-
Advanced Macro and Money Lecture notes (WS 09/10)
Prof. Dr. Gerhard Illing,
LMU University of Munich 1.1 Basic Tools: The Problem of Optimal
Control – An intuitive Guide 1.1.1 Dynamic optimisation problem
with inequality constraints – Discrete time ∑ == Tt t
t
tccUV 0)(
][)0(max β
)1/(1 ρβ += discount factor with ρ as time preference.
Constraints: Law of motion tttt ckfkk −=−+ )(1
tc control variable; tk state variable (capital). Boundary
constraints: Initial capital stock 0k given. Terminal constraint:
11 ++ ≥ TT kk . Furthermore, 0;0 ≥≥ tt ck . Characterise the
optimal time path { } { } Ttkc tt ,...,0*;* = Necessary /
sufficient properties of U(.) and f(.) to guarantee existence of
optimum? Assume and ][ tcU )( tkf to be strictly concave
functions.
-
For optimisation problems, we use Lagrangian method.
∑∑ = += −−−+= Tt tttttTt tt
tckkckfcUL 0 10)(
)](][[][max λβ
Recipe: Construct the Hamiltonian:
, ]][[][),,( ttttt
tttt ckfcUckH −+= λβλ Using Ht, reformulate the Lagrangian:
∑ −∑ −= = +=Tt ttt
Tt tttttc
kkckHL 0 10)()(],,[max λλ
∑ −+−∑ += = +−= Tt TTtttTt tttt kkkckH 1 10010 )(],,[ λλλλλ
1.1.2 How to handle inequality constraints? Kuhn Tucker conditions!
Kuhn Tucker: Static optimisation with inequality constraints.
Introduce Lagrange multiplier with complementary slackness
],...,[max 1)( ntcccU s. t. mjaccg jnj ,...,1),...,( 1 =≤
Lagrangian: += ],...,[ 1 nccUL [ ]),...,( 11 njjmj j ccga −∑ = ν
with optimality conditions:
(1) i
jmj j
i cg
cU
∂∂
=∂∂
∑ =1ν
(2) mjga jjj ,...,1;0)( ==−ν complementary slackness: mjccga njj
;0),...,( 1 j ,...,1;0 =≥≥− ν
This way, Lagrange multipliers are non-negative!
2
-
Examples: Characterize optimal consumption bundle with
inequality constraints: (1) Economic intuition for shadow
prices:
],[max 21 ccUc s.t. 2211 ; cccc ≤≤
][][)(: 222111 ccccUL −+−+ νν FOC: 2211 /;/ νν =∂∂=∂∂ cUcU ; 0;0
≥≥− iii cc ν ;
2,10] =[ =− iciciiν
→ 2
1
2
1
2
1
1
2
//
//
cUcU
cUcU
cdcd
∂∂∂∂
==∂∂∂∂
=−νν
Relative shadow price for resource constraints (2) Economic
intuition for complementary slackness
],[max 21 ccUc=U(c1)+ß U(c2) s.t. 11221 ; ccYcpc ≤≤+
][][)()(: 1122121 cccpcYcUcUL −+−−++ νλβ ν: Shadow value for the
constraint 11 cc ≤
(1) 221
; pcU
cU λβνλ =
∂∂
+=∂∂ ;
(2) 0;;0][ 221221 ≥≤+=−− λλ YcpccpcY ; 0;;0][ 1111 ≥≤=− νν
cccc
→
( )2222
1
1
2 1/111//)1(
pppcUcU
cdcd
≥+=+
=∂∂∂∂
+=− λνλ
νλρ
If the shadow value ν is positive, the constraint 11 cc ≤ is
binding. If the constraint 11 cc ≤ is not binding, ν=0
3
-
1.1.3 The Maximum Principle: Hamiltonian conditions Dynamic
control problem We concentrate on the terminal value constraint 11
++ ≥ TT kk And ignore the other inequality constraints 0;0 ≥≥ tt
ckMaximise the Lagrangian:
∑∑ = +−= −+−+= Tt TTtttTt tttttckkkckHL 1 10010)(
)(],,[max λλλλλ +
)( 11 ++ − TT kkν This gives the FOC Hamiltonian conditions
(Maximum Principle):
A) 0=∂∂
t
t
cH t=1,…,T
B) 0)( 1 =−+∂∂
−ttt
t
kH λλ t=1,…,T
C) ttt
t kkH −=∂∂
+1λ t=1,…,T
D) Transversality condition (note: νλ =T )
0)( 11 =− ++ TTT kkλ with complementary slackness: If 0011
=→>− ++ TTT kk λ ; if 110 ++ =→> TTT kkλ
Interpretation: tλ as shadow price (in present value terms)
of the optimised path: t
t kL∂
∂=
(.)*λ ; 0(.)*1
≤−=∂
∂
+T
TkL λ
Lagrangian approach gives us – as additional information – the
shadow price of the state variable at each time t, once 0λ is
determined.
00
(.)* λ=∂
∂k
L . For nontrivial problems, 00
(.)* λ=∂
∂k
L >0
4
-
Economic intuition behind Hamiltonian: is the sum of ])([][],,[
tttt
ttttt ckfcUckH −+= λβλ
1) The direct impact of the state variable kt on the payoff
function at t plus
2) The change in the state variable (evaluated at present value
λt) according to the law of motion: Try to maximise this sum, but
take also into account that value of state variable may change:
][ 1−−+= ttttt kHL λλ Ht: at t realised returns;
][ 1−− tttk λλ : Indirect impact via capital gains of state
variable Maximise, at each t, the sum of returns and capital gains
→ arbitrage equation:
B) )( 1−−−=∂∂
ttt
t
kH λλ
KEY economic insights: C): Characterises the law of motion for
the state variable
(necessary condition, it is part of the basic constraints) B)
Arbitrage condition: Characterizes the path (law of
motion) for shadow value D) Transversality condition: Either the
value of the state
variable is positive, then the endpoint constraint is binding
for the state variable; if endpoint constraint not binding, shadow
value must be zero.
5
-
1.1.4 Note: Hamiltonian in current value vs. present value
Shadow price in terms of present value (evaluated at initial
stage): Alternative: Hamiltonian in current value terms:
])([][ ttttt
t ckfcUH −+= λβ
]))([][~ tttttt ckfcUH −+= μ Shadow price μ in terms of current
value (price of capital in terms of current utility)!
tt
t HH~β= with ] t
tt μβλ =)([][ ttt
tt
t ckfcU −+= μββFOC:
A*) 0~
=∂∂
t
t
cH since t
t
t
t
t
cH
cH −
∂∂
=∂∂ β
~
B*) )(~
11 −− −−=∂∂
tttt
t
kH μμρμ
t
tt
t
t
kH
kH
∂∂=
∂∂ ~β = )()( 1
11 −
−− −−=−− t
tt
ttt μβμβλλ
( ) Inserting in B) gives tt
t μβλ = )(~
11
−−−−=
∂∂
ttt
t
kH μβμ
C*) ttt
t kkH −=∂∂
+1
~
μD*) Transversality condition 0)( 11 =− ++ TT
TT kkβμ
Straightforward interpretation of B*) as arbitrage
condition:
ρμμ
μ=
Δ+
∂∂ kH /~
B*): → Value1 of capital kt: TtTtT
iit
itit k
Hμββμ −−=
+
+ +∂∂
= ∑ 1~
in present value terms: TtTiit
itt k
H λλ +∑∂∂
= −=+
+1
1 μt Value of state variable evaluated at t; λt value of state
variable evaluated at initial stage 6
-
1.1.5 Continuous time analysis Apply the same method to
continuous time problem
dtettcuV tTtc
ρ−∫= 0)(
]),([)0(max s.t. constraints:
a) Law of motion: transition equation for state variable:
(Intertemporal resource constraint) )),(),(()( ttctkgtk =&b)
0)0( 0 >= kk initial value of state variable given c) Bounds:
k(t) ≥ 0; c(t) ≥ 0. Fixed endpoint T given. We ignore bounds other
than bound at endpoint: 0)( ≥Tk c1) Sometimes impose (in market
economies) No Ponzi game condition: 0)( )( ≥⋅ − TTreTk Optimal
solution: (continuous) path for control c*(t) and state variable
k*(t). Necessary/sufficient conditions for existence of optimum:
Strict concavity of u (.) and g (.). Technique: use generalized
Lagrangian:
dttkttctkgtettcuL tT )]}()),(),(()[(]),([{0 &−+=−
∫ λρ Partial integration by parts
dttFtdtGdaGaFbGbFdttG
tdtFd b
aba )(
)()()()()()()( ∫−−=∫
gives:{ }
)()()0()0(
)()()),(),(()(]),([0TkTk
dttktttctkgtettcuL T t
λλ
λλρ
−+
++= ∫ − &
7
-
Construct Hamiltonian: )),(),(()(]),([ ttctkgtettcuH t λρ +=
−
Maximum Principle: First order conditions for the Hamiltonian
(Necessary conditions):
A) 0)()(
=∂∂
tctH
B) )()()( t
tktH λ&−=
∂∂ Arbitrage equation:
C) )()()( tk
ttH &=
∂∂
λ Law of motion
D) k(0)=k0; Transversality condition 0)()( =TkTλ or 0)()(lim
=
∞→tkt
tλ
Complimentary slackness: If 0)(0) =→>( TTk λ ; if 0)(0)
=→>( TkTλ
Sufficient condition for Maximum: Concavity conditions:
0)()(
2
2<
∂∂
tktH ; 0
)()(
2
2<
∂∂
tctH
Necessary conditions are sufficient (unique global maximum) if
