t (st ) + k̃t+1 (st ) ≤ w̃t (st )ñt (st ) + [r̃t (st ) + 1 − δ]k̃t (st−1 )

79
 Useless Capital Producer

▶ The capital producer is entirely owned by the household and there is no

financing friction between the two agents
▶ Therefore, the household can play the role of the capital producer by renting
out the capital stock directly to the goods producer (firm I)
▶ In other words, we can get rid of the capital producer without changing
anything to the equilibrium conditions

80
 Equivalent Model Diagram

Labor
market
la bor Re
nt
ut lab
n to or
Re

Goods Goods
Households Purchase goods Sell goods
market Producers
Re l,
nt ita
ou cap nd
tc
ap e nt ivide
ita R
l yd
pa

Capital
market

81
8. Recursive Formulation
 Equivalence

▶ We established identical equilibrium allocations in the

1. Complete-market Arrow-Debreu economy with all trading at time 0
2. Complete-market Arrow economy with sequential trading

83
 Arbitrary Process

▶ The finding holds for any arbitrary technology process

▶ At (st ) is a measurable function of the history of events st
▶ These events st , in turn, are governed by some arbitrary probability
measure πt (st )

84
 Huge State Space

▶ In this general setup, all prices {Q̃t (st+1 |st ), w̃t (st ), r̃t (st )} and quantities
{kt+1 (st ), ct (st ), ℓt (st )} depend on the entire history of events st
▶ They are time-varying functions of all past events {sτ }tτ =0
▶ To obtain a recursive formulation, we need to make further assumptions on
the exogenous process for technology

85
 The Stochastic Event Is Markov

▶ The first assumption we make is that the stochastic event st is governed by

a Markov process, [s ∈ S, π(s′ |s), π0 (s0 )]

π0 (s0 ) = 1
πt (st ) = π(st |st−1 )π(st−1 |st−2 ) . . . π(s1 |s0 )π0 (s0 )
πt (st |sτ ) = π(st |st−1 )π(st−1 |st−2 ) . . . π(sτ +1 |sτ ) for t > τ

86
 Technology Is a Time-Invariant Function

▶ The second assumption is that aggregate technology is a time-invariant

function of its level in the last period and the current stochastic event st

At (st ) = A[At−1 (st−1 ), st ]

▶ Let’s consider the multiplicative version

At (st ) = st At−1 (st−1 ) = s0 s1 . . . st A−1

87
 State Variables

▶ What are the state variables?

88
 State Variables

▶ What are the state variables?

▶ There are two exogenous aggregate state variables
1. The current value of the stochastic event s
2. The current technology level A
▶ There is one endogenous aggregate state variable
3. The beginning-of-period capital stock K

88
 Aggregate State of the Economy

 
▶ Thanks to our two assumptions, the state vector X ≡ K A s is a
complete summary of the economy’s current position
▶ It is all that is needed for a planner to compute an optimal allocation
▶ It is all that is needed for the “invisible hand”, ie households and firms, to
call out prices and implement the first-best allocation

89
9. Recursive Formulation: Central Planner
 Problem of the Central Planner

▶ Let C, N, K denote objects in the planning problem that correspond to c, n, k

in the decentralized economy
▶ The central planner chooses C, N , K ′ to maximize the utility function of the
representative household

91
 Bellman Equation

▶ The Bellman equation writes

( )
X
v(K, A, s) = max ′ u(C, 1 − N ) + β π(s′ |s)v(K ′ , A′ , s′ )
C,N,K
s′

subject to

K ′ + C ≤ AF (K, N ) + (1 − δ)K
A′ = As′

92
 Policy Functions

 
▶ Using the definition of the state vector X ≡ K A s , we denote the
optimal policy functions as

C = ΩC (X)
N = ΩN (X)
K ′ = ΩK (X)

▶ Equation A′ = As′ and the Markov transition density π(s′ |s) induce a
transition density Π(X ′ |X) on the state X

93
 First-Order Conditions

▶ Define for convenience

Uc (X) ≡ uc [ΩC (X), 1 − ΩN (X)] Fk (X) ≡ Fk [K, ΩN (X)]

Uℓ (X) ≡ uℓ [ΩC (X), 1 − ΩN (X)] Fn (X) ≡ Fn [K, ΩN (X)]

▶ The first-order conditions are

Uℓ (X) = Uc (X)AFn (X)

X Uc (X ′ ) ′
1=β Π(X ′ |X) [A FK (X ′ ) + 1 − δ]
′
Uc (X)
X

94
10. Recursive Formulation: Sequential Trading
 Endogenous State

▶ Relative to lecture 8, we now have an endogenous state variable, namely

the aggregate capital stock Kt
▶ How do we deal with this in a competitive economy?
▶ We use a “Big K, little k” device

96
 Price Taker vs Price Maker

▶ So far we have assumed that each individual firm and household is a price
taker, ie each acts as if their decisions do not affect current or future prices
▶ In sequential market setting, prices depend on the state, of which Kt is part
▶ But of course, in the aggregate, agents choose the motion of capital Kt , and
so trough their combined actions they determine prices, ie are price makers
▶ “Big K, little k” is a device that makes them ignore this fact when they
solve their individual decision problem

