The Implicit Function Theorem (IFT): key points

1
The solution to any economic model can be characterized as the level set
corresponding to zero of some function
1
Model: S = S(p; t ), D = D(p), S = D; p =price; t =tax;
2
f (p; t ) = S(p; t ) D(p) = 0. Level Set: {(p, t ) : f (p; t ) = 0}.
2
When you do comparative statics analysis of a problem, you are studying
the slope of the level set that characterizes the problem.
1
Intuitive comp. stat. question: change t , what happens to p?
2
Intuitive idea: t ; how much p is needed to keep on LS?
3
Math answer is the slope of LS
3
The implicit function theorem tells you
1
when this slope is well dened
2
if it is well-dened, what are the derivatives of the implicit function
4
Its an extremely powerful tool
1
explicit function p(t ) could be nasty; no closed form
E.g., : f (p; t ) = tp
15
+t
13
+p
95

p = 0; whats p(t )?
2
dont need to know p(t ) in order to know
dp(t )
dt
.
3
can compute
dp(t )
dt
from partials of f (; ).
() December 5, 2013 1 / 21
Level Sets locally well-dened diff
able
functions

0
0
0
x
x
x
g()
g()
g()
( , x)
( , x)
Case 1: x is a well-dened, dierentiable function of
Case 2: x is a well-dened, but not dierentiable function of
Case 3: Locally, x is a well-dened, dierentiable function of , except at ( , x)
Figure 1. Deriving x as a function of from level sets
() December 5, 2013 2 / 21
Implicit function theorem (single variable version)
Theorem: Given f : R
2
R
1
, f C
1
and ( , x) R
2
, if
f ( , x)
x
= 0,
nbds U

of , U
x
of x & a unique g : U

U
x
, g C
1
s.t. U

,
f (, g()) = f ( , x) i.e., (, g()) is on the level set of f through ( , x)
g

() =
f (, g())

_
f (, g())
x
to slide 4 to slide 8
Trivial Proof of the second line:
f (, x) := f (, g()) = 0
d(f , g())
d
=
f (, g())

+
f (, g())
x
dg()
d
= 0
dg()
d
=
f (,g())

f (,g())
x
Note that the following is not true: if
f ( , x)
x
= 0, nbd U

of , & and a unique

R, g C
1
s.t. U

, f (, g()) = f ( , x).
to slide 4
() December 5, 2013 3 / 21
U

and U
x
to slide 3

0
x
g()
( , x)
U

U
x
1
Note that
There are two C
1
functions g
1
and g
2
mapping U

to R s.t.
for i = 1, 2, for U

f (, g
i
()) = f ( , x).
but only one of these maps U

into some nbd of x.

p
2
f
t
= p
15
+13t
12
;
f
p
= 15tp
14
+95p
94
1/2
1

p
3
if
f
p
= 0,
dp
dt
=
_
p
15
+13t
12
__
_
15tp
14
+95p
94
1/2
1

p
_
Also true (if you are a mathematician but not an economist):
1
if
f
t
= 0,
dt
dp
=
_
15tp
14
+95p
94
1/2
1

p
__
_
p
15
+13t
12
_
() December 5, 2013 5 / 21
Implicit function theorem examples II
Example: (slightly less dramatic)
Illustrates that U

needs to be chosen carefully: cant be too big

= 2;
f (x;)
x
= 2x
3
if
f (x;)
x
= 0, then
dx
d
=
df (x;)
d
_
df (x;)
dx
=/x.
Suppose = 0.9, x =

10.81 0.44.
Use IFT to estimate effect of a 100% increase in to 1.8
dx
d
=0.9/0.44 2;
x(1.8) 0.44+(20.9) 1.36
Clearly, no x exists such that (x; +0.9) LS
0
conclude: hazardous to base policy decisions on comp stat analysis
d is unlikely to be small enough for the theorem to be applicable.
() December 5, 2013 6 / 21
Implicit function theorem examples III
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
1.5
1
0.5
0
0.5
1

x
() December 5, 2013 7 / 21
Implicit function theorem (intermediate version)
Theorem: Given f : R
n+1
R
1
, f C
1
and (

, x) R
n
R
1
, if
f ( , x)
x
= 0,
j , nbds U
j
of
j
, U
x
of x & a unique g : U
j
U
x
, g C
1
s.t.
j
U
j
,
f (, g()) = f (

, x) i.e., g puts us on the level set of f containing (

, x)
dg()
d
j
=
f (, g())

j
_
f (, g())
x
or in vector form
g() =

f (, g())/f
x
(, g()).
to slide 3
This is a straightforward extension of the previous theorem. Example:
1
Model: S = S(p; t ), D = D(p; y), S = D; t =tax, y =income;
2
f (p; t , y) = S(p; t ) D(p; y),
3
If
f
p
= 0, can use IFT to compute p = (p
t
, p
y
)
4
Can use IFT to compute p = (p
t
, p
y
)
5
Once you have gradient, can get any directional derivative
6
Use IFT to approx impact of any combination of param changes
() December 5, 2013 8 / 21
Equilibrium price as a function of tax and income
P
Q
p(t, y)
p(t