functions f(.) and g(.) are strictly concave and tt ∀≥ 0)(λ ~~
Hamiltonian is strictly concave
8
-
Again: Current value vs. present value Hamiltonian Up to now:
Shadow price (λ) defined in terms of present value (evaluated at
initial stage) (.))((.) gteuH t λρ += − Alternatively: construct
Hamiltonian in current value terms: Shadow price (μ(t)) – price of
capital in terms of current utility:
(.)(.)~ guH μ+= HeH t ~ρ−= = (.))((.) gteue tt μρρ −− + )()( tet
t μλ ρ−=
FOC of the optimal control problem:
A*) 0)()(~
=∂∂
tctH since te
tctH
tctH ρ
)()(
)()(~
∂∂
=∂∂
B*) μρμ &−=∂∂
)()(~
tktH
Because: tetctH
tctH ρ−
∂∂
=∂∂
)()(~
)()( ;
μλ ρ te−= → μρμλ ρρ tt ee −− −= && →B)
C*) )()()( tk
ttH &=
∂∂
μ
D*) Transversality condition: 0)()(lim =−∞→
tket tt
ρμ
Arbitrage condition:
ρμμ
μ=+
∂∂ &kH /~
Marginal contribution of capital to utility + capital gain =
marginal return on current consumption
9
-
1.1.6 Proof of Optimality: Barro/Sala-i-Martin or Chiang Dynamic
problem with endpoint constraint 0)( ≥Tk Let )}(*{)},(*{ tctk be
the optimal path for state and control variable. Then, any small
perturbation of the optimal control path )}()(*{ tptc ε+ with the
corresponding perturbation of the path for the state variable
)}()(*{ tqtk ε+ and
)(( )(*) TkdTkTk ε+= should leave the maximum value L
unchanged: 0=∂∂
εL (given the endpoint constraint 0)( ≥Tk )
])}(.,(.))(.,),(.,[{)(., 0 dtkckHLT ελεεε &+∫=
),()()0()0( ελλ TkTk −+ + ),( εν Tk Take derivative wrt ε :
0)())((0 =∂∂
−+∫⎭⎬⎫
⎩⎨⎧
∂∂
+∂∂
=∂∂
ελν
ελ
εεTkTtdkHL T &
Since )()();()( TkdTktqkHtp
cHH =
∂∂
∂∂+
∂∂=
∂∂
εε, we get:
0)()]()([)()(0 =−+∫⎭⎬⎫
⎩⎨⎧
⎥⎦
⎤⎢⎣
⎡+
∂∂
+∂∂
=∂∂ TkdTTtdtq
kHtp
cHL T λνλ
ε&
Must hold for arbitrary small perturbations paths );();( tqtp
)(Tkd , so for each component, the Hamilton conditions have to
hold:
cH
∂∂ =0
0=+∂∂ λ&
cH
)()( TT λν = resp. CS: 0)()( =TkTλ or 0)()(lim =∞→
tktt
λ
10
-
1.1.7 Example: Optimal extraction path for exhaustible
resources
dtetCuV tρ−∞∫= 0 )]([ s.t. constraints:
a) Intertemporal resource constraint )()( tCtK −=& b) 0)0( 0
>= KK initial capital stock; c) ttK ∀≥ 0)( So: 00 )( KdttC ≤∫ ∞
Construct Hamiltonian: )()()]([ tCtetCuH t λρ −= − First order
conditions:
A) 0)()(
=∂∂
tCtH → tetCu ρλ −= ))(('
B) )()()( t
tKtH λ&−=
∂∂ → 0)( =tλ&
C) )()()( tK
ttH &=
∂∂
λ → )()( tCtK −=&
D) K(0)=K0 ; Transversality condition: 0)()(lim =
∞→tKt
tλ
B): Arbitrage equation: 0)( =tλ& present value constant →
Hotelling rule: current value must increase at discount rate:
ρμμμμ ρ=
)()(;)0()(
t=
tet t &
Reformulate conditions using the current value Hamiltonian:
)()()]([~ tCttCuH μ−= !
11
-
Example: Calculate optimal path for CES payoff function:
σσ
11
/11
1 −−
= tt cU ;
Example (σ=1):
)]([ln)]([ tCtCu = → tetC ρλ
−=1)( ;
λ1)0( =C
000111)( KdtedttC t ==∫=∫ ∞ −∞ ρλλ
ρ
(using the resource constraint)
Thus 0)0( KC ρ= and teKtC ρρ −= 0)(
][0Tt eKtd
; so that 1[ 00K)( TeTK
ρρρ −− =−= ∫ Transversality condition holds: 0)( >= λλ t
=∞→
)()(lim tKtt
λ
∞→t
lim ][lim1[ 000T
tT t eKtdeK ρρ λρλ −
∞→
− =− ∫ =0
Hotelling rule: Current value shadow price: tet ρμμ 0)( = ;
tett ρμλ −= )()(
12
-
1.2 The Basic Framework Modern Macro is about dynamic stochastic
general equilibrium in a set up with representative, optimising
agents in an economy with imperfect competition and possibly
incomplete asset markets. Here, in order to gain the basic
intuition for intertemporal optimisation, we start by reducing the
problem to the simplest case: We consider first a pure endowment
economy (1.2.1) with one good (consumption) and no uncertainty. We
allow for intertemporal trade on perfect capital markets via real
bonds with real interest rate rt. In this setup, we analyse
incentives for consumption smoothing. Later, we extend this setup
by introducing capital accumulation, uncertainty, government
activity and finally money (section 1.3). In the next step (1.2.2),
we look at the Ramsey growth model. It characterises the optimal
intertemporal path for consumption and capital formation {Ct*}
{kt*} (t=0,1,…,T) for a representative agent, starting with an
initial capital stock k0. We solve it as optimal control problem.
With a complete set of markets, a decentralised market economy can
implement the efficient allocation. Section 1.2.3 looks at some
applications, such as introducing government spending and taxes and
international trade of small countries. In both cases, we
concentrate on wealth constraints and try to figure out the
implications for sustainability of (public or international) debt.
Section 1.2.4 then looks at some basics of dynamic stochastic
optimisation. A promising route for modern macro is to integrate
macro and finance. Stochastic optimisation is the starting point
for this research area.
13
-
1.2.1 Consumption Smoothing: The Euler Equation The
representative agent maximizes
∑ == Ttt
tCUU 0 )( β s. t. per period budget constraint: tttttt CBrYBB
−+=− −+ 11
with B0 and r-1 given and the end point constraint BT+1≥ 0 (No
Ponzi Game constraint). rt-1: real return of asset Bt invested
during the last period t-1. Pure endowment economy: Path {Yt} is
given.
Define gt as the growth rate of endowment t
ttt Y
YYg −= +1 .
Start with a two period model U = U (C1) +ß U (C2). Budget
constraints per period: tttttt CBrYBB −+=− −+ 11 t=1,2 Two Periods:
Starting point (t=0) is 1; end point T is T=2! Choose {C1; C2} and
{B2; B3} optimally! Constraints: (1+ r0) B1 given; terminal
constraint B3≥0. A straightforward way to solve is to merge both
per period budget constraints into a life time constraint
(eliminating B2):
101
21
1
3
1
21 )1(111
Brr
YYr
Br
CC +++
+≤+
++
+ ,
and take also the terminal constraint B3≥ 0 into account.
14
-
But let us instead use the optimal control approach. First,
define the Hamiltonian in present value:
)()( 1 ttttttt
t CBrYCUH −++= −λβ Key FOC for intertemporal optimality:
A*) 0=∂∂
t
t
cH → t
t
t
cU λβ =
∂∂
B*) )( 111
1tttt
t
t rBH λλλ −−==
∂∂
+++
+ → tt
t
r+=+
111
λλ
C*) ttt
t BBH −=∂∂
+1λ → tttttt CBrYBB −+=− −+ 11
D*) Transversality condition: 01 =+TT Bλ → 032 =Bλ By combining
A*) and B*), we derive the Euler equation:
E*) ttc
tc
rCUCU
+=
++
11
)()(
11 1
ρ
Define the MRS as internal discount factor and 1/1+rt as
market discount factor t
tt rq
+=+ 1
11, . Then, we can rewrite E*
ttt
t
t
tc
tc
rq
CUCU
+===
+ +++
11
)()(
11
1,11
λλ
ρ
Using t
tt r+=+ 1
11 λλ , D* gives 01
13
1132,11 =+
= Br
Bq λλ .
Because 01
1 >∂∂
=cUλ this simplifies to: 0
11
31
32,1 =+= B
rBq
15
-
An equivalent formulation of the budget constraint is in terms
of wealth: If I plan to have wealth Wt+1 at the beginning of next
period t+1, I must spend today 11, ++ ttt Wq (Think of buying zero
bonds at the price ). Note: Wt= (1+rt-1) Bt . 1, +ttq
This gives: ttt
tt WYr
WC +≤+
+ +1
1
with the end point constraints 0; 311 ≥= WWW . The wealth
constraint per period can be formulated as:
ttttttttt WqCYqWW )1/1()(/1 1,1,1 −+−≤− +++ since tttttt WCYWq
+−≤++ 11, .