97
 Big K, Little k

▶ Big K is an endogenous state variable, useful to forecast prices, but which

agents regard as beyond their control
▶ Small k is chosen by firms and consumers
▶ In the equilibrium, after firms and consumers have optimized, we set

K=k

98
 Price System

▶ We specify price functions (prices are functions of the aggregate state X)

▶ r(X) is the rental price of capital
▶ w(X) is wage rate for labor
▶ Q(X ′ |X) is the price of a claim to one unit of consumption next period
when next period’s state is X ′ and this period’s state is X
▶ All are measured in units of this period’s consumption good

99
 Perceived Law of Motion

▶ We take as given an arbitrary perceived law of motion for K

K ′ = G(X)

▶ This equation together with A′ = As′ and a given subjective transition

density π̂(s′ |s) induce a subjective transition density Π̂(X ′ |X) for state X
▶ The perceived law of motion of K and the transition probability Π̂(X ′ |X)
describe the beliefs of the household

100
 Household Problem

▶ Let J be the value function; the Bellman equation writes

( )
X
′ ′ ′
J(a, X) = max u(c, 1 − n) + β J[ā(X ), X ]Π̂(X |X)
c,n,ā(X ′ )
X′

subject to
X
c+ Q(X ′ |X)ā(X ′ ) ≤ w(X)n + a and ā(X ′ ) ≥ 0
X′
 
▶ X ≡ K A s is the vector of state variables
▶ a is the household’s individual wealth in units of current goods
▶ ā(X ′ ) is next period’s wealth in units of next period’s consumption goods

101
 First-Order Conditions
▶ The first-order conditions are

ūℓ (a, X) = ūc (a, X)w(X)

ūc [σ a (a, X; X ′ ), X ′ ]
Q(X ′ |X) = β Π̂(X ′ |X)
ūc (a, X)

where the household’s optimal policy functions are

c = σ c (a, X); n = σ n (a, X); ā(X ′ ) = σ a (a, X; X ′ )

and for convenience

ūc (a, X) ≡ uc [σ c (a, X), 1 − σ n (a, X)]

ūℓ (a, X) ≡ uℓ [σ c (a, X), 1 − σ n (a, X)]

102
 Problem of the Goods Producer

▶ The static problem of the goods producer writes

max{AF (k, n) − r(X)k − w(X)n}

k,n

▶ The zero-profit conditions are

r(X) = AFk (k, n)

w(X) = AFn (k, n)

103
 Problem of the Capital Producer

▶ The problem of the capital producer writes

( )
X
′ ′ ′
max
′
k −1 + Q(X |X)[r(X ) + 1 − δ]
k
X′

▶ The zero-profit condition is

X
1= Q(X ′ |X)[r(X ′ ) + 1 − δ]
X′

104
11. Recursive Competitive Equilibrium
 Equilibrium

▶ So far we have taken the price functions r(X), w(X), Q(X|X ′ ), the perceived
law of motion K ′ = G(X), and Π̂(X ′ |X) as given arbitrarily
▶ We now impose equilibrium conditions on these objects and make them
outcomes in the analysis; we impose

K=k

▶ Imposing equality afterward makes the household and firms be price takers

106
 Debt Supply and Debt Demand

▶ The supply of state-contingent debt issued by the capital producer must be

equal to the demand for debt coming from the household

ā(X ′ ) = [r(X ′ ) + 1 − δ]K ′

▶ Beginning-of-period assets must also satisfy

a(X) = [r(X) + 1 − δ]K

107
 Rewriting the Budget Constraint

▶ Plug the previous conditions into the household’s budget constraint

X
Q(X ′ |X)[r(X ′ ) + 1 − δ]K ′ = [r(X) + 1 − δ]K + w(X)n − c
X′
P
▶ Use the capital producer’s FOC X ′ Q(X ′ |X)[r(X ′ ) + 1 − δ] = 1 and the fact
that K ′ is predetermined when entering next period

K ′ = [r(X) + 1 − δ]K + w(X)n − c

108
 Rewriting the Budget Constraint

▶ Plug in the equilibrium prices

K ′ = [AFk (k, n) + 1 − δ]K + AFn (k, n)n − c

▶ Set K = k, N = n = σ n (a, X), C = c = σ c (a, X), and use Euler’s theorem

K ′ = AF [K, σ n (a, X)] + (1 − δ)K − σ c (a, X)

▶ Use the equilibrium condition a = [r(X) + 1 − δ]K

K ′ = AF {K, σ n ([r(X) + 1 − δ]K, X)} + (1 − δ)K − σ c ([r(X) + 1 − δ]K, X)

109
 Actual Law of Motion

▶ We have expressed K ′ only as a function of the current aggregate state

 
X= K A s

K ′ = AF {K, σ n ([r(X) + 1 − δ]K, X)} + (1 − δ)K − σ c ([r(X) + 1 − δ]K, X)

▶ This is the actual law of motion of K ′ that is implied by the household’s and
firms’ optimal decisions