, y

)
D(p, y)
D(p, y

)
S(p, t)
S(p, t

)
1
() December 5, 2013 9 / 21
The zero level set of S(p, t ) D(p, y)
0
0.2
0.4
0.6
0.8
1
0
0.2
0.4
0.6
0.8
1
0
0.5
1
1.5
2
2.5
3
tax
Zero Level set of Supplydemand model
income
e
q
u
i
l
i
b
r
i
u
m

p
r
i
c
e
() December 5, 2013 10 / 21
Implicit function theorem (multivariate version)
to slide 13
Theorem: Given f : R
n+m
R
m
, f C
1
& (

, x) R
n
R
m
, if
det (Jf
x
(

, x)) = 0, U

of

, U
x
of x & !g : U

U
x
, g C
1
s.t. U

,
f(, g()) = f(

, x) i.e., g puts us on level set of f containing (

, x)
_

_
g
1
()

1

g
1
()

n
g
2
()

1

g
2
()

n
.
.
.
.
.
.
.
.
.
g
m
()

1

g
m
()

n
_

_
= Jf
x
(, g())
1
_

_
f
1
(,g())

1

f
1
(,g())

n
f
2
(,g())

1

f
2
(,g())

n
.
.
.
.
.
.
.
.
.
f
m
(,g())

1

f
m
(,g())

n
_

_
where Jf
x
(, g()) is the Jacobian of f treating as parameters
() December 5, 2013 11 / 21
Multivariate IFT: graphical representation
0
0
0
0
dx
1
dx
2
0
0
0
0
dx
1
dx
2
0
0
0
x1 x1
x2 x2
d

Level set of f
1
corresponding to 0 Level set of f
2
corresponding to 0
The tangent plane to the 0-level set of f
1
The tangent plane to the 0-level set of f
2
The two tangent planes combined
The two tangent planes: enlarged
1
f : R
1+2
R
2
, i.e., two endog variables x, one exog
1 x is in the horizontal plane; on vertical plane
2 ( , x) LS
1
LS
2
is a soln, i.e., belongs to both zero level sets
3 as changes, x changes to keep vector in LS
1
LS
2
4 (, x) slides along the groove in the bottom right panel
() December 5, 2013 12 / 21
Example: apply IFT to rst order conditions I
Problem
max
L,K
(L, K) = pL

wLrK Soln: (

L,

K)
In this case, the level set LS
0
we stay on is the FOC, where
FOC(w, r ; L, K) =
_

K
_
=
_
pL
1
K

w
pL

K
1
r
_
= 0
Match concepts: general expression for IFT vs this example
to slide 11
f is FOC i.e., = (
L
(

L,

K),
K
(

L,

K))
x is (

L,

K).
is (w, r ).
Jf
x
(

, x) is Jacobian of (

L,

K) w.r.t. (L, K), or H, the Hessian of
Jf

, x) is Jacobian of (

L,

K) w.r.t. (w, r ).
g() is
_

L(w, r )

K(w, r )
_
.
() December 5, 2013 13 / 21
Example: apply IFT to rst order conditions II
Plug in all the pieces...
Jf
x
(

, x) is Jacobian of (

L,

K) w.r.t. (

L,

K),
Jf
x
(

, x) = H(

L,

K) =
_

2
(

L,

K)
L
2

2
(

L,

K)
LK

2
(

L,

K)
LK

2
(

L,

K)
K
2
_
Jf

, x) is Jacobian of (

L,

K) w.r.t. (w, r ).
Jf

, x) = J
w,r
=
_

2
(

L,

K)
Lw

2
(

L,

K)
Lr

2
(

L,

K)
Kw

2
(

L,

K)
Kr
_
g() is
_

L(w, r )

K(w, r )
_
is implicitly dened but not explicitly.
Jg() = Jf

, x)Jf

, x)
J
_

L(w, r )

K(w, r )
_
=
_

2
(

L,

K)
L
2

2
(

L,

K)
LK

2
(

L,

K)
LK

2
(

L,

K)
K
2
_
1
_

2
(

L,

K)
Lw

2
(

L,

K)
Lr

2
(

L,

K)
Kw

2
(

L,

K)
Kr
_
=
_

2
(

L,

K)
L
2

2
(

L,

K)
LK

2
(

L,

K)
LK

2
(

L,

K)
K
2
_
1
_
1 0
0 1
_
() December 5, 2013 14 / 21
What some other people do
Many economists dont apply IFT directly; two differences
they totally differentiate each line of f separately
they end up a different place from where the IFT ends up
The IFT gives a formula for the Jacobian of the implicit funct g
Total differentiators end up with an expression for differential of g
Relationship between Jacobian and differential is as it always is
Recall: for every n n matrix, unique L : R
n
R
n
.
dx = Jg(; x)d
Summary:
IFT route ends up at: Jg(; x)
Total differentiation route ends up at: dx = Jg(; x)d
() December 5, 2013 15 / 21
IFT via total differentiation
max
L,K
pL