The Hamiltonian gives us as FOC:
11,11, ~
~)~~()1/1(~
−+−+ =→−−=−=∂
∂
t
tttttttt
t
t qqWH
λλλλλ
Note: ttttt q λλλ 1,1~
++ ==Transversality condition: 011, =++++ TtTtt Wq or 033,1 =Wq
The Euler equation E* characterises intertemporal demand. Aggregate
Equilibrium: For the whole economy, demand must be equal to
endowment. Equilibrium path depends on:
a) Preferences (discount rate, intertemporal elasticity of
substitution)
b) Technology (endowment; productivity growth → real
intertemporal exchange rate)
Let us concentrate on Euler equation )(11)( 1++
+= tc
ttc CU
rCUρ
Marginal utility out of consumption, adjusted for the relevant
rate of return, has to be equal across time Due to concavity
(u”(c)
-
Elasticity of marginal utility reflects the concavity of the
utility function: How does the slope of the indifference curve
change with a change in the consumption ratio ? 1/ +tt cc
)('/)('/
}/{)}('/)('{)(
1
1
1
1
+
+
+
+−=tt
tt
tt
tt
cucucc
ccdcucudcε
Inverse to the intertemporal elasticity of substitution σ
)(')("
)(1)(
cuccu
cc −==
σε (for continuous case)
Example: CES payoff function:
σσ
11
/11
1 −−
= tt cU ;
Euler equation:
tt cU ln:1For0 ==> σσ
tt
t
rC +=⎟⎟
C⎠
⎞⎜⎜⎝
⎛+11
ρ
−+
11
1
1 σ or ttt CrC
σ
ρ⎟⎠
⎞⎜⎝
⎛++
=+ 11
1 ;
tC is increasing [decreasing] for ρ>tr [ ρ
-
Equilibrium: Yt+1=(1+g) Yt;→ grnt σρ
1+= → Ct+1=(1+g) Ct =Yt+1 The natural rate of interest:
Equilibrium real interest rate is characterised by equilibrium
condition tYC tt ∀= (aggregate consumption = endowment). Using the
growth rate: ttt YgY )1(1 +=+ , we can write E* as:
( ) ⎟⎟⎠
⎞⎜⎜⎝
⎛++
=+ρ
σ111
1 nt
trg
Log-linear approximation: or )( ρσ −= ntt rg tn
t gr σρ1+= .
The natural rate of interest is higher ntra) The higher ρ:
impatience requires stronger incentives to
save! b) The higher the growth rate g (because of incentives
for
consumption smoothing, high growth has to be balanced by
stronger incentives to save!
c) The lower the elasticity of substitution (the more concave
the utility, the stronger incentives to smooth consumption for a
given growth rate)!
18
-
1.2.2 The Ramsey Model: Optimal Growth – Continuous Time Version
New twist: Include production and capital formation! Neoclassical
linear homogenous production function F(K,N) (CRS)
N(t): Labour supply at t; K(t): capital stock at t Law of
motion: )()())(),(()( tCtKtNtKFtK −−= δ&
)1,)()(())(),((
)(1))((
tNtKFtNtKF
tNtkf ==
0)(";0)(');( = kfkfkfy ;
kkfkfNkkfNkfNKF
NKFkfNNKF
kN
K
)()(),(
,()(),(
∂∂
+=
→= kfKkkfN k
)(')(
);(')()
−=
=∂∂
=
Inada conditions: ∞=
→)('lim
0kf
k; 0)('lim =
∞→kf
k
tneNtN 0)( = ; δ rate of depreciation Representative agent
maximises per capita consumption:
dtetcuV tρ−∞∫= 0 ))(( with )(/)()( tNtCtc =
Strictly concave payoff function: ; 0)(";0)(' cucuρ discount
factor s. t. constraints:
a) Intertemporal resource constraint: Capital stock per head
evolves according to: )()())(()( tckntkftk −+−= δ& ,
because: NNKKkk /// &&& −=
nNKknk or
kKKk −=⋅−⋅= /&& /& ; ckkfNK −−= δ)(/& b) Initial
capital stock )0(k given; c) Endpoint constraint 0)(lim ≥
∞→Tk
T
19
-
1.2.2.1 Hamiltonian: )]()()([)()]([: tcknkftetcuH t −+−+= −
δλρ
FOC:
A) 0)()(
=∂∂
tctH → tecu ρλ −= )('
B) )()()( t
tktH λ&−=
∂∂ → λδλ &−=+− )]()('[ nkf
C) )()()( tk
ttH &=
∂∂
λ → )()()()( tcknkftk −+−= δ&
D) 0)()(lim =∞→
tktt
λ → 0)())((('lim =−∞→
tketcu tt
ρ
FOC give 2 differential equations (characterising the evolution
of the state variable kk /& and of the shadow price λλ /&
); B) λδλ &−=+− )]()('[ nkf C) )()()()( tcknkftk −+−= δ&
Using A), B) can be transformed into a differential equation for
the evolution of control variable : cc /&
ttt ecucu
cu⎢
cecueccu ρρρ ρρλ −−− ⎥⎦⎤
⎣
⎡('
("−=−= )('
)))(')(" &&&
Because ccu
cuc)(")(')( −=σ , we get: λρ
σλ ⎥⎦
⎤⎢⎣⎡ +=−
cc&& 1 or
AB) )])('[)( nkfccc
−−−= δρσ&
C) )()()()( tcknkftk −+−= δ& 2 boundary conditions give a
determined solution for the 2 differential equations: (1) )0(k and
(2) the transversality condition: 0)())((('lim =−
∞→tketcu t
t
ρ
20
-
1.2.2.2 Steady State and Transition Dynamics Digression:
Stability analysis
How can we solve a differential equation )(yfy =& ? Is there
convergence to a steady state? Intuition by graphical analysis: Two
steady states. One stable, one unstable
Explicit analytical solution for linear differential
equations:
byay =−&
Steady state 0=y& : aby /−=
21
-
Solution procedure: a) Solve for the homogeneous differential
equation 0=− yay&
with ayy
=& stable system: a0
General solution: with taeAty =)( )()( tyaeAaty ta ==&
Particular solution: taeAty )0()( = b) Take some particular
solution for the non-homogenous
differential equation: For instance, take steady state 0=y&
with aby /−=
General solution for the complete system: abeAty ta −=)(
The definite solution is determined by a specific boundary
condition: y(0) → abyA += )0(
→ Determines the unique solution: abe
abyty ta −+= ])0([)(
22
-
Steady State in the Ramsey Model: The equilibrium path is
characterised by the 3 conditions AB), C) and D). In a Steady
State:
0)])('[)( =−−−= nkfccc δρσ&
Modified golden rule: nkf ++= δρ*)(' 0)()()()( =−+−= tcknkftk
δ& or **)(*)( cknkf ++= δ
For the steady state {c*; k*}, the transversality condition is
fulfilled since k=k* and 0*)('))((('
∞→
−− →→T
TT ecueTcu ρρ .
Is the steady state stable? Graphical analysis:
23
-
Saddle point stability:
*)('0 kknkfcc
>≤
⇔++<≥
⇔<≥ δρ&
knkftckk )()()(0 +−
>≤
⇔<≥ δ
&
For any initial k(0), exactly one unique path of convergence to
the steady state {c*; k*} starting with initial consumption
))0((ˆ kc - all other paths are diverging. For ))0((ˆ)0( kcc
> : Savings too low → eventually cross →
at some finite T: k(T)=0 → jump in consumption to c=0 0)(
=tk&
If ))0((ˆ)0( kcc < : Excessive Savings → Overaccumulation →
in finite time: c=0 → violation of transversality condition.
Comparative statics: What happens, if some exogenous parameter
changes? What is the impact of such a shock?
↓→↑→↑ **)(' kkfρ Investment reduced; consumption goes up
(consumption path tilted more towards present consumption).
24
-
No further adjustment in steady state: gknkftc −+−= *)(*)()(*
δ
Increase in real rate of interest: consumption tilted more
towards future consumption
25
-
1.2.2.3 The Decentralised Market Economy The competitive
equilibrium is efficient in an economy with infinitely lived
agents. Before reading the next pages, try first to show this
yourself. Assume that some firms own the capital stock k(0),
financed by issuing bonds to households. Bonds b(t) pay a return
r(t) at t. The initial wealth of households is the value of the
firm b(0). Labour is supplied inelastically, household’s wealth per
capita b(t) evolves according to:
)()()()()()( tbntctbtrtwtb −−+=& Formulate the firm’s and
household’s optimisation problem. Then, characterize the
intertemporal price path. Note that in a one good economy, only the
intertemporal price (the rate of return on assets, the interest
rate r(t)) has to be determined. Use the Hamiltonian approach to
characterise the FOC Finally, show that the FOC for the optimum are
equivalent to the conditions of the Ramsey problem. Hints: 1)
Define })]([1)(
0νν dr
ttr
t
∫⋅= as average interest rate up to t giving
the present value factor: ttre ⋅− )( 2) Impose the No
Ponzi-Game-constraint:
→ 0)(lim ])([ ≥⋅−−∞→
TnTr
TeTb
Transversality condition: this inequality must be binding for
the optimal solution
3) The firm’s problem is not really dynamic. Value of the firm:
b(0)= k(0)!
26
-
1) Firms In the absence of adjustment costs, firms face not
really a dynamic problem. Firms rent labour and capital at each
point in time. Assume some firms own the capital stock, financed by
issuing bonds. Assume CRS (Constant returns to scale).