110
 Perceived Law of Motion

▶ Remember the perceived law of motion of capital

K ′ = G(X)

▶ We want G not to be arbitrary but to be an outcome

▶ We want to find an equilibrium perceived law of motion

111
 Rational Expectations

▶ For this we impose rational expectations: we require that the perceived and
actual laws of motions be identical, by equating the previous two equations

G(X) = AF {K, σ n ([r(X) + 1 − δ]K, X)} + (1 − δ)K − σ c ([r(X) + 1 − δ]K, X)

▶ The perceived law of motion G affects decisions σ c and σ n via the problem of
the household, therefore the right side is itself an implicit function of G
▶ In turn, G and prices imply an actual law of motion of capital
▶ Mathematically, G is a fixed point: the equation maps a perceived G and a
price system into an actual G

112
 Rational Expectations

▶ Rational expectations mean that the agent’s perception is consistent with

the equilibrium outcome
▶ The previous equation requires that the perceived law of motion for the
capital stock G(X) equal the actual law of motion
▶ The actual law is determined jointly by the decisions of the household and
the firms in a competitive equilibrium

113
 Recursive Competitive Equilibrium

A recursive competitive equilibrium with Arrow securities is a price system

r(X), w(X), Q(X ′ |X), a perceived law of motion K ′ = G(X) and associated
induced transition density Π̂(X ′ |X), a borrowing limit ā(X ′ ), a household value
function J(a, X), and decision rules σ c (a, X), σ n (a, x), σ a (a, X; X ′ ) such that

1. Given r(X), w(X), Q(X ′ |X), Π̂(X ′ |X), the functions σ c (a, X), σ n (a, X),
σ a (a, X; X ′ ) and the value function J(a, X) solve the household’s problem
2. For all X, r(X) and w(X) solve the goods producer’s problem

r(X) = AFk {K, σ n ([r(X) + (1 − δ)]K, X)}

w(X) = AFn {K, σ n ([r(X) + (1 − δ)]K, X)}

114
 Recursive Competitive Equilibrium

3. Q(X ′ |X) and r(X) satisfy the zero-profit condition

X
1= Q(X ′ |X)[r(X ′ ) + 1 − δ]
X′

4. G(X), r(X), σ c (a, X), σ n (a, X) satisfy the law of motion of capital

G(X) = AF {K, σ n ([r(X) + 1 − δ]K, X)} + (1 − δ)K − σ c ([r(X) + 1 − δ]K, X)

5. The perceived transition density equals the actual one

π̂ = π

115
 Remarks

▶ Item 1 enforces optimization by the household, given the prices it faces and
its expectations
▶ Item 2 requires that the goods producer break even at every capital stock
and labor supply chosen by the household
▶ Item 3 requires that the capital producer break even
▶ Market clearing is implicit when item 4 requires that the perceived and
actual laws of motion of capital be equal
▶ Item 5 and the equality of perceived and actual G imply that Π̂ = Π
▶ Thus, items 4 and 5 impose rational expectations

116
 Solving the System

▶ One could attack directly the fixed point problem at the heart of the
equilibrium definition
▶ Instead we guess a candidate G and a price system
▶ Then we verify that they form an equilibrium

117
 Using the Planning Problem

▶ Which candidates should we pick?

▶ Remember the welfare theorems: a competitive equilibrium is Pareto
efficient
▶ Thus as our candidates for G and prices we turn to the planning problem

118
 Using the Planning Problem

▶ For G we choose the planner decision rule for K ′

K ′ = ΩK (X)

▶ For prices we also choose those of the planner

r(X) = AFk (X)

w(X) = AFn (X)
Uc (X ′ ) ′
Q(X ′ |X) = βΠ(X ′ |X) [A FK (X ′ ) + 1 − δ]
Uc (X)

119
 Equivalence

▶ In equilibrium the household’s decision rules for consumption and labor

matches those of the planner

ΩC (X) = σ c ([r(X) + 1 − δ]K, X)

ΩN (X) = σ n ([r(X) + 1 − δ]K, X)

▶ The key to verifying the guesses is to show that the FOCs for firms and the
household are satisfied at these guesses; we leave this as an exercise

120
 Conclusion

▶ Economic phenomena are dynamic and uncertain

▶ We have studied two ways to model these phenomena
▶ The first way is to use Arrow-Debreu or Arrow general equilibrium
structures and search for optimal actions
▶ These optimal actions are conditional on the sequence of realizations of all
past and present random variables

121
 Conclusion

▶ The second way is to use recursive methods and search for equilibrium
decision or policy rules
▶ These rules specify current actions as a function of a limited number of
state variables that summarize all the necessary information
▶ Lucas and Prescott (1971) and Mehra and Prescott (1980) introduced the
notion of recursive competitive equilibrium
▶ It is widely used today in macroeconomics and finance

122
12. Exercise
 Exercise – Equivalence of Allocations

1. Prove the equivalence of allocations of Section 6 between the time 0 trading

and sequential trading equilibria.
2. Verify that the guesses in Section 11 are correct, ie that the recursive
competitive equilibrium with sequential trading matches the recursive
equilibrium of the central planner.

124