wLrK
Requirement: stay on the zero level set of f , where
f (w, r ; L, K) =
_

K
_
=
_
pL
1
K

w
pL

K
1
r
_
= 0
Totally differentiate:
0 =

L
L
dL +

L
K
dK +

L
w
dw
0 =

K
L
dL +

K
K
dK +

K
r
dr
Move exog variables to right hand side; put in matrix form:
_

L
L

L
K

K
L

K
K
_
_
dL
dK
_
=
_

L
w
0
0

K
r
_
_
dw
dr
_
=
_
1 0
0 1
__
dw
dr
_
Invert
_
dL
dK
_
=
_

L
L

L
K

K
L

K
K
_
1 _
1 0
0 1
__
dw
dr
_
() December 5, 2013 16 / 21
Apply IFT to an equilibrium system
Question: (from ARE prelim, 2008): A monopolist sells in two countries: 1 and
2. It produces a good at a constant marginal cost of c and cannot produce
more than a total of Q units. Country 1 imposes a per unit tax of units on the
good sold in that country. Assume that the constraint binds.
1
For a given value of , show how the equilibrium in the two countries are
determined.
2
Show how these equilibrium prices change as increases.
() December 5, 2013 17 / 21
Answer to ARE prelim, 2008 question
Answer: Assume that inverse demand curves are concave in price.
The monopolists optimization problem is
max
p
1
,p
2
(p
1
)D
1
(p
1
) + p
2
D
2
(p
2
) c(D
1
(p
1
) +D
2
(p
2
))
s.t. D
1
(p
1
) +D
2
(p
2
) Q
The Lagrangian is
L(p
1
, p
2
, ; ) = (p
1
)D
1
(p
1
) + p
2
D
2
(p
2
) c(D
1
(p
1
) +D
2
(p
2
))
+ (QD
1
(p
1
) D
2
(p
2
))
Assuming capacity constraint binds, the rst order conditions are
L
p
1
= D
1
(p
1
) + (p
1
c )D

1
(p
1
) = 0
L
p
2
= D
2
(p
2
) + (p
2
c )D

2
(p
2
) = 0
L

= QD
1
(p
1
) D
2
(p
2
) = 0
Solution is a triple (p

1
, p

2
,

) which solves this system, given .

)
c +

()
c +

)
1
() December 5, 2013 19 / 21
L
p
1
= D
1
(p
1
) + (p
1
c )D

1
(p
1
) = 0
L
p
2
= D
2
(p
2
) + (p
2
c )D

2
(p
2
) = 0
L

= QD
1
(p
1
) D
2
(p
2
) = 0
The Hessian of the Lagrangian w.r.t. endog vars is:
HL
p,
=
_
_
2D

1
(p

1
) +
(
p

1
c

)
D

1
(p

1
) 0 D

1
(p

1
)
0 2D

2
(p

2
) +
(
p

2
c

)
D

2
(p

2
) D

2
(p

2
)
D

1
(p

1
) D

2
(p

2
) 0
_
_
HL

=
_
_
D

1
(p

1
)
0
0
_
_
.
Hence from the implicit function theorem, we have
_

_
dp

1
d
dp

2
d
d

d
_

_
= HL
1
(p,)
_
_
D

1
(p

1
)
0
0
_
_
= HL
1
(p,)
_
_
D

1
(p

1
)
0
0
_
_
() December 5, 2013 20 / 21
Its straightforward to check that the determinant of HL
(p,)
is

_
D

2
(p

2
)
2
_
2D

1
(p

1
) + (p

1
c

)D

1
(p

1
)
_
+D

1
(p

1
)
2
_
2D

2
(p

2
) + (p

2
c

)D

2
(p

2
)
_
_
> 0
Applying Cramers rule, we have
dp

1
d
= det
_
_
_
_
_

_
D

1
(p

1
) 0 D

1
(p

1
)
0 2D

2
(p

2
) + (p

2
c

)D

2
(p

2
) D

2
(p

2
)
0 D

2
(p

2
) 0
_

_
_
_
_
_
_
det(HL
(p,)
)
= D

1
(p

1
)D

2
(p

2
)
2
_
det(HL
(p,)
) > 0
dp

2
d
= det
_
_
_
_
_

_
2D

1
(p

1
) + (p

1
c

)D

1
(p

1
) D

1
(p

1
) D

1
(p

1
)
0 0 D

2
(p

1
)
D

1
(p

2
) 0 0
_

_
_
_
_
_
_
det(HL
(p,)
)
= D

2
(p

1
)D

1
(p

2
)
2
_
det(HL
(p,)
) < 0
d

d
= det
_
_
_
_
_

_
2D

1
(p

1
) + (p

1
c

)D

1
(p

1
) 0 D

1
(p

1
)
0 2D

2
(p

2
) + (p

2
c

)D

2
(p

2
) 0
D

1
(p

1
) D

2
(p

2
) 0
_

_
_
_
_
_
_
det(HL
(p,)
)
=
_
2D

2
(p

2
) + (p

2
c

)D

2
(p

2
)
_
D

1
(p

2
)
2
_
det(HL
(p,)
) < 0
() December 5, 2013 21 / 21