)()()()()())(()()(),(
tNtwtKtrtKtkftNMaxtKtN
−−− δ
FOC: )())((' trtkf += δ or δ−= ))((')( tkftr ; )()())(('))((
twtktkftkf =− Zero Profits: Labour income: )()( tNtw ; Capital
income per worker: )()( tktr . With CRS, the value of the firm is
b(0)= k(0)! 2) Households: The representative household starts with
initial financial wealth b0 and earns income from interest on bonds
and labour income w(t): (→ human wealth). Assets per capita for the
household evolve according to:
)(])([)()()()()()()()( tbntrtctwtctbntbtrtwtb −+−=−−+=&
Let us follow the standard Hamiltonian approach:
Max dtetcudtetNtCuV tt ρρ −∞−∞ ∫=∫ ⎟
⎠
⎞⎜⎝
⎛= 00 ))(()(
)(
s.t. (1) the law of motion
)(])([)()()()()()()()( tbntrtctwtctbntbtrtwtb −+−=−−+=& (2)
The No Ponzi-Game-constraint
0})({lim}])([exp[)({lim ])([0 ≥=∫ −−⋅⋅−−
∞→∞→
tntr
t
t
tetbdnrtb νν
(2b) Initial financial wealth: 0)0( bb =
27
-
Hamiltonian)](])([)()([)()]([: tbntrtctwtetcuH t −+−+= − λρ
FOC:
mA) 0)()(
=∂∂
tctH → tecu ρλ −= )('
mB) )()()( t
tbtH λ&−=
∂∂ → λλ &−=− )])([ ntr
mC) )()()( tb
ttH &=
∂∂
λ → )(])([)()()( tbntrtctwtb −+−=&
mD) 0)()(lim =∞→
tbtt
λ → 0)())((('lim =−∞→
tbetcu tt
ρ
Compare to the planning solution: The planning problem was:
Max dtetcuV tρ−∞∫= 0 ))(( s.t. a) Law of motion: )()()())(()(
tkntktctkftk −−−= δ& b) Endpoint constraint 0})({lim ≥
∞→tk
t
c) Initial capital stock: 0)0( kk = with Hamiltonian:
)]()()())(([)()]([: tkntktctkftetcuH t −−−+= − δλρ FOC:
A) 0)()(
=∂∂
tctH → tecu ρλ −= )('
B) )()()( t
tktH λ&−=
∂∂ → λδλ &−=−− ])('[ nkf
C) )()()( tk
ttH &=
∂∂
λ → )()()())(()( tkntktctkftk −−−= δ&
D) 0)()(lim =∞→
tktt
λ → 0)())((('lim =−∞→
tketcu tt
ρ
So ntkf −−=− δλλ ))(('&
c
c
UU&&
−=− ρλλ
28
-
Equivalence: mA) is the same as in the planning solution. mB) is
equivalent to B) because in a market economy,
δ−= ))((')( tkftr mC) is equivalent to C) )()()()( tcknkftk −+−=
δ& because
δ−= ))((')( tkftr(')())((and b(t)=k(t) and under CRS
)())( tktkftwtkf += . But more generally, even in the absence of
CRS, the value of
the firm includes all profits, so we get the analogous result
also in that case.
mD) Transversality condition 0)()(lim =
∞→tkt
tλ is equivalent
to the No Ponzi game constraint being binding: 0})({lim ])([
=⋅−−
∞→
tntr
tetb , because b(t)=k(t) and
δ−= ))((')( tkftr , so from B): λλ &−=− ]n)([ tr or tntrt r
⋅−− ])([0 )([ ν edn ⋅=− )0(}] λνt ∫−⋅= exp[)0()( λλ
29
-
No Ponzi Game Constraint and Transversality Condition From the
(flow) budget constraint, integration gives evolution of wealth
across time:
tddnrtctwdnrbTb TtTT ]])([[exp)]()([]])([[exp)0()( 00 νννν ∫∫∫
−−+−=
Notation: ttre )(− = νν drt∫− 0 )(exp[ with νν drttr t∫⋅= 0
)(
1)(
So tdetctwebTb tTntrTTnTr )(])([0])([ )]()([)0()( −⋅−⋅− −∫+=
The intertemporal wealth constraint is: tdetwbeTbtdetc
tntrTTnTrtntrT ⋅−−⋅−−⋅−− ∫+=+∫ ])([0
])([])([0 )()0()()(
or for T→∞ tdetwbeTbtdetc tntrTnTr
Ttntr ⋅−−∞⋅−−
∞→
⋅−−∞ ∫∫ +=+ ])([0])([])([
0 )()0()(lim)(
No Ponzi-Game-constraint:
Finite Time: Impose condition 0)( ≥Tb for representative agent:
Intuition: Otherwise (that is if for some j 0)( Tb at the end of
the world. i
No Ponzi-Game-constraint with infinite horizon ( ∞→T ):
0})({lim}])([exp[)({lim ])([0 ≥=∫ −−⋅
⋅−−
∞→∞→
tntr
t
t
tetbdnrtb νν
Present value of assets must be asymptotically non-negative!
Motivation: lack of counterparty - no one is willing to hold assets
with positive present value forever. Otherwise (if constraint were
not imposed): it would be optimal to go into infinite debt at the
starting period (cannot be an equilibrium).
But from the Transversality condition (as part of the optimality
conditions): we know that this inequality must be binding for the
optimal path. 30
-
1.2.2.4 OLG Model: Dynamic Inefficiency Simplest case: Agents
live two periods; income only in first period U = U (C1t) +ß U
(C2t+1)
s. t. wealth constraint: C1t +1/(1+rt) C2t+1= wt
FOC : ttc
tc
rCUCU
++
=+11
)()(
1
12 ρ or ttt CrC 112 1
1 σ
ρ⎟⎠
⎞⎜⎝
⎛++
=+
Consider σ=1:
tt wC ρρ
++
=21
1
ttt wwrs ρ+=
21)(
; t
tt w
rCρ+
+=+ 2
112
Population Growth: tt NnN )1(1 +=+ Evolution of capital
stock:
12111 ),( −−+ −−+=−=− ttttttttttttt NCNCKrNwCNKFKK = 1211 ][ −−
−+− ttttttt NCKrNCw ; Consumption of old: ttttt KrKNC 112 −− +=
ttttttt wrsNCwNK )()( 11 =−=+ or in per capita terms: tttttt
t
tt wrsn
Cwn
CwNNk )(
11)(
11)( 11
11 +
=−+
=−=+
+
= [ ])(')()('(1
11 tttt kfkkfkfsn
−+ +
For σ=1 ρ+
=+ 21)('( 1tkfs ; for
αtt kkf =)( :
αα tt kw )1( −= so
ααρ tt
kn
k )1(2
11
11 −++
=+ → α
αρ
−⎟⎠
⎞⎜⎝
⎛ −++
=1
1
)1(2
11
1*n
k
)2()1(1
*)(*)(' 1 ρα
αα α ++−
== − nkfkf
Compare with golden rule path: nkf g =)(' ! Overaccumulation if
α is small: nkf
-
Dynamic Inefficiency ~ Reason? Market failure: missing markets
Future generations do not participate in active markets today
Incomplete markets Mathematical joke: Hotel infinity: one missing
bed nor problem, just ask all to move one bed ahead Paul Samuelson:
OLG model as basis for: a) social security system Intergenerational
Transfer as free lunch b) Introduce Money (Sargent/ Wallace): OLG
model as microfoundation for monetary theory (fad) no room for
money as soon as there is storage; dominated asset! (You need
transaction role for money)
32
-
1.2.2.5 Financial Wealth/ Asset Pricing: Value of firms
(shares): Basics of asset pricing (see Obstfeld/Rogoff 100ff;
Appendix 2B, 121) The value of a firm { } depends on the dividend
stream { } and the terminal value. Each period, the household
decides about the number of shares held until next period. She
purchases shares of the firm at the value Vt-1. At time t, these
shares yield a dividend dt and (after dividend payments, before
consumption) have the value Vt.
tV td
ts
At t, additional shares can be sold or bought at the current
value Vt. Given the dividend stream, how can we calculate the value
of the firm? Intuition: Intertemporal evolution of the firm’s value
should follow the intertemporal market prices.
Define 1,11 11
)()(
+++ =
+== tt
tt
t
tc
tc qrCu
Cuλ
λβ as the market
discount factor. Then, the asset evaluation equation should be:
1,11 )( +++ += ttttt qdVV = TttTittTi it qVqd ++= + +∑ ,,1
Let us derive this value function explicitly. With other income
and spending Nwt
−t Vs, available consumption in each period is:2
+1)( ++= ttttt sVdC Nwt So })({( 10 NwssVsduU ttttttt
t +−+= +∞=∑ β For choosing the number of shares held optimally,
we must have as FOC:
ts
][][)(')('
)('])[('
111
111
ttt
ttt
t
tt
ttt
tttt
VdVdCu
CuV
VCuVdCu
+=+=→
=+
−−−
−−−
λλβ
ββ
2 Budget constraint: tttttt CsVdsV −+=+ )(1 Nwt+ . Use the
standard Lagrangian:
]Nwt+)[( 1sVCsVd ttttttt −−+ +λ + ]111 NwsVC tttt −−−) 1st−[(
111 Vd ttt −−− +−−+λ . Alternatively, try Hamiltonian approach by
reformulating: tt ss −+1
33
-
By iterated substitution, the value of the firm is the present
value of the dividend stream (discounted by the marginal rate of
substitution between t and today [at each state s]):
01
00 λ
λλλ T
TTt
tt VdV +∑= = = = TTtTt t qVqd ,0,01 +∑ =
)('
)(')('
)('
001 Cu
CuVCu
Cud TT
Tt
tTt t
ββ +∑ =
Transversality condition: 0)(')('lim
0=
∞→ CuCuV TTTT
β , so
∑= ∞=10
0 tt
tdV λλ
tTt t qd ,01∑= =
Extension: Introduce an additional international asset with
constant interest rate r: With 2 assets and a constant rate of
interest r, real wealth is:
tttt BsVW += −1
tttttttttttttt CNwsVVsdrBsVsVBB −+−++=−+− −−++ )( 1111
})()1{( 110 NwssVsdBBruU tttttttttt +−++−+= −+∞=∑ β
FOC: )(')()(' 111 +++ += ttttt CudVCuV β and
)(')1()(' 1++= tt CurCu β thus:
)(1
11 ttt Vdr
V ++
=− or
∑+
++
=++
= =Tt TTtt Vr
dr
Vdr
V 1110 )1(1
)1(1)(
11
Using the transversality condition: ∑+
→ ∞=10 )1(1
t tt drV
34
-
Remarks: 1) Under CRS: Value of capital stock ~ Value of Firm:
tt VK =
)(),( 1 ttttt KKNwNKFd −−−= + , so firms maximise at t=0:
)}(),({)1(
11000 tttt tt KKNwNKFr
Vd −−−∑+
=+ +∞=
with rFwFtKtN == ; .
