Beruflich Dokumente
Kultur Dokumente
perturbation methods
Joe Steinberg
August 8, 2011
Overview
Each country has one Lucas tree. Bear same kind of fruit.
Endowment process
_
log y
t
log y
t
_
= N
_
log y
t1
log y
t1
_
+
_
t
_
t=0
to maximize lifetime
utility E
0
t=0
t
u(c
t
).
h,t
+p
f ,t
= (1)y
t
+
h,t1
(p
t
+y
t
) +
f ,t1
(p
t
+y
t
)
t
+p
h,t
+p
t
f ,t
= (1)y
t
+
h,t1
(p
t
+y
t
) +
f ,t1
(p
t
+y
t
)
Equilibrium
(c
t
) = E
t
_
u
(c
t+1
)(p
t+1
+ y
t+1
)
t
u
(c
t
) = E
t
_
u
(c
t+1
)(p
t+1
+ y
t+1
)
p
t
u
(c
t
) = E
t
_
u
(c
t+1
)(p
t+1
+ y
t+1
)
t
u
(c
t
) = E
t
_
u
(c
t+1
)(p
t+1
+ y
t+1
)
Market clearing:
c
t
+ c
t
+ y
t
+ y
t
,
h,t
+
f ,t
= 1,
f ,t
+
h,t
= 1
Accounting change
h,t
+ p
f ,t
= y
t
+
h,t1
(p
t
+ y
t
) +
f ,t1
(p
t
+ y
t
)
c
t
+ p
h,t
+ p
t
f ,t
= y
t
+
h,t1
(p
t
+ y
t
) +
f ,t1
(p
t
+ y
t
)
Change of variables
Let
h,t
and
f ,t
denote value of domestic agents holdings of
domestic and foreign stock at end of period t:
h,t
= p
t
h,t
,
f ,t
= p
f ,t
h,t
= p
h,t
,
f ,t
= p
t
f ,t
Market clearing:
h,t
+
f ,t
= 0,
f ,t
+
h,t
= 0.
Let w
t
=
h,t
+
f ,t
and w
t
=
h,t
+
f ,t
denote total wealth at
end of period t.
Market clearing: w
t
+ w
t
= 0.
Reframe maximization problem
t
) and value of domestic shares
h,t
(
f ,t
).
t
)
h,t1
+ R
t
w
t1
(1)
c
t
+ w
t
= (R
t
R
t
)
f ,t1
+ R
t
w
t1
(2)
Stock returns
R
t
=
p
t
+ y
t
p
t
R
t
=
p
t
+ y
t
p
t
Equilibrium
(c
t
) = E
t
_
u
(c
t+1
)R
(3)
u
(c
t
) = E
t
_
u
(c
t+1
)R
(4)
E
t
_
u
(c
t+1
)R
t
= E
t
_
u
(c
t+1
)R
(5)
E
t
_
u
(c
t+1
)R
t
= E
t
_
u
(c
t+1
)R
(6)
Market clearing:
w
t
+ w
t
= 0,
h,t
+
f ,t
= 0
First-order local approximation
Two steps:
1. Find the non-stochastic steady state. Trivial in most models, dicult
task here.
2. Approximate the models dynamics using a Taylor expansion of the
equilibrium conditions around the steady state. Generally pretty easy,
very dicult task here.
Why is the non-stochastic steady state nontrivial?
Normally, we just back out the steady state values from the
non-stochastic versions of the equilibrium conditions.
R =
R
R =
R
f
are equally good.
R
t
] = E
t
[
t
]
E
t
[
R
t
] = E
t
[
t
]
Inuition: Both assets have same expected return. Thats all that
matters in a rst-order model, so all feasible values of
h,t
and
f ,t
are equally good.
Outline of strategy:
1. Find a second-order condition that is sucient to tie down steady
state.
2. Exploit some properties of this equation that allow us to use a
rst-order approximation to the non-portfolio model to evaluate
sucient condition.
3. Find a closed-form solution for the steady state.
Second-order approximation of portfolio choice equations
Dene
R
x,t
=
R
t
R
t
. Assume u(c) = c
1
/1 .
Home country:
E
t
_
R
x,t+1
+
1
2
(
R
2
t+1
R
2
t+1
) c
t+1
R
x,t+1
_
= 0
Foreign country:
E
t
_
R
x,t+1
+
1
2
(
R
2
t+1
R
2
t+1
) c
t+1
R
x,t+1
_
= 0
Add together:
E
t
_
( c
t+1
c
t+1
)
R
x,t+1
_
= 0 (7)
R
x,t
+
1
w
t1
+ y y
t
h,t
doesnt show up in any other equations. So rst-order portfolio
dynamics dont aect rst-order behavior of c
t
, w
t
and
R
x,t
.