Under constant returns to scale tKN KFNFF t+=(.) , so
tt KrNwF =−(.) So there are zero profits and 00000 )1( rKKKrdV
+=+=+
2) No Bubble condition: In equilibrium, it must be true that
. 0)('lim =∞→
TT
TTCuV β
with constant r, equivalent to 0)1(
1lim =+∞→ TTT r
V because
ttt
rCuCu
)1(1
)(')('
0 +=β and 0)(' 0 >Cu
Proof: Suppose . Then, agents would
be better off (increase present value of consumption) by
short-selling shares (maintain forever a constant negative value of
the asset). But as long as , no
one would be willing to hold the positive counterpart (all
agents would like to do the same – so that cannot be an
equilibrium). Nobody wants to be the victim of a Ponzi game!! So we
impose the no bubble condition:
. Optimality with free disposal implies
that inequality is binding: Transversality condition:
0)('lim >∞→
TT
TTCuV β
lim∞→T
0) ≤
0) =
0)(' >TT
T CuV β
('lim∞→
TT
TTCuV β
('lim∞→
TT
TTCuV β
35
-
Value of capital in the case of adjustment costs: Tobin’s q See:
Barro/Sala-i-Martin, section 3.6 and Romer, section 8.2 In the
presence of adjustment costs, the value of capital may deviate from
replacement cost. Solve: Max
dtetKtItItNtwtNtKFV ttr )(0 )()(1)()()()](),([ −∞∫
⎭⎬⎫
⎩⎨⎧
⎥⎦
⎤⎢⎣
⎡⎟⎠
⎞⎜⎝
⎛Φ+−−=
s.t. )()()( tKtItK δ−=& ; 0)(";0)(';0)0( ≥⋅Φ>⋅Φ=Φ Shadow
price: q(t)=V/K is the price of the capital stock at period t in
current value (Market value of a firm/ capital stock). (Check with
present value Hamiltonian: ttretqt )()()( −⋅=λ !) Current value
Hamiltonian:
)]()([)()()(1)()()()](),([)(~ tKtItqtKtItItNtwtNtKFtH δ−+⎥
⎦
⎤⎢⎣
⎡⎟⎠
⎞⎜⎝
⎛Φ+−−=
FOC:
0)()(~
=∂∂
tNtH or 0)](
)()([ =−
∂∂ tw
tNtF
0)()(~
=∂∂
tItH or
)()(
)()('
)()(1)(
tKtI
tKtI
tKtItq ⎟
⎠
⎞⎜⎝
⎛Φ+⎟
⎠
⎞⎜⎝
⎛Φ+=
)()()()()(~ tqtrtq
tKtH
+−=∂∂
& or
)()())(()()(
)()('
)()(
2
2tqtqtr
tKtI
tKtI
tKtF
&−=+−⎥⎦
⎤⎢⎣
⎡⎟⎠
⎞⎜⎝
⎛Φ+
∂∂ δ
or: ⎥⎦
⎤⎢⎣
⎡⎟⎠
⎞⎜⎝
⎛Φ+
∂∂
−+= 22
)()(
)()('
)()()())(()(
tKtI
tKtI
tKtFtqtrtq δ&
36
-
⎥⎦
⎤⎢⎣
⎡⎟⎠
⎞⎜⎝
⎛Φ+
∂∂
−+= 22
)()(
)()('
)()(
)(1))((
)()(
tKtI
tKtI
tKtF
tqtr
tqtq δ&
Return on existing capital, deflated by cost of capital q(t): F’
+ marginal reduction in adjustment cost with rising K Without
adjustment costs: δ+=== rFtqtq ';0)(;1)( & Transversality
constraint (motivated by the no bubble constraint): 0)()(lim )(
=−
∞→
ttrt
etKtq
Example: Linear adjustment costs )()(
2)()(
tKtIb
tKtI =⎟
⎠
⎞⎜⎝
⎛Φ
FOC: )()(1)(tKtIbtq +=
btq
tKtI 1)(
)()( −
=→
)(1)()( tKbtqtK ⎟
⎠⎞
⎜⎝⎛ −
−= δ&
⎥⎦
⎤⎢⎣
⎡ −+−+=
btqtKFtqtrtq2
]1)([))((')())(()(2
δ&
Convergence of shadow price of capital q(t) to q*=1+δ b? Saddle
path stability:
37
-
1.2.3 Applications: 1.2.3.1 Permanent income hypothesis Milton
Friedman argued that a representative agent consumes a constant
proportion of his permanent income. Under what conditions will this
be optimal? Using the transversality condition, the intertemporal
wealth constraint for a representative agent is:
tdetwbtdetc tntrtntr ⋅−−∞⋅−−∞ ∫∫ += ])([0])([
0 )()0()( Using the optimality conditions derived in 1.2.1, we
can characterise the optimal consumption path in a straightforward
way:
Because of )][)( nrccc
−−= ρσ& , we get via integration:
tntrtcectc ⋅−−= ])([))(()0()( ρσ Given this path, c(0) is
determined by the intertemporal wealth constraint. For constant σ
we can solve explicitly for the consumption path:
dtectdec tntrtntrntr ⋅−−−∞⋅−−−−−∞ ∫∫ = }]1[])({[0)])(()([
0 )0()0(ρσσρσ
)0()0( HWb +=
A) Special case σ=1: )]0()0([)0( HWbc += ρ
B) If the real rate of interest r is constant, this gives as
optimal initial consumption level:
)]0()0([])([)0( HWbnrnrc +−−−−= ρσ Obviously, c(t)= c(0) only if
0=−− nr ρ :
)0()0()( ][ cectc tnr == ⋅−−ρσ
)0()0()0()0( ][0 Htnr Wb
nrctdec +=
−=∫ ⋅−−
∞
or pH wbnrWbnrc +−=+−= )0()()]0()0()[()0( Note: !ρ=− nr (steady
state analysis) Modern Version: Consumption growth follows random
walk 38
-
1.2.3.2 Government Spending and Ricardian equivalence 1.2.3.2.1
Basic Idea: Two Period Framework What changes if we introduce
government activities? Take the path of government spending and
taxes as given. The government can issue bonds to finance spending
via debt.
2121 ;;; TTGG
The government budget constraint per period is: 10211 )1( BrBTG
+−+= ; 21322 )1( BrBTG +−+=
Consider a pure endowment economy. The initial government debt
has to be equal to the initial wealth 10 )1( Br+ of private agents
who initially bought bonds . Since for private agents terminal
condition is binding, we have to impose the No Ponzi game
constraint for the government:
[resulting in the transversality condition ].
1B03 ≥B
03 ≤B 03 =B
)1(1)1(
1 13
1
21́10
1
21 r
Br
TTBrr
GG+
++
+=+++
+
The budget constraint for private agents is now:
101
2211
1
3
1
21 )1(111
BrrTYTY
rB
rCC ++
+−
+−=+
++
+
Ricardian equivalence: Government debt is no net wealth for
private agents! From the government budget constraint, we can
substitute for taxes in the private agent constraint:
1
2211
1
21 11 r
GYGYr
CC+−
+−=+
+
So the modified equilibrium conditions are now: 222111 ; GYCGYC
−=−= , and the natural real interest rate is
now determined by:
11,
1
11
22
11
)()(
11
rq
GYuGYu
ttt
t
c
c
+===
−−
+ ++
λλ
ρ
39
-
1.2.3.2.2 Infinite Horizon Growth Model Assume that the
representative household owns the capital stock )(tk and also holds
government bonds )(tb [both defined as per capita]. This defines
wealth. She has to pay taxes and may appreciate g(t). The growth
rate of labour is n. The household’s problem (with the capital
stock rented out to the firms) is:
Max dtetgtcuV tρ−∞∫= 0 ))(),(( s.t. law of motion a) Total
wealth Ω(t)=k(t)+b(t) evolves according to: with )()()( tbtkt
&&& +=Ω
)()()()(])([))(()( ttkntctbntrtkft τ−−−−+=Ω& b) No
Ponzi-Game-constraint 0})({lim ])([ ≥Ω ⋅−−
∞→
tntr
tet
c) Initial wealth out of government bonds: )0(bd) Initial
capital stock: )0(k e) Assume: ))(),(( tgtcu is additive
separable
Hamiltonian: )()()]([: ttetcuH t Ω+= − &λρ From Lagrangian:
)()()]([:)( ttetcuHL t Ω+= − &λρ - )]([)( tt Ωλ& So:
)]()()(])([)())(([)()]([: ttkntbntrtctkftetcuH t τλρ −−−+−+=
−
FOC:
A) 0)()(
=∂∂
tctH → tecu ρλ −= )('
B) )()()(
)()(
)()( t
tbtH
tktH
ttH λ&−=
∂∂
=∂∂
=Ω∂
∂
→ λλ &−=− ])([ ntr and λλ &−=− ])('[ nkf
40
-
C) )()()()()()( tktbtt
ttH &&& +=Ω==
∂∂ ω
λ
)()()(])([)())(()()()( ttkntbntrtctkftktbt τ−−−+−=+=Ω
&&&
D1) 0)()(lim =∞→
tbtt
λ → 0)())((('lim =−∞→
tbetcu tt
ρ
D2) 0)()(lim =∞→
tktt
λ → 0)())((('lim =−∞→
tketcu tt
ρ
So ntrntkf −=−=− )())(('λλ&
Equal return on capital and government bonds: )())((' trtkf
=
Rate of return on consumption:c
c
UU&&
−=− ρλλ
With fixed labour supply and utility additive separable in g(t)
( ) : 0=cgU cUU ccc && =
D1) is equivalent to: 0})({lim ])([ =⋅−−∞→
tntr
tetb
D2) is equivalent to: 0})({lim ])([ =⋅−−∞→
tntr
tetk
0)0(;)0(])([exp[)0()( ])([0 >=−−= ⋅−−∫ λλννλλ tntrt ednrt The
path of spending and taxation cannot be arbitrary: The government
has to take into account its budget constraint. Its flow constraint
is:
)()(])([)()( ttbntrtgtb τ−−+=& Starting with some initial
level of government debt b(0), debt evolves according to:
dtettgebTb tTntrTTnTr )(])([0])([ )]()([)0()( −−− ∫ −+= τ
or:
TntrT
tntrT ebdtetgtb ])([])([0 )]()([)0(−−−− +−= ∫ τ
41
-
When the transversality condition for households holds, the
government is subject to the “No Ponzi game condition:” 0})({lim
])([ ≤⋅−−
∞→
tntr
tetb
The transversality condition gives as intertemporal wealth
constraint for the government: (GW) dtetgtb tntr ])([0
)]()([)0(
−−∞ −= ∫ τ Independent of the timing of taxes and government
spending, the present value of net earnings (discounted tax
earnings minus government spending) must equal present outstanding
debt. Ricardian equivalence: How does government debt policy affect
private wealth? Using the transversality condition, the household’s
wealth constraint is:
tdettkfbktdetc tntrtntr ⋅−−∞⋅−−∞ ∫ −++=∫ ])([0])([
0 )]())(([)0()0()( τ Government bonds are not net wealth: Using
(GW), we can transform the household’s wealth constraint into:
tdetgtkfktdetc tntrtntr ⋅−−∞⋅−−∞ ∫ −+=∫ ])([0])([
0 )]())(([)0()( So the household’s wealth constraint is
independent of the tax and debt path. Intuition: Higher government
debt results in an equivalent increase in private savings in order
to offset future tax obligations. This result depends on strong
assumptions: No imperfections of capital markets, private equal to
social discount rate…!
Credit constrained agents: liquidity constraints OLG structure:
Distortionary taxes: Incentives for tax smoothing Non-optimising
agents (Keynesian type agents; rule of thumb ~ Akerlof: inclusion
of those norms (AER presidential address 2007).
42
-
1.2.3.2.3 Sustainability of Government Debt In the last section,
we characterised the government wealth constraint. Will a
government with a constant deficit run into solvency problems? Does
it make sense to impose some upper limit to the deficit ratio such
as in the Maastricht criteria? We have to distinguish between
primary deficit and total deficit. Consider as simplest case an
economy with constant growth rate y and constant real rate of
interest r. We assume the interest rate exceeds the growth rate:
r>y (only this assumption is essential!) When will the
government stay solvent, starting with some initial debt
(outstanding bonds)
>0? (Note: here b(t)=B(t)/Y(t); issuing bonds b>0 is
equivalent to debt!)
)0(b
With Pd)t
as primary deficit, the debt ratio evolves according to: b
)()()()()()()(( tbyrtdtbyrttg P −+=−+−= τ& giving the
differential equation:
dtettgebTb TtyrTTyr )()(0)(0 )]()([)(−−−− −+= ∫ τ or
TyrT
tyrT ebdtetgtb )()(00 )]()([−−−− +−= ∫ τ
0lim )( = Imposing the
transversality condition −−∞→
TyrTT
eb
dtetdb tyr
gives (GW)
P)(
00 )(−−∞∫ −= etgt0 )]()([ dttyr )( −−∞ −= ∫ τ = yr
g−−τ
The present value (discounted by the effective rate r-y) of
ratio of permanent future primary surplus to GDP has to be equal to
the current level of debt /GDP. Consider a path with a constant
primary deficit rate:
τ−= gpd such that: )()()( tbyrpdtb −+=& . Solution to this
linear differential equation: cbab =+& Homogenous part: taeAtb
−=)( ; given 0b
Particular solution: acbp =
43
-
Definite solution: aceacbtb ta /]/[)( 0 +−=−
Since a=-(r-y)0, we get )/()]/([)( )(0 yrpdeyrpdbtb
tyr −−−+= − Obviously, the government debt ratio is unstable
unless
)/(0 yrpdb −−= But: temporary deviations - ])([)( pdtpdtb
−=& - do no harm if offset later by additional surplus:3
dtepdtpdyr
pdTb TtyrT )()(0´ ])([)(−−−−+
−−= ∫
The government stays solvent if minmaxmax gsppd −=≤− τ with 1max
g as minimum sustainable spending rate. Since these rates are
somewhat arbitrary, as pragmatic approach, calculate as sustainable
tax rate the average tax rate which would be needed to finance
annuity value of future expected spending and transfers, plus the
real interest rate burden on current debt:
])()[( 0)(
0 bdtetgyr tyr +−= −−∞∫τ If ττ < , taxes have to be raised or
spending has to be cut when debt should be sustainable. The
magnitude ττ − gives a measure of the adjustment needed! How does
delay of adjustment until time t affect the size of needed policy
action? The timing of adjustment is given by
differentiating the equation above: )()( τττ −= − ttyre
tdd
Key ingredients: 1) The real interest rate exceeds growth rate:
r>y 2) Transversality condition holds: 0lim )( =−−
∞→
TyrTT
eb 3 Transversality condition 0lim )( =−−
∞→
TyrTT
eb implies:
0])([ )(0´ =− −−∞∫ dtedtd tyrpp 44
-
2) Maastricht treaty: constant total budget deficit rate The
Maastricht criteria impose a limit for the nominal debt ratio b≤0,6
and for the total nominal deficit ratio d≤ 0,03. So now we have to
analyse an economy with nominal debt and inflation. pY is nominal
GDP. We assume a constant rate of inflation π and the Fisher
equation i= r+π. We also include seigniorage μk in the government
wealth constraint (see section 1.3 for details). How will the debt
ratio evolve when the total budget deficit ratio (including debt
service) stays constant? Will the debt ratio converge to some
steady state?
Let constant==+−−= dbikgYp
B μτ&
Then bydb )( π+−=& . Note: d is total deficit (including
interest payments)!
Solution of the Differential equation bydb )( +−=•
π for any initial level : 0b
yde
ydbdtedebTb TyTtyTTy
++⎥
⎦
⎤⎢⎣
⎡+
−=+= +−−++− ∫ππ
πππ )(0
)()(0´
)(0 ][)(
So we have y
dTbT +
=∞→ π
)(lim , converging to the steady state:
ydb+
=π
*
Maastricht criteria: For d=0,03; y=0,03 and 02.0=π : b converges
to b=0,6 The lower y and π , the higher the steady state ratio b.
Implies: higher tax burden on economy (depending on real interest
rate r-y). The debt limit is feasible as long as
yrgbb
−−
=< minmaxτ ( 0;1 minmax >< gτ )
45
-
1.2.3.3 Current Account Deficits: Optimising Countries?
Frequently, economists analyse dynamics with constant interest rate
to gain a better intuition about key issues, such as solvency
constraints and sustainability of debt. In this section, we
consider simple examples of a pure endowment economy with infinite
horizon to get a feeling for solvency constraints and for the
implications of transversality conditions. We pose the following
questions:
1) Can we have an economy with optimising heterogeneous agents,
but a constant real rate of interest without violation of the
transversality condition? (Example: Countries with different time
preferences in open economies)?
2) Sustainability of Current account deficits: Under what
conditions is the debt dynamic of a small country sustainable?
Assume the country has an exogenous constant growth rate y with:
ytYtY =)(/)(& . Ignore capital formation and set n=0.
Preferences are characterised by a constant elasticity of
substitution σ. Let b(0)
-
We analyse the debt dynamic in a small country with
preferences
dtetcuV tρ−∞∫= 0 ))(( with σσ
11)(1)(
/11
−
−
= tctU
Initial wealth: b(0)= B(0)/ Y(0) a) Characterise the law of
motion for wealth/GDP:
Show that b(t= B(t)/Y(t) evolves according to:
)()()(1)()(/)]()()([)( tbyrtctbytYtBrtCtYtb −+−=−+−=&
b) Characterise FOC using the Hamiltonian! Show that
Consumption path for constant σ is characterised by: treCtC ⋅−=
][)0()( ρσ Under what conditions is the
problem well-defined? c) Characterize the autarchy solution. d)
Assume the country has initial debt b(0)= B(0)/ Y(0)n! For n−−−−
nrnr ρσ or ρ
σ−−
−<
nrnr
See also next section on sustainability analysis!