Key facts:
1. Only need 1st-order accurate dynamics for ( c
t
, c
t
,
R
x,t
) to evaluate
sucient condition (7).
2. Only need to know
h
to solve for 1st-order accurate dynamics for
( c
t
, c
t
,
R
x,t
).
Need to nd
h
such that 1st-order dynamics of non-portfolio model
satisfy sucient condition (7).
t
,
R
x,t
), check
whether (7) is satised, update
h
as necessary.
E
t
[
R
t
] = E
t
[
t
] E
t
[
R
x,t
] = 0.
R
x,t
is iid and mean zero.
Replace
1
R
x,t
in budget constraint with temporary exogenous
iid variable
t
:
c c
t
+ w
t
=
t
+
1
w
t1
+ y y
t
t
:
R
x,t
= X
1
t
+ X
2
t
c
D
t
= D
1
t
+ D
2
t
+ D
3
z
t
where z
t
= [exo states
t
, endo states
t+1
]
. X
1
, D
1
are scalars, X
2
and D
2
are 2 1 and D
3
is #z
t
1.
Solving for
h
R
x,t
where
h
=
1
h
:
R
x,t
=
X
2
t
c
D
t
=
D
2
t
+ D
3
z
t
where
X
2
=
X
2
1 X
1
h
D
2
=
D
1
h
1 X
1
h
X
2
+ D
2
Solving for
h
D
2
t+1
+ D
3
z
t+1
)
X
2
t+1
_
= 0
Since
t+1
is iid and mean zero,
E
t
[
t+1
z
t+1
] = 0
E
t
[
t+1
t+1
] =
Reduces to
X
2
2
= 0
Solving for
h
_
D
1
1 X
1
X
2
+ D
2
_
Solve for :
=
_
X
2
D
2
X
1
D
1
X
2
X
1
X
2
D
Basic idea similar to solving for steady state but a lot more involved.
Home:
0 = E
t
R
x,t+1
+
1
2
(
R
2
t+1
R
2
t+1
) c
t+1
R
x,t+1
+
1
6
(
R
3
t+1
R
3
t+1
)
2
2
c
2
t+1
R
x,t+1
2
c
t+1
(
R
2
t+1
R
2
t+1
)
Foreign:
0 = E
t
R
x,t+1
+
1
2
(
R
2
t+1
R
2
t+1
) c
t+1
R
x,t+1
+
1
6
(
R
3
t+1
R
3
t+1
)
2
2
c
2
t+1
R
x,t+1
2
c
t+1
(
R
2
t+1
R
2
t+1
)
( c
t+1
c
t+1
)
R
x,t+1
+
2
2
( c
2
t+1
c
2
t+1
)
R
x,t+1
2
( c
t+1
c
t+1
)(
R
2
t+1
R
2
t+1
)
(8)
R
x,t
] = E
t
1
2
(
R
2
t+1
R
2
t+1
)
1
6
(
R
3
t+1
R
3
t+1
) +( c
t+1
+ c
t+1
)
R
x,t+1
2
2
( c
2
t+1
+ c
2
t+1
)
R
x,t+1
+
2
( c
t+1
+ c
t+1
)(
R
2
t+1
R
2
t+1
)
(9)
Note we only have products of two and three terms, so we only need
1st and 2nd-order accurate terms to evaluate them to 3rd-order
accuracy.
Second-order approximation of excess portfolio returns
R
x,t
in budget constraint:
h,t1
R
x,t
1
R
x,t
+
2
(
R
2
t+1
t+1
) +
1
h,t1
R
x,t
h,t1
R
x,t
is also iid and mean zero.
h,t1
R
x,t
with a temporary iid mean zero exogenous variable
t
in the
budget constraint.
Solving for
We are looking for a vector that captures the law of motion for
portfolio dynamics of the form
h,t
=
z
t+1
, where
z
t+1
= [exo states
t
, endo states
t+1
]
In other words,
h,t
is function of exogenous state in period t and
the choice of tomorrows endogenous state (made in period t).
Need to nd that satises (8) for all possible states of the world.
The tensor algebra is very messy. If you really want to see it read
the paper.
To make a long story short, we rst solve for the coecients for
2nd-order laws of motion for ( c
t
, c
t
,
R
t
,
R
t
,
R
x,t
).
Let me know if you want to apply this method and need some help.