47
-
1.2.3.3.2 Sustainability analysis of current account deficits.
Using the transversality condition 0})({lim =⋅ ⋅−
∞→
Tr
TeTb , the
intertemporal wealth constraint is: =+= ⋅−∞⋅−∞ ∫∫ tdetYBtdetC
trtr 00 )()0()([
yrYBtdeYB tyr
−+=+ ⋅−−∞∫
)0()0()0()0( )(0
The consumption path for constant σ is characterised by: treCtC
⋅−= ][)0()( ρσ ;
)()0()0( 0
)]([
ρσρσ
−−=∫∞ ⋅−−−
rrCtdeC trr
The problem is well-defined for ρ
σρσ−
−−r
rrr ;0)(
(see 1.2.3.3.1) Otherwise there would be no convergence of
consumption. So we get:
=+−
−−= )]0()0([)]([)0( Byr
YrrC ρσ
)0()]([)0()( BrrYyr
rr ρσρσ −−+−
−−
tyrtyr eyr
rreYC
tYtC ⋅−−⋅−−
−−−
== }][{}][{)(
)0()0(
)()( ρσρσ ρσ
>∞
=−→
<
∞→yrforconst
tYtC
t][
)()(
0
ρσ
48
-
A) In the case of autarchy, we must have:
yrσ
ρ 1+= ; B(0)=0; tyty eCeYtYtC ⋅⋅ === )0()0()()(
For yrσ
ρ 1+= : consttYtC
=)()( =1+(r-y)b0; consttY
tB=
)()( ;
)0()0(
)()(
)()(lim
YB
tYtB
tYtB
t==
∞→= b0
B) If yrσ
ρ 1+> , the consumption path is tilted towards the
future. There is no convergence towards a steady state, but the
value of consumption will be bounded for σ : ∞→)()(
tYtC ; ∞→
)()(
tYtB
49
-
In that case, consumption grows faster than income, but
initially the country builds up assets. Ratio of net foreign
assets to GDP shoots off towards infinity: ∞=∞→ )(
)(limtytb
t.
Note: no stable steady state, but the transversality condition
holds even in that case as long as )( ρσ −> rr (compare
Obstfeld/Rogoff, exercise 2.7)! C) Consider now the case that the
world interest rate is less
than the autarchy rate yrσ
ρ 1+< . That is the interesting case
for solvency issues, since in that case, the consumption path of
a small open economy is tilted towards the present relative to the
autarchy case (C(0)>Ca(0). So, initially, there is a current
account deficit B(t)r C(t)- Y(t) CA(t)(t)B +==&
B(t)r ;
CA(t) C(t)- Y(t))( −==tTB . What will happen to a country
borrowing at the world rate r across time? Where does the debt/GDP
ratio converge to?
50
-
Steady state: B(t)g CA(t)(t)Blim ==∞→&
t; B(t)r)-(yTB(t)lim =
∞→t
In the long run: yrtY
tBt −
−=∞→
1)()(lim ; 1
)()(lim =
∞→ tYtTB
t; 0
)()(lim =
∞→ tYtC
t;
Limit of debt/ GDP ratio converges to the entire present value
of current and future output (debt has to be repaid out of future
production).
Proof:
)()(
)()(
)()()(;
)()(
)()(
)()(;
)()()(
tYtB
tYtY
tYtBtb
tYtY
tBtB
tbtb
tYtBtb
&&&
&&&−=−== ; Thus:
)()()(
)()(1
)()(
)()()()()(
tYtByr
tYtC
tYtBy
tYtBrtCtYtb −+−=−+−=& .
In steady state: or 0)( =tb& ⎟⎠
⎞⎜⎝
⎛−
−−=
∞→ )()(11
)()(lim
tYtC
yrtYtB
t;
For yrσ
ρ 1+< : 0)()(lim =
∞→ tYtC
t;
yrtYtB
t −−=
∞→
1)()(lim
Crucial issue: Is the limit 0)()(lim =
∞→ tYtC
t sustainable?
51
-
Notes: a) Consumption does not need to converge to 0 since Y
grows at rate y. But 0)()(lim =
∞→ tYtC
t for growth rate of
consumption: yr y must hold (otherwise
infinite consumption would be feasible), we must have: σ ρσ for
a stable
solution, the country’s growth rate has to exceed world
rate!
d) Unrealistic feature of this thought experiment: Assumption of
a constant interest and growth rate despite high growth economy.
More realistic: assume that growth rate converges to world growth
rate! (Otherwise small country assumption will be violated at least
in the long run)!
52
-
1.2.3.3.3 Mechanics of Accounting: Sustainability Analysis of
Current Account Deficits - The case of nominal assets and inflation
Current account: equivalent to an increase in foreign assets
)()( tBtCA &= with BiAYtCA +−=)( π;y growth rate of real
output and inflation
Growth of nominal BIP: π+= yYY&
byY
CAbYY
YBb )( π+−=−=
&&&
Sustainability Simulation: what would happen if CA/Y is
sustained for constant π;y ? Will b (ratio of net foreign assets to
GDP) converge to a steady state ratio ( )? 0=b&
Differential equation: ππ
π
++
+−= +−
yYCAe
yYCAbtb ty /)/()( )(0
YCA
yb
π+=∞
1)(
YBb = is constant if BytB )()( π+=&
Example: 05,0/;05,0 −==+ YCAy π b* = - 100 % With constant b,
real net external asset stock evolves according to:
PBy
PB
PB
tdPBd
=−= π&)/(
Steady state share of absorption A= C+I+G or a=A/Y: BiAYtBtCA
+−== )()( &
PBy
PBr
PA
PY
tdPBd !)/(
=+−= ; byra )(1 −−=−
53
-
If there is a critical lower bound aa > , yrabb
−−
−=>1 (present
value of excess production over minimum absorption level)
Obstfeld/Rogoff: Perspectives on OECD Economic Integration:
Implications for U.S. Current Account Adjustment What is the long
run equilibrium? 0)( =tCA ? Balanced current account? Note:
Nominally balanced current account implies (when inflation is
positive and with net foreign debt) an inflation adjusted current
account surplus! Rate of change of the economy’s real net foreign
asset position with =0 is: )()( tBtCA &=
PB
tdPBd π−=)/( >0 for B
-
Assume that absorption (trade balance) is such that it repays a
small fraction of the interest burden. The rest will be rolled over
into the next period.
BrAYTB ξ=−= , so BrBrAYCAB )1( ξ−=+−==& Thus, treBtB
)1(0)(
ξ−= Transversality condition holds:
000001)( Br
BrdteBrdtetTB trtr −=−=∫=∫ −∞−∞ ξξξ ξ
But notice that debt (and thus interest burden) grows at
rate
)1( ξ−r In order for repayments not to exceed total production
at some stage BrBrAY )1( ξ−=+−
)1(
, Y must grow at a rate above
ξ−> ry . (with y1ξ
55
-
1.2.4 Stochastic Optimisation 1.2.4.1 Bonds and market discount
rates How to extend market discount factor to stochastic problems?
Wealth constraint per period: If I plan to have wealth Wt+1 at the
beginning of next period t+1, I must spend today .
is the market discount factor between t and t+1. 11, ++ ttt
Wq
1, +ttqSince tttttt WCYWq +−≤++ 11, , the wealth constraint per
period is:
ttttttttt WqCYqWW )1/1()(/1 1,1,1 −+−≤− +++ From the
Hamiltonian, we get the FOC:
11,11, ~
~)~~()1/1(~
−+−+ =→−−=−=∂
∂
t
tttttttt
t
t qqWH
λλλλλ
and the transversality condition 011, =++++ TtTtt Wq Two
periods: Maximise )()( 21 CUCUu β+= subject to the per period
budget constraints
111
21 1
WYr
WC +=+
+ ; 222
32 1
WYr
WC +≤+
+ ;
with end point constraints 0; 311 ≥= WWW . FOC:
a) t
ttt
t
tc
tc
rq
CuCu
+===
+ +++
11
)()(
11
1,11
λλ
ρ
b) 033,1 =Wq Transversality condition Equilibrium condition:
2211 , YCYC == Imposing equilibrium conditions, qt is determined
by:
2,111
2
11
)()(
11 q
rYuYu
c
c =+
=+ ρ
What is the impact of uncertainty on consumption/
equilibrium?
56
-
1.2.4.2 Stochastic Discount Rates Simple example: 2 Periods;
uncertainty about the outcome of the second period with 2 possible
states: high or low income. Arrow Debreu economy: Assume a complete
set of risk markets – there is a price defined for each state of
the world.
hp probability of high endowment; otherwise =1- lp hp
Representative agents maximises expected utility:
[ ])()()( 221 llhh CUpCUpCUu ++= β Intertemporal buget
constraint with Arrow Debreu securities: At present value price ,
buy assets in period 1 giving a real claim of 1 unit in state s
next period:
sq2
With claims, wealth in state s is sD2 ss DW 22 = . So )( 222 sss
DYC +=
122122 YYqCCq sss
sss
+=+ ∑∑ or )( 1122 CYDq sss
−=∑
FOC:
'2
2
'2'2'
22
/)(/)(
s
s
sss
sss
qq
CCUpCCUp
=∂∂
∂∂
sss
s qCCUCCU
p 211
22
/)(/)(
=∂∂
∂∂β
Define s
ss p
qQ 22 = as stochastic discount factor
),();(),( 11,111, +++++ = ttssttttss zzqzzprobzzQ
ss
sss Qpq
CCUCCU
22
11
22
/)(/)(
==∂∂
∂∂β
With complete markets the stochastic discount factor is unique.
sQ2
57
-
How does uncertainty affect the price of a risk free asset (the
current value price for one unit of consumption in all states)? The
stochastic discount factor for each state is characterized by:
⎟⎟⎠
⎞⎜⎜⎝
⎛= ++++ );(
);(),( 1111,ttc
ttcttss cu
cuzzQξξβ
The risk free rate between some state s and s+1 is:
== ++
+++ ∑ 11
111, ),(( sss
sttss QpzzQE ⎟⎟⎠
⎞⎜⎜⎝
⎛=
+++
);();(
11 11
ttc
ttct
t cucuE
r ξξβ
Explicit analytical solutions can be obtained only for some
examples with special preferences, combined with specific
distributions.
58
-
Example 1) Assume (1) CRRA and (2) consumption growth is log
normally distributed. In that case, we had without risk:
σ
ρ
1
1
11
11
−+
⎟⎟⎠
⎞⎜⎜⎝
⎛+
=+ t
t
t CC
ror log-linearized: tt gr σ
ρ 1+=
When future consumption is risky, the Euler equation in expected
terms is:
σ
ρ
1
1
11
11
−
+⎟⎟⎠
⎞⎜⎜⎝
⎛+
=+ t
t
t CCE
r
In the presence of consumption risk, a safe asset provides
insurance for risky future consumption: risk averse agents want to
do precautionary saving. From Jensen’s inequality, we guess that
the safe return will be lower than a risky consumption profile with
same expected value (incentive to save for precaution drives down
the sure rate of interest). Define ttt CCC lnlnln 11 −=Δ ++ .
Taking logs, we can rewrite
the FOC condition: ( )1ln1
11ln +Δ−=+
+t
tCE
r σρ .
If )),((~ 2xxENx σ , then ( )221)(exp))((exp xxExE σ+= .
Assuming that consumption growth is log normally
distributed, here we have: 1ln1
+Δ−= tCx σ, so we get:
( ) ⎟⎠⎞
⎜⎝⎛ +Δ−=
++
+2
221
11ln1exp
11
Ctt
CEr
σσσ
ρ
Or, log linearising, we get:
( ) 222111ln1 Ctt CEr σσσ
ρ −Δ+= + = ( ) 222111
CCgE σσσρ −+
Precautionary saving With risky consumption people save more in
order to insure against risk of being in a state with low
consumption, driving down the risk-free interest rate. The more
risk averse people are (the lower σ ), the stronger this
effect!
59
-
Example: 2) Quadratric utility function (certainty
equivalence):
221 )( tCBU −−= gives:
[ ]tt
t CBrCEB −
++
=− + 11)( 1
ρ
Br
CCEt
tt ⎥⎦
⎤⎢⎣
⎡−
++
−=+ 111)( 1
ρ ;
for ρ=tr : tt CCE =+ )( 1 or 11 ++ += ttt CC ε With 0)( 1 =+tE ε
- consumption follows a random walk Example 3) Exponential utility
function:
)(exp1)( tt CCU αα−−=
This gives the Euler equation:
( ))(exp11)(exp 1+−+
+=− t
tt CE
rC αρ
α
Note: For )),((~ 2xxENx σ ( )221)(exp))((exp xxExE σ+= . So we
get:
⎥⎦
⎤⎢⎣
⎡+−
++
=− ++ )(2)(exp
11)(exp 1
2
1 ttt
t CVarCErC ααρ
α
or in logs: )(2
)()( 12
1 ++ +−−=− tttt CVarCErCααρα , so:
)(2
)(])([ 12
1 ++ +−=− tttt CVarrCCEαρα
Calculate equilibrium for the stochastic endowment process: 11
++ += ttt YY ε (persistent effect of shocks); ),0(~
21 σε Nt+
60
-
1.2.4.3 Stochastic Returns Let us now extend the analysis to
holding arbitrary assets with random return 2,~ir . Again, assume
complete contingent prices. Using the stochastic discount factor
sQ2
~ the MRS between consumption in state s and s+1 is
characterised by:
11
2
((
CC 2
2 /)/)
CUCU
Q sss ∂∂∂∂
= β . So the Euler equation gives:
A) ( )sissi QrECCUCCU
rE 22,11
222,
~)~1(/)(/)(
)~1(1 +=⎟⎟⎠
⎞⎜⎜⎝
⎛∂∂
∂∂+= β
Rewrite condition A) as the product of expected variables plus
the covariance A+) ( ) ( )sisi QrCovQErE 22,22, ~;~1)~(~11 +++= .
For a risk-less asset with safe return r, the covariance with the
stochastic discount factor (and any other random variable) is zero.
Thus:
B) t
ttss rzzQE
+=++ 1
1),(~( 11,
Using B) in A), we have for arbitrary risky assets: C) ( ) ( )(
)sii QrCovrrE 22,12, ~;~11)1(~1 +−+=+ Equation C) is the key
relation for the consumption CAPM Model: The expected return
premium of an asset relative to risk-free assets is negatively
[positively] proportional to its covariance with MRS [consumption].
Correlation (covariance) between the asset’s payoff and consumption
as decisive criterion for holding the asset: An asset with high
returns is attractive when consumption is low ( is high). → Demand
for such an asset will be high; its return will be lower than a
risk free asset.
sQ2
61
-
Applications: 1) Evaluation of risky shares (dividend
streams)
Without risk, the value of the firm is ∑∞== 10
0 tt
tdV λλ
The dividend stream is evaluated with )(')('
0,0 Cu
Cuq tto
tt βλ
λ== .
With )1(
1...)1(
1)1(
1)1(
1
1,1,20,11
1,,0
−=
− +++=
+= ∏
tt
ti
iit rrrr
q
How to evaluate a risky dividend stream? In case of risk
neutrality, utility is linear in consumption, so
1)(')('
0=
CuCu t . Thus t
o
tt Eq βλ
λ=⎟⎟
⎠
⎞⎜⎜⎝
⎛=,0
Evaluation equation: ( ) )(1,00 ∑∞=== t tttt dEdqEV β With risk
aversion, expected market discount factor depends on the covariance
between dividends and the discount factor. To see this write the
Euler equation as:
( ) ⎟⎟⎠
⎞⎜⎜⎝
⎛+++⎟⎟
⎠
⎞⎜⎜⎝
⎛
=⎟⎟⎠
⎞⎜⎜⎝
⎛+=
+++
+++
+++
][)(')('][
)(')('
][)(')('
111
111
111
ttt
ttt
t
t
ttt
tt
VdCu
CuCovVdECu
CuE
VdCu
CuEV
ββ
β
If dividends are positively correlated with the discount factor,
the price of the claim, Vt, is higher (a positive correlation with
the discount factor means a negative correlation with consumption
growth). Hence, if the dividends are negatively correlated with the
cycle, Vt, the price of the shares, is higher, since the claim on
the firm provides a good hedge against the variability of
consumption.
62
-
2) Lognormal distribution and homoskedasticity Characterise the
exact solution for lognormal distributions of the random variable
If )),((~ 2xxENx σ , then ( )221)(exp))((exp xxExE σ+= or
( )221)(log xx σ+))((exp ExE = So if Y is log normally
distributed, (that is, X= log Y is normally distributed), (
))(log)(log)(log 221 YYEYE σ+= If Y is homoskedastic, then
)](log[log)](log[log)(log 2 YEYVarYEYEYVar −=−= Logarithmic
Version: Taking the log of A) ( )si QrE 22, )~1(1 += , we get:
Log A) ( )
)]log;~(2)(log)~([21
)(log~0
22,22,
22,
sisi
si
QrCovQVarrVar
QErE
+++
+=
For the riskless asset r with Var (r) =0 and )log;( 2 sQrCov =0,
the log-version of B) is:
Log B) )](log[21)(log 22 ss QVarQEr −−=
For risky assets, the log-version of C) is:
Log C) ( ) ( ) ( )siii QrCovrVarrrE 22,2,12, log;~~21~ −−=−
63
-
Explicit solution for CRS preferences
With CRS preferences, σ1
1
2
11
22
/)(/)( −
⎟⎟⎠
⎞⎜⎜⎝
⎛=
∂∂
∂∂
CC
CCUCCU sss , the
stochastic discount factor is: σβ1
1
22
−
⎟⎟⎠
⎞⎜⎜⎝
⎛=
CC
Q ss . The inverse of σ is
the coefficient of relative risk aversion. The lower σ, the
higher the risk aversion. Defining 1+g as the gross growth rate of
consumption
, we get in A+): ttst CgC )1( 11 ++ +=
( ) ⎟⎟⎠
⎞⎜⎜⎝
⎛+++++=
−−σσβ1
2,
1
2, )~1();~1()~1()~1(1 grCovgErE ii
Log B) gives for the risk-free rate: )]([1
21)(1 121 ++ −+= tt gVargEr σσ
ρ
Log C) gives [ ] ),(1 criskyfreeriskriskyt grCovrrE σ≅−−
For these preferences, the Euler-equation gives a linear
relation between expected growth of consumption and the expected
return of any asset:
)]([121)()( 11 ++ +−= tt gVarrgE σ
ρσ or
1,1)( ++ +=Δ tWtttt rEcE σμ With tμ determined by variance and
covariance between consumption and the asset’s return. If both Wt
rE and μ are constant across time, growth of consumption follows a
random walk.
64
Advanced Macro and MoneyLecture notes (WS 09/10)Prof. Dr.
Gerhard Illing, LMU University of Munich1.1 Basic Tools: The
Problem of Optimal Control – An intuitive Guide resp. CS: or 1.1.7
Example: Optimal extraction path for exhaustible resources